A systematic review of evidence on employment transitions and weight change by gender in ageing populations.

BACKGROUND: Becoming unemployed is associated with poorer health, including weight gain. Middle- and older-age adults are a growing segment of workforces globally, but they are also more vulnerable to changes to employment status, especially during economic shocks. Expected workforce exits over the next decade may exacerbate both the obesity epidemic and the economic burden of obesity. This review extends current knowledge on economic correlates of health to assess whether employment transitions impact body weight by sex/gender among middle-aged and older adults.
METHODS: Eight bibliometric databases were searched between June and July 2021, supplemented by hand-searches, with no restriction on publication date or country. Longitudinal studies, or reviews, were eligible when examining body weight as a function of employment status change in adults ≥50 years. Data extraction and quality appraisal used predefined criteria; reported findings were analysed by narrative synthesis.
RESULTS: We screened 6,001 unique abstracts and identified 12 articles that met inclusion criteria. All studies examined retirement; of which two also examined job-loss. Overall, studies showed that retirement led to weight gain or no difference in weight change compared to non-retirees; however, reported effects were not consistent for either women or men across studies or for both women and men within a study. Reported effects also differed by occupation: weight gain was more commonly observed among retirees from physical occupations but not among retirees from sedentary occupations. Few studies assessed the role of health behaviours; sleep was the least studied. Most studies were medium quality.
CONCLUSIONS: Existing studies do not provide a clear enough picture of how employment transitions affect body weight. Firm conclusions on the impact of employment transitions on weight cannot be made without further high-quality evidence that considers the role of gender, job-type, other health behaviours, and other transitions, like job-loss.

Introduction

Work confers to individuals a sense of identity and coherence [1], social networks [2], and a stream of income that provides stability in meeting one’s material needs [2,3]. It also represents a significant amount of time spent away from other activities. In older adults, work also serves as a way of staying socially, mentally, and physically active and to continue contributing to the community and the broader economy [3,4]. Across OECD countries, the average share of the workforce aged 45 to 59 is expected to grow from 26% (1995) to 34% by 2030, and the share of workers aged 60 and above is projected to increase from 4% to 9% [5]. Coupled with the rise in labour force participation of middle-aged and older workers are unique challenges such as lower rates of enrolment into continued education and skills training [5], employer preference for hiring younger employees [5], and ageism at work [5,6]. Thus, more middle-aged and older workers may find themselves experiencing job-loss and challenges with re-employment following economic shocks, as observed following the Great Recession in the late 2000s and currently with the COVID-19 pandemic [7,8].

Beyond the direct financial and economic consequences of employment transitions, middle-aged and older workers who experience later-life career loss are also likely to experience worsening health [9]. In particular, involuntary job-loss in middle-aged and older adults is associated with declines in self-reported physical functioning [9], and increased risk of stroke and cardiovascular disease [10,11]. One of the risk factors for cardiovascular disease that has received some attention in the employment literature is body weight change [12-16], and how it differs following different changes in employment status. Changes in employment status, or employment transitions, span a range of events from exiting the labour force through job-loss and retirement to re-entry into the labour force by re-employment. Loss of a job without compensation has been found to be associated with both decreases and increases in body weight;[12,15] while voluntary and planned retirement has been found to be associated with weight gain [16]. However, these results have not been consistently supported in the literature [17,18]. It has been noted that change in body weight may depend on gender [12], occupation type [14], and changes in physical activity levels following a transition [14].

One additional post-transition behaviour change that may also be important in the effect of employment transitions on body weight is sleep [19-25]. For older adults (age 65 and older), the National Sleep Foundation recommends between 7 and 8 hours of sleep per night [26]. Sleep durations that are shorter and longer than recommended ranges are associated with increased risk of incident obesity, cardiovascular disease, type 2 diabetes, and all-cause mortality [27].

In a cohort of public employees in the U.S. [19], individuals who transitioned from full-time employment to full retirement had bedtimes and wake times that were 30 minutes and 63 minutes later in their first year post-retirement, respectively [19]. These delays increased to 36 and 78 minutes respectively by the third year [19]. The later bedtime and much later wake time led to a significant increase in overall sleep duration during weekday nights after retirement [19]. This increase in overall sleep duration may also be coupled with improvements in sleep quality [25] which may be protective rather than harmful, especially among those who were at higher risk of sleep disturbances due to job demand, fatigue, and poor mental health initially [25]. Additionally, in a study of Finnish public employees [20], a decline in sleep difficulties such as non-restorative sleep and waking up too early was observed in the period immediately following retirement [20]. Aside from retirement, the experience of job-loss is also associated with differential sleep outcomes. In one longitudinal study of aggregated employment data in the U.S. [28], those who experience periods of unemployment were more likely to have shorter and longer sleep durations compared to those who were employed continuously. Those who experienced unemployment were also more likely to report sleep disturbances and a greater number of nights of insufficient sleep [28].

In turn, poor sleep quality and both too much and too little sleep have been associated with increases in body weight in older adults. A literature review published in 2018 reviewed the existing evidence of sleep parameters in association with obesity in older adults [29]. Included longitudinal studies suggested that both shorter and longer sleep durations were associated with increased risk of weight gain,[30-32] and increased odds of obesity compared to those with recommended sleep durations in the order of 2.3-fold among women and 3.7-fold among men [33].

In sum, there is a potential mediating role that sleep may play in the impact of employment transitions on subsequent body weight gain. Sleep may contribute to explaining the heterogeneity that appears to exist in the literature on job-loss and retirement and body weight gain in older adults, yet it has not received as much attention as physical activity [14,18], or other modifiable health behaviours (e.g., alcohol consumption and smoking) [15,16]. To address this gap, this review has two objectives: (1) to systematically search the existing literature on the impact that employment transitions have on body weight by women and men among middle-aged and older adults, and (2) to examine whether sleep confounds, mediates, or modifies the effect that employment transitions have on body weight in the identified studies.

Methods

Search and data sources

Journal articles were systematically searched using eight bibliometric databases (Ovid MEDLINE/PubMed, Scopus, PsycINFO, Web of Science, Embase, EconLit, CINAHL, and Applied Social Sciences Index and Abstracts). Table 1 contains the search terms used in the databases after consulting with a reference librarian. Free-text thesaurus and MeSH terms were used with Boolean operators to capture the concepts of “middle-age and older”, “employment”, “transition”, and “body weight”. No restrictions were placed on publication date, country, or original publication language. Screening, assessment, and inclusion were limited to studies published in English, French, and Chinese. Searches were performed by ACTT, VAS, SJ, and TYL between June and July 2021. Search strategies in each of the databases are described in supplementary material (S1 Table). This systematic review was not registered.

Table 1

Search terms used in literature search.

Concept	Search terms (“/” indicating “OR”)
Middle-age and older adults	old* adult*/ ag?ing / aged / elder* / geriatric* / senior
Employment	employ* / job* / unemploy* / work* / retir*
Transition	change / loss / transition / terminat* / dismiss* / lay-off / reduc* / becom* / enter* / adjust*
Weight	weight / bmi / body mass index / adipos* / “met* syndrome” / “cardiovascular disease”

* indicates a search of the provided root and any ending.

? indicates a wildcard replacement of zero or one character. Variation in wildcard symbols were accounted for and the appropriate alternative symbols were used (e.g., $ or #) on a per database basis.

Screening

Two reviewers (ACTT and VAS, SJ, or TYL) screened titles and abstracts independently. Records were removed based on exclusion criteria and eligible full-texts were retrieved and screened for inclusion and reference-tracing adhering to the same division of work between two independent reviewers. All disagreements were managed with discussion among reviewers and resolved by consensus.

Criteria

A study was eligible for inclusion if it examined the longitudinal association of employment transition (e.g., a change in employment status between two time-points) and subsequent body weight outcomes in adults aged 50 years of age and older. Studies were excluded if the exposure of interest was employment status at only single point in time as the objective of this review is to examine how changes to employment status may be associated with body weight outcomes. Examples of changes in employment status are job-loss (employed at baseline of study and unemployed at follow-up) and retirement (employed at baseline of study and retired and not working at follow-up). No restrictions were placed on how studies defined the transitions or body weight outcomes. Additional exclusion criteria include: cross-sectional design; qualitative studies; broad age groups with no stratification of results (e.g., 18 years to 65 years); non-body weight outcomes; and, clinical populations. Studies that also included an analysis with sleep parameters were considered for our second objective; however, the absence of sleep as a variable in the analysis was not a criterion for exclusion. A summary of the criteria used are in Table 2.

Table 2

Inclusion/exclusion criteria used.

	Inclusion criteria	Exclusion criteria
Population	Adults in their middle-age or older	Adults in early adulthood; broad age groups, unless results are stratified; clinical populations
Exposure	Change in employment status	Only static employment status (e.g., baseline employment as exposure)
Outcome(s)	Change in self-reported or objectively measured body weight	None
Types of studies	Longitudinal studies (RCTs, cohort, panel/ecological, case-control)	Cross-sectional studies, qualitative studies, editorials, and validation studies
Setting	Any settings	None
Publication year	Any year	None
Publication language	English, French, and Chinese	Not published in English, French, and Chinese

Risk of bias assessment, data extraction, and analysis

Studies were appraised using the Effective Public Health Practice Project (EPHPP) tool given its suitability for assessing quantitative studies based on observational data [34]. The EPHPP quality assessment tool contains 21-items with sections that cover components such as selection bias, study design, confounding, data collection, and analysis considerations. Of the 21 items, 11 are directly relevant in assigning a global rating of “strong”, “moderate”, or “weak” to our included studies. The remaining 10 items from the tool are not relevant to the studies that were included in our systematic review and thus, had no bearing on the scoring assignment. For example, the items include questions related to randomized controlled trial design, the integrity of the intervention received, and intention-to-treat analytic approaches [34]. ACTT and VAS independently assessed study quality and initial assessments were compared, with final ratings established by consensus.

VAS, SJ, and TYL extracted data from studies into a standardised evidence table: author; objectives; study time frame; setting; research design; sample characteristics; description of exposure; description of outcome; time to follow-up; job-types specified; and, body weight results. Where relevant, gender and job-type stratified results are also populated into the results field. ACTT reviewed the extracted data and provided feedback to the three extractors during the process. A narrative synthesis was used to summarize the findings. The narrative synthesis revealed additional health behaviours beyond sleep that were studied as mechanisms; thus, a post-hoc analysis was conducted to summarise additional results.

Results

The results of the database searches are presented in a flow diagram in Fig 1. Database searches and citation tracing identified 6,001 unique records after duplicates were removed (n = 1,754). In total, 27 original research articles remained after title and abstract screening and were read in full. Of the 27, 12 were eligible for inclusion and were assessed for risk of bias, extracted from, and synthesized.

Fig 1

Modified PRISMA flow diagram of literature search and study selection [35].

Fifteen studies were excluded after full-text review. Ten studies were excluded because results were not relevant to the outcomes of interest (body weight) [36-45]. Three studies used employment status at a single time point and thus, did not measure the exposure of interest (employment transition) [46-48]. The two remaining studies investigated body weight as a determinant of retirement decisions and did not have results relevant to the desired direction of association [49]; or was a protocol document [50].

Characteristics of included studies

The characteristics of the included studies are presented in Table 3. All 12 included studies analyzed the transition from employment into retirement, while only two additionally considered transitions from employment into unemployment (job-loss) [12,16]. Analytic samples ranged in sizes from 288 to 11,168 respondents [18,51]. Eight of the studies used change in body mass index (BMI) between time points as their body weight outcome [14,16,51-56]. Additional outcomes included odds of weight change [17,57], odds or risk of overweight or obesity [53,56], and annualized weight change [12]. Measured body size data was used in five studies [12,16,18,51,54], while the other seven used self-reported measures [14,17,52,53,55-57]. Over half (n = 7) of the included studies utilized cohort data [14,16-18,52,53,55], while four used panel data [51,54,56,57], and one study used both cohort and panel data [12]. Time to follow-up also varied from 2 years [56] to 10 years [51]. Two studies restricted their study samples to men [16,18], and three studies did not report results by sex/gender [52,55,57].

Table 3

Characteristics of included studies.

Author / source	Stated study objective	Years	Setting	Study design, duration (data)	Study population (n)	Description of exposure	Outcome measures
Morris et al. 1992 [16][ProQuest]	To assess the effect of unemployment and early retirement on cigarette smoking, alcohol consumption, and body weight in a group of middle-aged British men	1978–1980; 1983–1985	UK,primary care	Prospective cohort5 y follow-up(British Regional Heart Study)	Men (40−59 y)(n = 6,191)	Baseline employment status and reasons for subsequent change were used to categorize men into transitions:(1) Continuously employed(2) Discontinuously employed(3) Unemployed due to illness(4) Retired due to illness(5) Unemployed not due to illness(6) Retired not due to illness	Anthropometric measures: weight (Kg), height (cm)Self-reported body mass index (BMI)% change in body weightSmoking (never, light, moderate, heavy),alcohol consumption (non-drinker, occasional, light, moderate, heavy),physical activity (PA) (regular, recreational, sporting)
Nooyens et al. 2005 [18][MEDLINE]	To evaluate the impact of retirement on diet, physical activity, body mass index (BMI) and waist circumference over a 5-year follow-up period in a population-based cohort	1987–1991; 1993–1997; 1998–2002	Town of Doetinchem, The Netherlands, at the municipal health centre	Prospective cohort5 y follow-up	Men (50−65 y)(n = 288)	Retired: respondent indicated they are retired at follow-upContinue working: respondent indicated they are still working at follow-up	Anthropometric measures: weight (Kg), height (cm), waist circumference (cm)Leisure-time PA (hrs/wk from EPIC physical activity questionnaire), diet (portion and frequency of items from EPIC Food Frequency Questionnaire)
Forman-Hoffman et al. 2008 [17][MEDLINE]	To determine whether retirement is associated with either weight gain or weight loss	1994–2002	USA	Prospective cohort2 y follow-up(Health & Retirement Study)	Adults (53−63 y)(n = 10,150)	Continuing to work: respondents reported working at least 20 hrs/month in adjacent interviews (e.g. 1994 and 1996)Recently retired: respondents reported currently working at the time of the first interview but reported retiring in the previous 2 years at the time of the next biennial interview	% change in self-reported BMIVigorous PA (started, stopped, no change)
Zheng 2008 [14][ProQuest]	To investigate the long-term effects of physical activity on body weight and the effect of food price on body weight	1992–2006	USA	Prospective cohort2 y follow-up(Health & Retirement Study)	Men (51−61 y)Women (50−61 y)(n = 3,936)	Retired: respondents indicated they are retired as opposed to working or unemployedNot retired: respondents indicated they were working or unemployed	Self-reported BMIVigorous PA (≥3 times/week, <3 times/week)
Chung et al. 2009 [52][MEDLINE]	To investigate the effect of retirement on the change in body mass index (BMI) in diverse groups varying by wealth status and occupation type	1992–2002	USA	Prospective cohort2 y follow-up(Health & Retirement Study)	Adults (50−71 y)(n = 10,565)	Retired: respondents who at the time of the interview were not working for payCurrently working: respondents who at the time of the interview were working for pay	Self-reported BMI
Gueorguieva et al. 2011 [55][Hand search]	To determine whether retirement influences BMI patterns according to occupational affiliation	1992–2002	USA	Prospective cohort2 y follow-up(Health & Retirement Study)	Adults (50+ y)(n = 2,096)	Retired: respondents indicated they were working during their initial interview and subsequently indicated they had retired during follow-up surveys	Self-reported BMIAnnualized change in BMI
Monsivais et al. 2015 [12][MEDLINE]	To examine the association between employment change and weight change among three groups: Those who maintained employment over the follow-up period, those who retired and those who became unemployed	British Household Panel Survey:2004–2005; 2006–2007EPIC-Norfolk: 1993–1997; 1998–2000	UK	Longitudinal2.2 y follow-up(British Household Panel Survey)3.5 y follow-up(EPIC-Norfolk cohort)	Adults (18+ y), BHPS(n = 4,539)Adults (39−75 y), EPIC-Norfolk(n = 7,201)	Remained employed:BHPS: respondent reporting being employed at follow-upEPIC-Norfolk: respondent reported working at follow-upRetired:BHPS: respondent reported being retired at follow-upEPIC-Norfolk: respondent reported being retired and either working or not working at follow-upLost job:BHPS: respondent reported being unemployed at follow-upEPIC-Norfolk: respondent reported being unemployed; unemployed and retired; not retired or unemployed but out of work for other reasons	Anthropometry: weight (Kg), height (cm) (EPIC-Norfolk)Annualized change in body weightSleep loss due to worry (4 categories to GHQ12: “Have you recently lost sleep over worry?”)
Godard 2016 [56][Hand search]	To estimate the causal impact of retirement on BMI, on the probability of overweight or obesity and on the probability of obesity	2004; 2006; 2010	Europe(Austria, Spain, Germany, Italy, Sweden, France, Switzerland and Belgium)	Panel2 y follow-up4 y follow-up	Employed adults (50−69 y)(n = 2,599)	Retired: respondents indicated they were currently retired and were previously currently workingContinuously employed: respondents indicated they were currently working for all waves of data	Self-reported BMIChange in obesity (kg/m2) (underweight, normal, overweight, obese)Leisure-time PA (4 categories to “How often do you engage in activities that require a moderate level of energy, such as gardening, cleaning the car, or going for walk?”)
Stenholm et al. 2017 [53][MEDLINE]	To examine changes in body mass index during years preceding retirement, during retirement transition and after retirement	Pre-retirement: 2000–2002; 2004; 2008Post-retirement: 2005, 2009, 2013	Ten towns (Tampere, Espoo,Turku, Vantaa, Oulu, Raisio, Naantali, Valkeakoski,Nokia, and Virrat),Finland	Prospective cohort4 y follow-up	Retired adults (late 50 y)(n = 5,426)	Statutory retirement: respondents indicated they fully retired at the statutory retirement age and did not report retirement due to health grounds, unemployment, or partial retirement plans	Self-reported BMIChange in obesity (kg/m2) (underweight, normal, overweight, obese)
Syse et al. 2017 [57][PsycINFO]	To assess the association between retirement and 5-year changes in health and health behaviour	2002–2007	Norway	Panel study5 y follow-up	Employed adults (57−66 y)(n = 546)	Retired: respondents who were collecting pension at T2 (2007) were considered retired voluntarily or involuntarilyStill working: respondents who reported still working at T2	Change in 1 Kg (self-reported)
Feng et al. 2020 [54][MEDLINE]	To investigate the effect of retirement on body mass index and body weight	2011–2015	China	Panel study2 y follow-up(China Health and Retirement Survey)	Retired adults (45+ y)(n = 3,090)	Retired: respondents indicated they are completely retired, do not work any hours in the labour market, and held a paid job at one point in the pastNon-retirees: respondents indicated they work in the labour market and are not retired	BMI from measured weight (Kg), height (cm)Leisure-time PA (yes, no to “during a usual week, did you do any physical exercise for at least ten minutes continuously.”), diet (food consumption and spending per capita)
Pedron et al. 2020 [51][MEDLINE]	To estimate the causal effect of retirement on a large set of biomedical and behavioral risk factors for cardiovascular and metabolic disease	Baseline: 1994–1995; 1999–2000Follow-up: 2004–2005; 2006–2008; 2013–2014	Germany	Panel study6–10 y follow-up(Cooperative Health Research in the Augsburg Region study)	Adults (45−80 y)(n = 11,168)	Retired: respondents reported their current employment status as “retired”Not retired: respondents indicated any status that was not “retired” such as employed, unemployed, homemakers, and unemployed due to long-term sickness	BMI from measured weight (Kg), height (cm)

Quality of included studies

Two studies were rated as weak [53,54]; a majority of the studies (n = 8) were considered moderate in quality [12,14,16-18,51,55,57]; and two were rated as strong [52,56] (Table 4). The reasons for the weak rating included an absence of description and count of those lost-to-follow-up or drop-outs in the data source, as well as a lack of discussion on the validity or biases of the relevant data collection tools. All studies controlled for many relevant confounders through design (stratification of sample by health status or initial body weight class) or through statistical adjustment in models, while only three studies described the validity of using self-reported body weight in their respective populations [17,52,56].

Table 4

Quality appraisal of included studies.

Section	Question number	Question	Stenholm et al. 2017 [53]	Syse et al. 2017 [57]	Zheng 2008 [14]	Chung et al. 2009 [52]	Feng et al. 2020 [54]	Forman-Hoffman et al. 2008 [17]	Pedron et al. 2020 [51]	Nooyens et al. 2005 [18]	Godard 2016 [56]	Gueorguieva et al. 2011 [55]	Morris, Cook, and Shaper 1992 [16]	Monsivais et al. 2015 [12]
A	Q1	Are the individuals selected to participate in the study likely to be representative of the target population?	1—Very likely	2—Somewhat likely	1—Very likely	1—Very likely	1—Very likely	1—Very likely	1—Very likely	2—Somewhat likely	2—Somewhat likely	1—Very likely	1—Very likely	1—Very likely
A	Q2	What percentage of selected individuals agreed to participate?	5 Can’t tell	2 60–79% agreement	5 Can’t tell	1 80–100% agreement	5 Can’t tell	1 80–100% agreement	5 Can’t tell	2 60–79% agreement	2 60–79% agreement	5 Can’t tell	2 60–79% agreement	5 Can’t tell
A	Rate this section	1—strong2—moderate3—weak	1—strong	2—moderate	1—strong	1—strong	1—strong	1—strong	1—strong	2—moderate	2—moderate	1—strong	2—moderate	1—strong
B	Q1	STUDY DESIGNIndicate the study design	5—Cohort	7—Panel	5—Cohort	3—Cohort Analytic	7—Panel	3—Cohort Analytic	7—Panel	3—Cohort Analytic	7—Panel	3—Cohort Analytic	3—Cohort Analytic	3—Cohort Analytic
B	Rate this section	1—strong2—moderate3—weak	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate
C	Q1	Were there important differences between groups prior to the intervention?	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes
C	Q2	If yes, indicate the percentage of relevant confounders that were controlled (either in the design (e.g.stratification, matching) or analysis)?	2–60–79% (some)	1–80–100% (most)	2–60–79% (some)	1–80–100% (most)	1–80–100% (most)	1–80–100% (most)	2–60–79% (some)	2–60–79% (some)	1–80–100% (most)	2–60–79% (some)	2–60–79% (some)	1–80–100% (most)
C	Rate this section	1—strong2—moderate3—weak	2—moderate	1—strong	2—moderate	1—strong	1—strong	1—strong	2—moderate	2—moderate	1—strong	2—moderate	2—moderate	1—strong
D	Q1	Was (were) the outcome assessor(s) aware of the intervention or exposure status of participants?	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes
D	Q2	Were the study participants aware of the research question?	2—No	2—No	2—No	2—No	2—No	2—No	2—No	2—No	2—No	2—No	2—No	2—No
D	Rate this section	1—strong2—moderate3—weak	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate	2—moderate
E	Q1	Were data collection tools shown to be valid?	3—Can’t tell	3—Can’t tell	3—Can’t tell	1- Yes	3—Can’t tell	1- Yes	3—Can’t tell	3—Can’t tell	1- Yes	3—Can’t tell	3—Can’t tell	3—Can’t tell
E	Q2	Were data collection tools shown to be reliable?	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell	3—Can’t tell
E	Rate this section	1—strong2—moderate3—weak	3—Weak	3—Weak	3—Weak	2—Moderate	3—Weak	2—Moderate	3—Weak	3—Weak	2—Moderate	3—Weak	3—Weak	3—Weak
F	Q1	Were withdrawals and drop-outs reported in terms of numbers and/or reasons per group?	2—No	1—Yes	1—Yes	1—Yes	3—Can’t tell	3—Can’t tell	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes	1—Yes
F	Q2	Indicate the percentage of participants completing the study. (If the percentage differs by groups, record the lowest).	4—Can’t tell	2–60–79%	1–80–100%	1–80–100%	4—Can’t tell	4—Can’t tell	2–60–79%	2–60–79%	2–60–79%	1–80–100%	1–80–100%	2–60–79%
F	Rate this section	1—strong2—moderate3—weakNot applicable	3—Weak	2—Moderate	1—Strong	1—Strong	3—Weak	3—Weak	2—Moderate	2—Moderate	2—Moderate	1—Strong	1—Strong	2—Moderate
Global Rating			3—Weak	2—Moderate	2—Moderate	1—Strong	3—Weak	2—Moderate	2—Moderate	2—Moderate	1—Strong	2—Moderate	2—Moderate	2—Moderate

Note. The global rating is based on the number of “Weak” ratings across the six sections. An article is given a global “Weak” rating if 2 or more sections receive a “Weak” rating; a “Moderate” rating is given if only one section was considered “Weak”; and, a “Strong” paper has no section “Weak” ratings. The Effective Public Health Practice Project tool contains two additional sections that do not contribute to the global rating so the scores are omitted from this table.

Table 5 provides a summarised overview of data extracted from included studies with respect to the direction of change in the body weight outcome by employment transition, gender, occupation type, and comparator (pre-post or against a control). Detailed extraction of included study results and definitions of occupation types are provided as online supporting information (S2 and S3 Tables).

Table 5

Summarised results from included studies by study design, employment transition, and occupation type.

	Measured anthropometry				Self-reported anthropometry
	Nooyens et al [18]	Monsivais et al [12]	Feng et al [54]	Pedron et al [51]	Morris et al [16]	Forman-Hoffman et al [17]	Zheng [14]	Chung et al [52]	Gueorguieva et al [55]	Godard [56]	Stenholm et al [53]	Syse et al [57]
Weight change relative to not retired/continued to work
Retirement
Active	/	/	/	/	/	W +M +	W +M +	A +	/	W +M +	/	/
Sedentary	/	/	/	/	/	W + -M -	W +M -	A +	/	W +M +	/	/
Overall	/	W +M -	W + -M +	W +M +	M +	W +M + -	/	A +	/	W +M +	/	A -
Lost job
Active	/	/	/	/	/	/	/	/	/	/	/	/
Sedentary	/	/	/	/	/	/	/	/	/	/	/	/
Overall	/	W +M +	/	/	M +	/	/	/	/	/	/	/
Weight change within employment transition (pre-post change)
Retirement
Active	M +	/	/	/	/	/	/	/	A +	/	W +M -	/
Sedentary	M +	/	/	/	/	/	/	/	A +	/	W +M -	/
Overall	/	W + M + A +	/	/	M + -	/	/	/	/	/	W + M -	/
Lost job
Active	/	/	/	/	/	/	/	/	/	/	/	/
Sedentary	/	/	/	/	/	/	/	/	/	/	/	/
Overall	/	W + M + A +	/	/	M + -	/	/	/	/	/	/	/

“W” or “M” indicate result is applicable to women or men, respectively; “A” indicates results that are not stratified for sex/gender.

“Active” or “Sedentary” refer to occupation types related to specific estimates. Definitions of occupation types differed between studies (see S2 Table). “Active” refer to occupations held before the employment transition that may include physically demanding tasks or primarily standing work or manual work, while “Sedentary” refer to occupations that involve primarily sitting/desk work or no manual work. “Overall” indicates results are not stratified by occupation type. Results are based on fully-adjusted models from included studies (see S4 Table for covariates each study used).

For weight changes relative to not retired/continued to work: “+” indicates higher or greater body weight outcome relative to a referent group of continually employed or non-retired participants; “-” indicates lower or lesser body weight outcome relative to a referent group of continually employed or non-retired participants; “/” indicates no applicable results. Bolded symbols indicate a statistically significant comparison where p-values are at least 0.05 or smaller.

For weight changes within employment transition (pre-post change): “+” indicates higher or greater body weight outcome relative to baseline; “-” indicates lower or lesser body weight outcome relative to baseline; “/” indicates no applicable results. Bolded symbols indicate a statistically significant comparison where p-values are at least 0.05 or smaller. Italicized symbols indicate a reported pre-post result where statistical significance was not tested.

Body weight outcomes within employment transitions by gender

A third (n = 4) of the included studies reported within-group weight change among the employment transitions [12,18,53,55]. Two of these studies reported results by occupation type only, which will be discussed in a subsequent section [18,55].

Among studies that reported the weight effect of retirement regardless of occupation type, women tended to experience weight gain while men both gained and lost weight. Monsivais et al [12] found women gained an average of 0.48 kg per year (95% confidence interval (95% CI) 0.28 to 0.68) over a mean follow-up period of 3.5 years, and Stenholm et al [53] found women experienced a mean BMI increase of 0.15 kg/m2 (95% CI 0.10 to 0.20) and had higher risk of obesity (RR = 1.15, 95% CI 1.09 to 1.21) in the survey wave right after retirement compared to the wave right before retirement (an average of 4 years apart between waves). However, among men, Monsivais et al [12] found men increased their weight when they entered retirement (0.52 kg/year, 95% CI 0.34 to 0.70) but Stenholm et al [53] found men experienced a decrease in BMI (-0.11 kg/m2, 95% CI -0.22 to -0.01), although, the quality of the second study is lower than the first.

In addition to analyzing weight change following retirement, Monsivais et al [12] found women and men who lost their jobs also gained weight at an average of 0.69 kg (95% CI 0.46 to 0.92) and 0.68 kg (95% CI 0.43 to 0.92) per year, respectively.

In addition to the four studies that examined weight change following employment transitions, Morris et al [16] reported proportions of men who experienced weight gains or losses of more than 10% following job-loss and retirement; however, statistical comparisons were limited to across-group differences only. While the proportions are not discussed in this section, they can be found in S3 Table.

Body weight outcomes of entering retirement compared to continued employment or non-retired groups by gender

Among five studies that reported results for women who retired relative to women who continued employment or were not otherwise retired [12,17,51,54,56], results were mixed: two studies found that retirement was associated with higher body weight outcomes [17,51], while three studies did not find that outcomes differed between those who retired and those who remained employed [12,54,56]. Using the U.S. Health and Retirement Study (HRS), Forman-Hoffman et al [17] found that among women with normal weight at baseline, retirees were more likely to gain 5% or more in BMI compared to those who continued to work at least 20 hours a week two years later (OR = 1.30, 95% CI 1.01 to 1.69). Pedron et al [51] used data from the KORA study in Germany and found that women gained 0.82kg/m2 more compared to women who were not retired over a follow-up period of between 6 to 10 years.

Studies that did not find a statistically significant effect of retirement on weight outcomes had periods of follow-up ranging from 2 to 4 years [12,54,56], and settings generally differed from the two studies above, with one exception [56]. Godard [56] examined weight outcomes between retirees and those who continued to work using data from the Survey of Health, Ageing and Retirement in Europe (SHARE), which included Germany among seven other European countries. In contrast to the findings of Pedron et al [51], Godard [56] found a non-statistically significant increase of 0.31 kg/m2 over a 2 to 4 year follow-up period. While Godard [56] used respondents who reported being employed or self-employed as their comparator, Pedron et al [51] included a broader group of non-retirees in more countries which included those who were unemployed and those experiencing long-term illnesses. However, the findings in Pedron et al [51] remained robust in sensitivity analyses when the comparator was limited to those who were employed. Of the three studies that did not find a statistically significant effect, two of the studies used objectively measured anthropometry [12,54], and one used self-reported body weight [56], while one of the two studies that found a statistically significant effect used self-reported body weight [17] and the other used objectively measured weight [51].

Results for men were similarly mixed. Among six studies that reported overall retirement effects for men [12,16,17,51,54,56], three studies found increased body weight outcomes [16,54,56], and three did not find a statistically significant difference [12,17,51]. Of the three studies that found a statistically significant effect, one was rated as weak in quality (Feng et al [54]), one was rated moderate (Morris et al [16]) and one was rated as strong (Godard [56]), while the remainder of the three studies were all rated as moderate. In a study using data from the China Health and Retirement Longitudinal Study (CHARLS), Feng et al [54] found male retirees experienced weight gain of 2.10 kg (p<0.10) and a BMI increase of 0.92 kg/m2 (p<0.05) higher compared to non-retirees over a two-year period based on objectively measured anthropometry. Morris et al [16] found that 7.5% of men who were retired for reasons unrelated to illness gained more than 10% of body weight five years later compared to 5.0% of men who were continuously employed (significantly different at p<0.05). While Godard [56] did not find men who retired experienced a significantly higher increase in BMI compared to those continuously employed, retirees did have a 11.5% increase in the probability of obesity compared to their working counterparts (p<0.05).

Comparing studies that reported a significant difference between men who retired relative to men who continued working or were otherwise unemployed with those that reported a non-significant difference in the same settings, some differences in analytic approach are notable. Monsivais et al [12] found men who entered retirement in the EPIC-Norfolk sample in the UK gained slightly less weight per year compared to those who remained employed, though the difference was not statistically significant. Model adjustments for important health behaviours relevant to weight outcomes, such as change in smoking and energy intake, were conducted in Monsivais et al [12]. This is in contrast to Morris et al [16], where adjustments for similar health behaviours were not done, and a significant difference was found. Morris et al [16] also found significantly more heavy drinkers and smokers among men who were not continuously employed compared to those who were, indicating non-negligible differences in the baseline characteristics of the two groups that may have explained the significant finding for body weight. Similar to the results among women discussed earlier, the results for men between Pedron et al [51] and Godard [56] were also mixed, but to a lesser extent. While Godard [56] found an increase in the probability of obesity among men, both studies reported non-significant results when the outcome variable was defined as BMI change [51,56]. Two of the three studies using objectively measured anthropometry did not report a significant finding [12,51], while two of the three studies using self-report did find that retirees differed in their weight gain from continued workers [16,56]. Where findings were statistically significant, the direction of change was generally consistent between studies using objective measurement and self-report [16,54,56].

Body weight outcomes of job-loss compared to continued employment by gender

Among included studies, only two investigated transitions other than retirement [12,16], with one study sampling only men [16]. That study found that British men who became unemployed for reasons not due to illness or retirement were 7.1% more likely to gain over 10% in body weight 5 years later at follow-up compared to 5.0% among men who were continuously employed over 5 years (statistically different at p<0.05). The second study considered the effects of employment transitions on annualized weight gain separately for British women and men [12]. Monsivais et al [12] found that annualized weight gain was significantly higher among women who lost their job (0.72 kg/year) compared to those who remained employed (0.42 kg/year) (p = 0.007). For men, weight gain did not significantly differ between those who remained employed and those who lost their job (0.63 kg/year and 0.68 kg/year, respectively) [12].

Body weight outcomes of employment transitions by occupation type and by gender

Seven studies stratified analyses by occupation types [14,17,18,52,53,55,56]. All studies defined occupation types based on the physical demands of the work, which often led to categorizing jobs as active versus sedentary, or manual versus professional/managerial (see S2 Table for our classification of results by occupation type).

Gueorguieva et al [55] did not stratify their results by sex/gender, but they found retirement was associated with yearly BMI increases that differ by occupation type (see S3 Table). Retirees from service and other blue-collar occupations tend to have a higher BMI per year change (0.12 kg/m2, p<0.05, and 0.13 kg/m2 per year, p<0.01, respectively) relative to retirees from professional and managerial occupations (0.04 kg/m2 per year). Similarly, Chung et al [52] found weight gain among retires to be higher compared to those who were currently working in those occupations and subgroup analyses found this was primarily due to those retiring from physically demanding occupations (p<0.05).

Stratification of results by job-type accounted for some of the differences in the general effect of employment transitions on body weight by gender; however, findings remain mixed.

Weight gain in women who retired from a physically demanding job was reported in two studies [17,53]. Physically heavy work was associated with an increase in BMI of 0.30 kg/m2 (95% CI 0.15 to 0.46 kg/m2) and an increased risk of obesity (RR = 1.20, 95% CI 1.07 to 1.34) during retirement transition in a Finnish cohort [53]. In a US prospective cohort, female retirees from blue collar occupations were 1.58 times more likely to gain at least 5% in BMI relative to those who remained working in those occupations (OR = 1.58, 95% CI 1.13 to 2.21) [17]. Two additional studies found BMI increased following retirement from physically demanding jobs; however, findings were not statistically significant [14,56].

Among men, weight increases following retirement from a physically demanding job are supported by three studies (Zheng [14], Nooyens et al [18] and Godard [56]) and not supported by two others (Forman-Hoffman et al [17] and Stenholm et al [53]). Zheng [14] and Forman-Hoffman et al [17] both utilized a similar dataset derived from the HRS in the US but had different findings. This difference may be reconciled through a difference in the studied outcome. Forman-Hoffman et al [17] set a clinically significant threshold of a 5% increase in BMI over a 2 year period as their outcome, while Zheng [14] used change in BMI as a continuous variable in their analysis. In addition to finding retirees from physically demanding jobs gained more weight compared to those who were not retired, Zheng [14] also found that the BMI change experienced by retirees from strenuous jobs was significantly different from the BMI change experienced by retirees from sedentary jobs. Nooyens et al [18] examined change in weight (kg) and waist circumference (cm) per year and Stenholm et al [53] examined BMI change and relative risk of obesity. Nooyens et al [18] found that men who retired from active occupations experienced greater annualized weight gain and waist circumference growth compared to retirees from sedentary jobs. In contrast, Stenholm et al [53] found that men who retired from sedentary jobs experienced decreases in their BMI and retirement from physically demanding occupations did not significantly change BMI or risk of obesity. Finally, Godard [56] found that men retiring from strenuous occupations had an increased probability of obesity compared to those who were continuously employed, and that this increase was statistically significantly more than retirees from sedentary jobs.

Role of sleep and other health behaviours as potential mediators

Only one study included analyses of sleep. Monsivais et al [12] reported those who experienced job-loss had greater odds of subsequent sleep loss due to worry relative to those who remained employed (OR = 4.5, p<0.01). In contrast, the odds of sleep loss were not statistically different between those who entered retirement and those who remained employed [12]. Despite a lack of a formal analysis of mediation, Monsivais et al [12] suggests that sleep may play a mediating role in the relationship between employment transitions and body weight change for future research to consider.

In light of the limited evidence related to sleep, we further explored the included studies for other health behaviours that may mediate the impact of employment transitions on weight change. While most studies adjusted for health behaviours associated with body weight, several studies examined change in smoking and alcohol consumption, physical activity, and diet as outcomes following employment transitions [12,14,16-18,54,56], but only Nooyens et al [18] included separate models that examined the effect of change in health behaviours on weight change. None of the other included studies statistically examined health behaviours as mediators, but instead offered conceptual interpretations that the examined health behaviours may be potential mechanisms or channels through which employment transitions may impact subsequent weight change.

Morris et al [16] assessed changes in alcohol consumption and smoking following employment transitions, and found that men who became unemployed for reasons unrelated to illness were more likely to reduce their drinking compared to those who were continuously employed, but were not any more or less likely to reduce their cigarette smoking than continuously employed men. In the discussion of the results, Morris et al [16] did not link these findings with their weight change findings, but instead suggested that the weight gain observed among men who experienced job-loss may be due to changes in diet or physical activity.

Physical activity was the most common health behaviour examined [14,16-18,54,56]. An increase in physical activity post-retirement was commonly found; however, the findings by gender are mixed [14,17,18,54,56]. Godard [56] found women were more likely to engage in moderate leisure-time physical activity following retirement while Forman-Hoffman [17] and Zheng [14] both found that retirees did not engage in more vigorous physical activity. However, Forman-Hoffman [17] and Zheng [14] assessed change in vigorous physical activity as inclusive of work-based physical activity, which may mask how leisure-time physical activity changes following retirement. Among men, studies found an increase in vigorous physical activity [14,17]. Zheng [14] found that the increase in vigorous physical activity was mainly among those who retired from sedentary occupations. However, Nooyens et al [18] and Godard [56] did not find that men increased their leisure-time physical activity after retirement.

Diet was also assessed as a hypothetical mediator. Feng et al [54] found that men increased food consumption following retirement and suggested that this may explain why men experienced increases in BMI following retirement while women did not. Nooyens et al [18] found that male retirees increased their frequency of intake of vegetables and rice and decreased their intake of potatoes specifically. When stratified by occupation type, retirees from sedentary jobs also decreased their intake of proteins while retirees from active jobs decreased their overall energy intake and fibre density [18]. Of these changes in diet following retirement, only a decrease in fibre density was associated with subsequent weight gain [18].

Discussion

In this systematic review, we screened 6,001 records and identified 12 studies that investigated the longitudinal association between employment transitions and body weight change in middle-aged and older adults. Few studies added to our understanding of job-loss and its effect on body weight as the most commonly examined transition was retirement. Many of the studies reporting the effects of employment transitions on subsequent body weight change recognized that the physical demands of different occupations can influence body weight [14,17,18,52,53,55-57]. Almost all studies were in moderate quality and accounted for both gender and occupation type differences through study design or analyses. The results of the included studies generally fall into two categories: pre-post changes in weight within employment transitions or changes in weight for employment transitions relative to a comparison group (i.e., those who remained employed or did not retire). Findings from pre-post analyses suggest women tend to experience weight gain [12,53], and men may experience either weight loss [16,53], or weight gain following retirement [12,16,18], but there is much uncertainty associated with these findings due to the few pre-post only studies available and even fewer studies that included women in the study sample. When stratified by occupation type, weight gain may be related to retirement from physically demanding occupations [18,53], while weight loss may be related to retirement from sedentary occupations [53]. The modifying effect of occupation type, however, on weight change due to retirement is also likely to further differ by gender. As for example, men retiring from physically demanding occupations may lose, not gain, weight if that weight is predominantly muscle from occupations involving weight-bearing activity [58]. While these studies provide us with some insight into potential trajectories in weight change within employment transitions, evidence is more robust when a comparison group is used to determine weight change from employment transitions.

Among between-group studies, we found that women who enter retirement tended to experience more weight gain [17,51], or similar weight change, relative to their working or non-retired counterparts [12,54,56]. However, weight gain was not consistently supported among studies using objectively measured anthropometry [12,54]. The literature also appears inconsistent in reported effects of retirement for among men [12,17,51,54,56]. The evidence on gender differences, however, is not clear given that both the increases and decreases in the weight of women and men were not consistently found across different studies, potentially owing to differences in how studies defined their outcomes [12,14,17,51,54,56]. Moreover, we also found that occupation type modified the effect of retirement on weight, but results were more clear for the weight gain after retirement from physically demanding occupations [14,17,52,56], than the reported weight change following retirement from sedentary jobs [14,17,52,56]. Finally, job-loss appears to lead to more weight gain than continuous employment [12,16]; however, more studies are required to ascertain gender differences.

The findings of weight change among included studies may be relatively modest when compared to age-related weight trajectories of middle-aged and older adults [59-61]. Within the HRS, women and men generally experienced linear increases in BMI from ages 51 to 75 years [60]. Zheng [14] simulated BMI change 10 years post-retirement using the results from their analysis of retirees within the HRS for sedentary and strenuous occupations and found the weight loss experienced among retirees from sedentary occupations was reversed by age effects, leading to modest BMI gains post-retirement transition. However, this linear rise in BMI may not be true for settings outside of the USA. Using data from the Canadian National Population of Health Survey (NPHS), Wang et al [61] found women and men in Canada generally experience steady decreases in BMI from ages 65 to 96 years. Further, Dugravot et al [59] found women and men in the French GAZEL study experienced a gradually slower rate of BMI increase between the ages of 45 to 65 years, suggesting a plateauing of weight change. It is important to note the differences in how variables were defined across the included studies and how it impacts both the statistical and clinical significance of findings. For example, the inclusion of those who are long-term unemployed in a “not retired” group may not make it a suitable comparator as long-term unemployment may be associated with chronic illnesses that have consequences on body weight; however, Pedron et al [51] showed that their results were statistically robust even with different definitions of their comparator. Operationalization of weight gain as a categorical or as a continuous variable, however, does seem to influence whether a statistically significant finding is detected, as was the case with Zheng [14] and Forman-Hoffman et al [17]. Change in BMI was measured on a continuous scale in Zheng [14] while change was categorized as 5% or greater gain, 5% or greater loss, or no change in Forman-Hoffman et al [17]. A statistically significant increase in BMI was found among retired men in Zheng [14], but the odds ratio for 5% or greater gain was not significant in Forman-Hoffman et al [17], suggesting men may gain weight following retirement that does not exceed clinical thresholds. Indeed, many of the findings of weight gain measured on a continuous scale are below clinically significant thresholds. Gains in excess of 1 kg/year have been associated with elevated risk of all-cause and cardiovascular disease-related mortality in older adults (age 50+) [62]. Similarly, changes of 5% in weight or +/- 1 BMI unit have also been found to be associated with excess mortality risk [63].

The difference in continuous and categorical measures of body weight in middle-aged and older adults is discussed in Paige et al [64]. In examining the effect of education level on annual weight change in a population-based cohort of adults aged 45 and older in Australia, Paige et al [64] defined weight change in four different ways: a categorical variable using percentage change, a categorical variable using absolute change, a continuous variable using percentage change, and a continuous variable using absolute change. The authors found no statistically significant or clinically meaningful change in body weight when the continuous variables were modeled [64]; however, they found that those with higher educational attainment had lower risk of both weight gain and loss of more than 1kg compared to those with no school certifications [64]. Future research that examines employment transition and weight change should consider a similar approach to modeling by including alternative measures and definitions of the outcome to avoid obscuring potentially important directions and clinically meaningful changes. Additionally, future literature reviews of this relationship should consider alternative methods to synthesize information from studies, such as individual participant data meta-analyses which can help address differences between studies in outcome definition, factors adjusted for, comparator groups used, and setting [65].

Our search only identified one study, that examined sleep in addition to the effect of employment transitions on body weight. While no other study included sleep in their analyses, there was some acknowledgement that sleep might also be an important underlying mechanism that has not yet been examined [51,56]. Both Godard [56] and Pedron et al [51] cited previous work from Eibich [66] who examined the effect of retirement on health and health behaviours. Eibich [66] found that improved overall health status following retirement was mediated through improvements in sleep and increases in physical activity. Monsivais et al [12] found that those who lost their job were more likely to experience sleep loss compared to those continuously employed. Taken together, sleep may be an important mediator for health outcomes across both involuntary transitions like job-loss and voluntary transitions like planned retirement. Given the paucity of studies that considered sleep and that the existing literature remains mixed with respect to mediation by physical activity and moderation by occupation type, future research ought to explore sleep further as suggested by authors in this area.

There are several limitations with this review. The first limitation is the difficulty in identifying studies that investigated “employment transitions” as this did not appear to be a term that was used explicitly in many articles. Often, a thorough reading of the methods section was required to verify whether a study truly examined a change in employment status over, at least, two points of time. However, our use of a dual independent reviewer approach and practice screenings would have mitigated most of this problem. However, some studies may still have been missed. A second limitation is related to our quality appraisal tool. Observational studies that used either pre-post or pre-post with comparison were both given the same rating under the design domain, despite the latter offering potentially more robust findings when well-executed. Assigning a different score to each of the studies would have helped to further differentiate the quality of the evidence. Thirdly, publication bias may be present. As this review has found, null results have been observed when considerations around sex/gender, occupation type, and direction of body weight change are not adequately addressed. Lastly, the studies identified were mostly based in high-income countries. With respect to differences in labour force markets, cultural norms of working and ageing, health and pension systems, we expected the impacts of employment transitions on body weight to vary by countries. This highlights another important gap in the literature that future research should address.

The findings from our systematic review suggest that the effect of employment transitions on body weight outcomes in middle-aged and older adults differ by gender, occupation types and countries. The existing body of literature brings to attention the large research gaps. Potential key areas of focus in future studies include: the causal relationship and sleep as a third variable, the effect of job-loss as opposed to retirement, whether the effect of employment transitions differ by levels of socioeconomic status or across comorbidities related to overweight or obesity and, the impact of alternative operational definitions of weight change on study results. More rigorous evidence synthesis methods are also needed in future literature reviews to address study and topic complexity.

(DOCX)

Click here for additional data file.

Search strings used in each bibliographic database.

(DOCX)

Click here for additional data file.

Details of employment categories in included studies.

(DOCX)

Click here for additional data file.

Results of included studies by sex/gender.

(DOCX)

Click here for additional data file.

Covariates used for main model in included studies.

(DOCX)

Click here for additional data file.

12 Apr 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Stephanie Prince Ware, PhD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Additional Editor Comments (if provided):

The manuscript refers to body “weight” throughout, but the correct term is body mass. Please refer to this paper:

Winter EM, Abt G, Brookes FB, Challis JH, Fowler NE, Knudson DV, Knuttgen HG, Kraemer WJ, Lane AM, van Mechelen W, Morton RH, Newton RU, Williams C, Yeadon MR. Misuse of "Power" and Other Mechanical Terms in Sport and Exercise Science Research. J Strength Cond Res. 2016 Jan;30(1):292-300. doi: 10.1519/JSC.0000000000001101. PMID: 26529527.

Ensure that middle-aged is consistently hyphenated.

Page 5, please include in brackets examples of the “other modifiable health behaviours”.

Methods: Please provide the full bibliographic database-specific search strategies; these can be placed in a supplement.

Was full-text screening also done by two independent reviewers?

Can the authors please expand on what they mean by 11 of the 21 items were directly relevant to assigning a global rating for risk of bias? What did they do with the other 10 items?

Was this review protocol registered or published prior to commencement? Please include the proposal registration information.

Was a meta-analysis considered?

Page 7, BMI is incorrectly spelt out as “body weight index” rather than “body mass index”

Why were other measures of body mass status not incorporated (e.g., waist circumference, waist-to-hip ratio, body fat %)

Table 4 – why was a p<0.10 chosen for identifying “significance”? Also, the text for Results includes other p-values (e.g., p<0.05, p<0.01).

Table 4 – suggest changing O to A (all)

Table 4 – Feng et al., why is there a + and – for women? Same for Forman-Hoffman et al. for men.

It’s not clear if the findings are based on the statistical significance of null or fully adjusted models? Ideally they should be from fully adjusted models and the covariates included in the adjustment should be provided. Were comorbidities and socioeconomic status considered as potential effect modifiers?

Page 19, the discussion of results for Monsivais et al. was a bit repetitive for men and women.

Page 19, four studies regarding weight change following employment transitions are mentioned, but only one (Morris et al.) is discussed.

Page 22, The first paragraph contains content that is best suited for the Introduction or Discussion.

Page 24, this is a post-hoc analysis regarding other health behaviours. This should be outlined in the Methodology.

The authors should interpret all results relative to the certainty of the evidence. Additionally, many of the broad stroke findings are related to 1 or 2 studies.

The PRISMA checklist says the “certainty of evidence” was assessed, but there is no description of this in the Methods or the Results.

The PRISMA checklist item 13f is missing a description.

If the study protocol was not registered, this should be stated as a major limitation of the study.

Was there any difference in outcomes based on whether body mass was self-reported or objectively measured?  There is no discussion about how this may affect results.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

Reviewer #2: N/A

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is a helpful review summarising the literature on this important topic. I have the following suggestion to improve the manuscript.

Abstract:

'Employment is strongly associated with weight' is a very general statement, and does not tell us anything about the direction of the association. Maybe this could be more specific?

'they are also more vulnerable to employment transitions' Since employment transitions have not be defined, it may be more helpful to say 'job loss'.

Introduction:

line 9 'lower incidence' may not be the appropriate term for education and skills training.

It would be helpful early on to define the term 'employment transitions' and introduce the range of employment transitions that are of interest.

Methods:

Was the review protocol registered before starting - if yes please add details in the methods section.

It may be helpful for clarity to add an inclusion/exclusion table that follows a standard format e.g. Setting/Participants/Exposure/Outcomes/Study type/Publication type/Publication year/Language.

Results:

Further details need to be added to Table 4 so that this table can be understood without reference to the article text. For example 'Active, Sedentary, Overall' presumably refer to the type of job that the participant held before retirement or job loss, but this is not stated. Also some details should be provided in the methods of how jobs were categorised into active or sedentary, - the type of information which was extracted from the papers and how this was split into the categories presented.

Table 4 could be a very helpful overview of the evidence, but needs a little work to make it more easy to understand.

Conclusions

Given the wide range of factors that may contribute to changes in weight status at job loss or retirement, it may be worth considering an individual participant data meta-analysis in future work. (ref: https://www.bmj.com/content/340/bmj.c221)

Reviewer #2: This is a relevant review investigating the impact of employment transitions on body weight. Twelve studies were included in the review, all considering transitions from employment to retirement, however only two studies considered transitions from employment to unemployment. The conclusion of the review is that the existing studies do not provide a clear enough picture to draw any conclusions on the impact of employment transitions and weight change.

Overall, the review is well written, and I believe the topic is relevant and the review will contribute to the literature by highlighting important areas where more research is needed.

Comment #1

The conclusion is that no firm conclusions can be drawn, however, when looking at table 4, it seems pretty consistent that women gain weight after retirement. Do you believe that there is not enough evidence to draw this conclusion?

Comment #2

Odds ratios are not presented in the same way throughout the manuscript. Please try to align this.

Comment #3

You assess the quality of each study; however, you do not mention this when describing the results or when you draw conclusions. It would be nice to see some considerations of the quality – do the good quality studies weigh more when drawing conclusions? Or do you weigh the results from each paper equally?

Comment #4

Introduction page 3: I do not understand the sentence: Coupled with the rise in labour force participation of middle-aged and older workers are unique challenges such as lower incidence of continued education and skills training. Perhaps you could elaborate why this is relevant.

Comment #5

Methods page 5: The sentence: Screening, assessment, and inclusion of studies published in English, French, and Chinese. I feel like there is something missing to this sentence?

Comment #6

Results page 7: you write body weight index, but it should be body mass index.

Same page: The sentence: odds or risk of becoming overweight or obese. There is a consensus within the research field of obesity that obesity is not something you are or become – it is a condition that you have. Like a disease. Whatever disease a person may have, it may not define them as persons or individuals. For example, having a chronic disease such as cancer does not make a person identify as a cancerous person. I would rewrite the sentence: odds or risk of overweight or obesity. Please go through the manuscript to make sure you use person-first language throughout.

Comment #7

Table 2: It should be outcome(s) measures instead of outcome(s) measured. Sometimes you use a full stop and sometimes you don´t.

Comment #8

A few times (e.g. on the bottom of page 19) you describe women who continued employment or were not otherwise retired/unemployed. Please elaborate what you mean by not otherwise retired/unemployed.

Comment #9

Results page 24: Suggestion for rephrasing the sentence: suggests that there may be a mediating role that sleep has in the impact of employment transitions on body weight change… to: suggests that sleep may play a mediating role in the relationship between employment transitions and body weight change…

Same page: you write: dietas outcomes. I have never come across this word before. Is it a typo or is it a real phrase?

Comment #10

Discussion page 26: you write: However, this linear rise in BMI may not be true for other countries. Perhaps you should mention here where the study is conducted, as this is not fresh in the readers memory.

Comment #11

Discussion page 27: you mention that operationalization of weight gain as a categorial or continuous variable has influenced the results in the studies by Zheng and by Forman-Hoffman et al. It seems that you only describe changes in BMI as a continuous variable and do not mention the categorical changes? I do not see from your discussion how categorical or continuous weight change has influenced the results.

Additionally, in the following section, you mention a study examining weight change as both categorical and continuous (Paige et al.); however, you only mention the results for the continuous variable.

I am a bit confused what you wish to discuss with these two paragraphs, and I do not think that the message is coming across very clear. Perhaps you should consider rewriting these paragraphs.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

27 May 2022

We have copied our responses from the attached Response to Reviewer document into this field. We wish to thank the editor and the reviewers again for their careful reading of our manuscript.

*Please note the page numbers referenced are mapped to the clean version of the manuscript.

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf

and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

We have revised our manuscript in accordance with the style requirements and file naming conventions.

*** Editor comments ***

1. The manuscript refers to body “weight” throughout, but the correct term is body mass. Please refer to this paper:

Winter EM, Abt G, Brookes FB, Challis JH, Fowler NE, Knudson DV, Knuttgen HG, Kraemer WJ, Lane AM, van Mechelen W, Morton RH, Newton RU, Williams C, Yeadon MR. Misuse of "Power" and Other Mechanical Terms in Sport and Exercise Science Research. J Strength Cond Res. 2016 Jan;30(1):292-300. doi: 10.1519/JSC.0000000000001101. PMID: 26529527.

Thank you for sharing an article that highlights the difference between body mass and Newtonian weight and the continued incorrect usage of “weight” in sport and exercise science. We acknowledge that the use of body weight in our manuscript to refer to measures of kilograms or body matter would be imprecise in the Winter et al. sense. In our experience, within obesity epidemiology, the discourse around changes to anthropometrics (measured in e.g., kg/BMI) primarily expresses it as changes to body weight, and not as body mass – see examples from Hu et al 2018; Mozaffarian et al 2011; and Tasali et al 2022.

The use of body weight as the term used to capture kilograms also bears clinical relevance. For example, clinical practice guidelines for managing overweight and obesity from North American and European clinical professional groups make similar statements using body weight rather than body mass when referring to anthropometric changes. Guidelines have also identified clinically meaningful thresholds for weight change targets to manage cardiovascular and metabolic risks and other comorbidities; gaining 5 kg or more in body weight during adulthood is a risk for many major chronic conditions (WHO 2004).

We acknowledge that PLOS One appeals to a wide variety of disciplines and to both clinical and non-clinical audiences alike. We believe that the term “weight” is the more accessible word choice; however, we leave it the editor to make the final decision regarding the use of “weight” or “mass” in our manuscript.

American Psychological Association, Clinical Practice Guideline Panel. (2018). Clinical practice guideline for multicomponent behavioral treatment of obesity and overweight in children and adolescents: Current state of the evidence and research needs. Retrieved from http://www.apa.org/obesity-guideline/obesity.pdf

Hu Y, Zong G, Liu G, et al. Smoking Cessation, Weight Change, Type 2 Diabetes, and Mortality. N Engl J Med. 2018;379(7):623-632. doi:10.1056/NEJMoa1803626

Mozaffarian D, Hao T, Rimm EB, Willett WC, Hu FB. Changes in diet and lifestyle and long-term weight gain in women and men. N Engl J Med. 2011;364(25):2392-2404. doi:10.1056/NEJMoa1014296

Tasali E, Wroblewski K, Kahn E, Kilkus J, Schoeller DA. Effect of Sleep Extension on Objectively Assessed Energy Intake Among Adults With Overweight in Real-life Settings: A Randomized Clinical Trial. JAMA Intern Med. 2022;182(4):365-374. doi:10.1001/jamainternmed.2021.8098

Wharton S, Lau DCW, Vallis M, et al. Obesity in adults: a clinical practice guideline. CMAJ. 2020;192(31):E875-E891. doi:10.1503/cmaj.191707

Yumuk V, Tsigos C, Fried M, et al. European Guidelines for Obesity Management in Adults [published correction appears in Obes Facts. 2016;9(1):64]. Obes Facts. 2015;8(6):402-424. doi:10.1159/000442721

World Health Organization. Global strategy on diet, physical activity and health. Geneva, Switzerland: WHO Publications, 2004

2. Ensure that middle-aged is consistently hyphenated.

We have made edits to consistently hyphenate middle-aged throughout the manuscript.

3. Page 5, please include in brackets examples of the “other modifiable health behaviours”.

We have provided examples of the health behaviours that the cited studies used. The sentence now reads:

Sleep may contribute to explaining the heterogeneity that appears to exist in the literature on job-loss and retirement and body weight gain in older adults, yet it has not received as much attention as physical activity [14,18], or other modifiable health behaviours (e.g., alcohol consumption and smoking) [15,16].

4. Methods: Please provide the full bibliographic database-specific search strategies; these can be placed in a supplement.

We have placed the search strategies into a supplemental table (new Table S1). We have referenced it in our Methods section.

5. Was full-text screening also done by two independent reviewers?

Yes, full-text screening (as well as reference-tracing of included studies) was also done by two independent reviewers following the same division of work as with title and abstract screening. We have updated the manuscript to clarify this. The change can be found on page 8:

“Records were removed based on exclusion criteria and eligible full-texts were retrieved and screened for inclusion and reference-tracing adhering to the same division of work between two independent reviewers.”

6. Can the authors please expand on what they mean by 11 of the 21 items were directly relevant to assigning a global rating for risk of bias? What did they do with the other 10 items?

Thank you for the opportunity to clarify. Within the quality appraisal tool that we selected, there are 11 items that the authors of the tool deem to be relevant to their scoring algorithm. These 11 questions are presented in Table 3. In addition to these items, there are a total of 10 additional questions in the tool that have no bearing on how the studies are scored. The topics covered by these questions are described below. We excluded presenting the responses to these items as the core focus of the excluded items pertain to clinical trials, which our responses to would be “not applicable” and are not consequential to the final rating.

Section B asks 3 questions specific RCTs and whether the selected design was appropriate for the research question.

Section G asks 3 questions regarding the integrity of the intervention received (e.g., percentage of participants who were given treatment, likelihood of contamination).

Section H asks 4 questions regarding the appropriateness of the analytic approach selected by the authors (e.g., unit of analysis, intention to treat).

We have updated the manuscript with text to clarify why these 10 items were not relevant and not used. The text is on page 10 and reads:

“The remaining 10 items from the tool are not relevant to the studies that were included in our systematic review and thus, had no bearing on the scoring assignment. For example, the items include questions related to randomized controlled trial design, the integrity of the intervention received, and intention-to-treat analytic approaches [34].”

7. Was this review protocol registered or published prior to commencement? Please include the proposal registration information.

The proposal was not registered prior commencement. Literature review work had begun as a part of ACTT’s thesis as background and rationale for the research and the suggestion to complete the work as a systematic review was made after. ACTT wrote a PROSPERO-style protocol document which was reviewed by the supervisory committee (listed as co-authors) and used to guide the actual systematic review work; however, we did not register it to the PROSPERO database.

8. Was a meta-analysis considered?

We did not consider a meta-analysis. Based on some limited literature searches conducted by ACTT for his thesis proposal, we did not believe that we would have identified enough studies to allow for a quantitative synthesis like a meta-analysis. This was confirmed after only 12 studies were identified in the systematic review. Also, the included studies varied in how outcomes were defined (e.g., weight change, BMI change, categorical weight change of 1kg or 5%) and which subgroups were analysed, if any (gender or type of occupation). Therefore, our strategy for synthesizing results was narrative in nature. Our findings showed wide heterogeneity and paucity of data that would confirm this choice.

9. Page 7, BMI is incorrectly spelt out as “body weight index” rather than “body mass index”

We have corrected this typographical error.

10. Why were other measures of body mass status not incorporated (e.g., waist circumference, waist-to-hip ratio, body fat %)

We were interested in change in body weight over time given the important role that excess weight gain of 5 kg or more has for chronic disease (WHO 2004). While we did not explicitly exclude other measures of body weight, our search terms were selected to reflect our research question: ‘what is the impact of employment transitions on body weight in women and men?’ Anthropometric measures such as waist circumference, waist-to-hip ratio and body fat % are other important obesity outcomes that reflect body composition and could be the focus of a future review. We note, however, that we did find papers that included such anthropometric outcomes in addition to Kg.

World Health Organization. Global strategy on diet, physical activity and health. Geneva, Switzerland: WHO Publications, 2004

11. Table 4 – why was a p<0.10 chosen for identifying “significance”? Also, the text for Results includes other p-values (e.g., p<0.05, p<0.01).

The use of p<0.10 in Table 5 (formerly Table 4, please see response to reviewer #1’s comment on our Methods) was originally based on some of the included studies where the selected and

presented p-value was p<0.10, so we applied this cut-off to all studies. However, after discussion and re-consideration, we have decided to update this table to reflect results that were significant at p<0.05. The table description has been updated to reflect the new cut-off, and the results have been cross-checked with the studies in line with the new p-value.

We wish to clarify that the p-values reported in the Results text refer to statistical significance level chosen by the authors of the included studies that met our review criteria. Thus, the range of values reflects the diversity in reporting of significant findings and thus the overall evidence.

12. Table 4 – suggest changing O to A (all)

We have updated Table 5 (formerly Table 4) using the suggestion of A to represent “all” instead of O for “overall”. The note underneath Table 5 has also been updated to reflect the updated symbol.

13. Table 4 – Feng et al., why is there a + and – for women? Same for Forman-Hoffman et al. for men.

In Feng et al., the authors modelled both BMI and body weight in kilograms. For women, the BMI model showed weight gain of +0.25 kg/m2 while the kilogram model showed weight loss (-0.06 kg). Both results were not statistically significant at the authors’ chosen value of p<0.10. Therefore, we indicated this by using both the + and the – symbols for retired women, without boldface.

In Forman-Hoffman et al., the results are based on a multinomial logistic regression model using a three-level categorical outcome of ≥5% weight gain, ≥5% weight loss, and no change serving as the reference. The authors reported an odds ratio above 1 for both losing ≥5% and gaining ≥5% in weight among retiring normal weight men compared to men who continued to work, but the ORs were not significant at the authors’ chosen value of p<0.05. Therefore, we indicated this by using both the + and – symbols for men who retired, without boldface.

Detailed study results from each study can be found in Table S3 (formerly Table S2).

14. It’s not clear if the findings are based on the statistical significance of null or fully adjusted models? Ideally they should be from fully adjusted models and the covariates included in the adjustment should be provided. Were comorbidities and socioeconomic status considered as potential effect modifiers?

We agree and confirm that the findings reported are based on fully-adjusted models reported by the included studies. We have added a note to the bottom of Table 5 (formerly Table 4) and Table S3 (formerly Table S2) to indicate this. We have placed the covariates from the adjusted models into a supplementary table (Table S4).

Socioeconomic status variables were often considered by included studies only as confounders. We did not find the use of socioeconomic status variables as effect modifiers in any of the included studies. We had not previously explicitly investigated comorbidities as a potential effect modifier; however, we have revisited the included studies and we can confirm that the studies did not consider this either.

We would like to mention that Morris et al was perhaps the closest to considering effect modification by underlying disease. Their retired and unemployed groups were further stratified by whether the reason for the employment status change was due to illness or not; however, this was based on a self-reported reason for not working. The results for both reasons for employment transitions were placed in Table S3 along with the rest of the extracted results. We also added a comment in our Discussion as an area of future investigation. It can be found on page 37:

“Potential key areas of focus in future studies include: the causal relationship and sleep as a third variable, the effect of job-loss as opposed to retirement, whether the effect of employment transitions differ by levels of socioeconomic status or across comorbidities related to overweight or obesity and, the impact of alternative operational definitions of weight change on study results.”

15. Page 19, the discussion of results for Monsivais et al. was a bit repetitive for men and women.

Thank you for identifying an opportunity to improve the readability of the manuscript. We understand that this section uses repetitive language when presenting results. To rectify this, we have retained the language for the retirement results, but we have simplified the language in subsequent passages that cover results for job-loss. It can be found in the lower half of page 23:

“In addition to analyzing weight change following retirement, Monsivais et al [12] found women and men who lost their jobs also gained weight at an average of 0.69 kg (95% CI 0.46 to 0.92) and 0.68 kg (95% CI 0.43 to 0.92) per year, respectively.”

16. Page 19, four studies regarding weight change following employment transitions are mentioned, but only one (Morris et al.) is discussed.

In the section that describes results “within employment transitions”, we limited our discussion to Monsivais et al and Stenholm et al because both studies present results for employment transitions without occupation types as effect modifiers. The two other studies that we counted among the four were Nooyens et al and Gueorguieva et al and they did not present results of their analysis for employment transitions overall, opting to only present results specific to certain occupation types. We decided to save the discussion of these two studies for a later section where we explore differences in outcomes by occupation type. We mentioned we would revisit Nooyens et al and Gueorguieva in the last sentence of the first paragraph on the top of page 23: “Two of these studies reported results by occupation type only, which will be discussed in a subsequent section […]”. The second paragraph does cover the results of Monsivais et al and Stenholm et al.

The Morris et al paper was not counted among the four papers of this section as their analysis compared across group differences (e.g., discontinuously employed vs continuously employed), rather than within each transition. However, the authors did present pre and post values where we could get a sense of the direction of change. We opted to discuss these briefly and reserved the majority of the discussion of across-group differences for a later section. We can remove it from the manuscript, if the editor considers this paragraph a distraction to the reader

17. Page 22, The first paragraph contains content that is best suited for the Introduction or Discussion.

We have deleted this paragraph and moved content of it into the first paragraph of the Discussion section as suggested.

18. Page 24, this is a post-hoc analysis regarding other health behaviours. This should be outlined in the Methodology.

We agree that the component of the synthesis related to other health behaviours beyond sleep is post-hoc, and we have updated the Methodology section with text that describes this additional work we carried out. It has been added to the bottom of page 10 and is provided here for reference:

“The narrative synthesis revealed additional health behaviours beyond sleep that were studied as mechanisms; thus, a post-hoc analysis was conducted to summarise additional results.”

19. The authors should interpret all results relative to the certainty of the evidence. Additionally, many of the broad stroke findings are related to 1 or 2 studies.

We agree that our broad stroke summary of findings for within-group pre-post changes in the Discussion section were originally related to only 1 or 2 studies. However, we believe that the paragraph that follows, regarding across-group differences, does sufficiently cover a wider range of studies (2 to 6 studies). We owe this to our search identifying more across-group studies (e.g., retired vs not retired) than pre-post/within-group studies, which enabled us to summarise findings from multiple studies that can be more general. More so, with respect to job-loss, we identified one study that included both women and men and a second study that only included men. Thus, we were limited again by the few studies that addressed this research area. To reflect the uncertainty due to limited evidence and the instance where we only connected our broad stroke findings to 1 or 2 studies, we have added new text on page 32 as follows:

“Findings from pre-post analyses suggest women tend to experience weight gain[12,53], and men may experience either weight loss [16,53], or weight gain following retirement [12,16,18], but there is much uncertainty associated with these findings due to the few pre-post only studies available and even fewer studies that included women in the study sample.”

We only included one sentence regarding the findings for job-loss, and we feel that it accurately highlights what we found in the two studies, so we have not edited it. We have provided an excerpt from page 33: “Finally, job-loss appears to lead to more weight gain than continuous employment [12,16]; however, more studies are required to ascertain gender differences.”

20. The PRISMA checklist says the “certainty of evidence” was assessed, but there is no description of this in the Methods or the Results.

Thank you for the opportunity to clarify. It was not possible to assess the certainty of evidence because the outcomes assessed – BMI, kg, kg/year, and odds of change – differed between studies. As noted earlier, we believed that generating a single estimate of effect would not be feasible and thus no measurement of certainty.

We elected to, instead, use the Discussion section to re-present the conflicting findings and the methodological differences between studies that might explain mixed findings. This is what we wanted to indicate in the PRISMA checklist file by stating “N/A – narrative synthesis” and “Acknowledged mixed results.”. We have re-written the description in the PRISMA checklist (item 15 and 22) to the following:

“No certainty assessment was conducted for our narrative synthesis. We discuss the mixed results in the Discussion.”. For item 15, this is followed by some excerpts from the text that reflect this discussion.

21. The PRISMA checklist item 13f is missing a description.

We apologise for missing the description for item 13f in the PRISMA checklist. A description has been added. We have provided that text here for reference:

“No sensitivity analyses were conducted for our narrative synthesis.”

22. If the study protocol was not registered, this should be stated as a major limitation of the study.

We did not register the study protocol; however, we do not consider this to be a limitation in any way since registration in PROSPERO or OSF is not peer-reviewed. Furthermore, we followed a strict protocol for the search that was a priori and hence criteria for inclusion/exclusion and search terms are defined beforehand to ensure accurate, precise and repeatable searching and results of this review; we pilot-tested our search strategy with preliminary searches before finalising the protocol for the final search. A great deal of oversight was provided to this process by ACTT’s thesis supervisory committee (listed as co-authors: RAM, WZ, and AIC) of experienced researchers with topic expertise. In addition to ACTT’s checks, WZ also provided a secondary high-touch review of extracted information to ensure accuracy of the independent reviewers’ work. ACTT also consulted with a health science reference librarian at their institution in organizing this review. We would defer to the editor’s decision on how this presents a limitation.

23. Was there any difference in outcomes based on whether body mass was self-reported or objectively measured? There is no discussion about how this may affect results

Thank you for the question. We had not considered this angle of the results, despite capturing which studies used self-report versus objectively measured outcomes in our characteristics table. To better highlight this, we have revisited our Table 5 (formerly Table 4) and added two headers to separate the studies that used measured outcomes and the studies that used self-report. The studies have been re-ordered under their appropriate header.

We have also updated our manuscript with some new text to explain some of the differences between studies using self-report vs objectively measured when comparing across employment transition groups. For men, 2/3 studies using self-report found a statistically significant effect of retirement on the outcome and 1/3 studies using objectively measured outcomes found a statistically significant effect. For women, there was also a difference between the two measurements: 1/3 studies using objective measurement found a statistically significant effect and 1/2 studies using self-report found a statistically significant effect. The new text can be found on pages 25 through 26 of the results and we added one comment to our discussion on page 33.

*** Reviewer #1 comments ***

Abstract:

'Employment is strongly associated with weight' is a very general statement, and does not tell us anything about the direction of the association. Maybe this could be more specific?

'they are also more vulnerable to employment transitions' Since employment transitions have not be defined, it may be more helpful to say 'job loss'.

We have revisited the background section of our abstract. The direction of association is bidirectional, and the direction of the change in weight is also dependent on the employment exposure itself. To give the readers something more specific, we have re-written the first sentence to reflect one area within the employment-weight relationship.

“Becoming unemployed is associated with poorer health, including weight gain.”

With respect to the use of “employment transitions”, we wished to capture job-loss, involuntary retirement, and other changes to employment status in a succinct term. However, as the reviewer has identified, we have yet to define the term within the abstract. We have replaced “employment transitions” with “changes to employment status” instead, which should help us retain our intention to refer to multiple kinds of transitions that can occur. The sentence now reads:

“Middle- and older-age adults are a growing segment of workforces globally, but they are also more vulnerable to changes to employment status, especially during economic shocks.”

Introduction:

line 9 'lower incidence' may not be the appropriate term for education and skills training.

It would be helpful early on to define the term 'employment transitions' and introduce the range of employment transitions that are of interest.

We have revisited the use of the term “lower incidence” from the cited work and have come up with “enrolment” as a more appropriate term to capture new entrants into skills training programs. Here is an excerpt of the updated text for your reference:

“Coupled with the rise in labour force participation of middle-aged and older workers are unique challenges such as lower rates of enrolment into continued education and skills training [5], employer preference for hiring younger employees [5], and ageism at work [5,6].”

We agree with the reviewer that we should introduce employment transitions earlier. We have added the following text when we first discuss the differential impacts of job-loss and retirement in our Introduction. Here is an excerpt:

“One of the risk factors for cardiovascular disease that has received some attention in the employment literature is body weight change [12-16], and how it differs following different changes in employment status. Changes in employment status, or employment transitions, span a range of events from exiting the labour force through job-loss and retirement to re-entry into the labour force by re-employment. Loss of a job without compensation has been found to be associated with […]”

Methods:

Was the review protocol registered before starting - if yes please add details in the methods section. It may be helpful for clarity to add an inclusion/exclusion table that follows a standard format e.g. Setting/Participants/Exposure/Outcomes/Study type/Publication type/Publication year/Language.

As indicated above, our review protocol was not registered before starting thus there is no number to add to the manuscript However, we did establish a clear protocol following standard PROSPERO procedures to ensure accurate, precise and repeatable searching and results of this review: 1) a priori criteria for inclusion/exclusion and search terms defined beforehand; 2) pilot-testing of search strategy with preliminary searches before finalising the protocol for the final search; and 3) separate searches and review by two independent researchers. Following this Reviewer’s suggestion, we have also summarised inclusion and exclusion criteria as the new Table 2. We have offset the subsequent table numbers (characteristics of included studies, quality appraisal, and summary of results) by 1 to accommodate this addition.

Results:

Further details need to be added to Table 4 so that this table can be understood without reference to the article text. For example 'Active, Sedentary, Overall' presumably refer to the type of job that the participant held before retirement or job loss, but this is not stated. Also some details should be provided in the methods of how jobs were categorised into active or sedentary, - the type of information which was extracted from the papers and how this was split into the categories presented.

Table 4 could be a very helpful overview of the evidence, but needs a little work to make it more easy to understand.

We thank the reviewer for sharing useful ways to improve the content of Table 5 (formerly Table 4). As suggested, we have added more details that allow Table 5 to stand alone without reference to manuscript text.

We have updated the title (new text in red) to better set up the reader for the content that it contains.

“Table 5. Summarised results from included studies by study design, employment transition, and occupation type.”

We have added a description of the occupation types presented in the table to the note under the table. Exact classifications of “active” or “sedentary” occupations on a per study basis are provided in full in a separate table (Table S2), but we have provided the breadth of criteria that were used to categorise them into the table as a note in Table 5.

‘“Active” or “Sedentary” refer to occupation types related to specific estimates. Definitions of occupation types differed between studies (see Table S2). “Active” refer to occupations held before the employment transition that may include physically demanding tasks or primarily standing work or manual work, while “Sedentary” refer to occupations that involve primarily sitting/desk work or no manual work. “Overall” indicates results are not stratified by occupation type.’

Conclusions

Given the wide range of factors that may contribute to changes in weight status at job loss or retirement, it may be worth considering an individual participant data meta-analysis in future work. (ref: https://www.bmj.com/content/340/bmj.c221)

We thank the reviewer for highlighting a methodological consideration for future studies. We agree that while resource intensive, the output of such a meta-analysis would be useful in processing the many factors that contribute to weight change following employment status change. We have added a line to our discussion to reflect this. Here is the text for your reference:

“Additionally, future literature reviews of this relationship should consider alternative methods to synthesize information from studies, such as individual participant data meta-analyses which can help address differences between studies in outcome definition, factors adjusted for, comparator groups used, and setting [66].”

*** Reviewer #2 comments ***

Comment #1

The conclusion is that no firm conclusions can be drawn, however, when looking at table 4, it seems pretty consistent that women gain weight after retirement. Do you believe that there is not enough evidence to draw this conclusion?

We agree that it may appear there is consistency results suggesting women gain weight after retirement. However, not all these studies can be directly compared, and we believe there is not enough evidence to draw a firm conclusion that women who enter retirement gain weight (pre-post) because only two studies provided estimates for women – one of which was rated as a weak study because of incomplete reporting of methods and sample selection. When we look at comparative studies (retirees vs continued workers), we also hesitate to draw any conclusions given the inconsistency in the weight gain finding. While the general direction is indeed positive, few studies found that the weight gain was statistically more than those who continued to work. For non statistically significant point estimates, they were generally small and close to the null; therefore, we are also unsure if drawing a conclusion on the direction (more vs less) is appropriate given current evidence.

While we spend an appreciable amount of space in the Discussion section trying to unpack differences between studies that may help us contextualize why there were different findings, we feel as though our final consensus remains and we cannot say for certain that the retirement transition is associated with more gain than continuing to work for women. We think that a “no firm conclusion” conclusion accurately represents the truth of our review’s findings.

Comment #2

Odds ratios are not presented in the same way throughout the manuscript. Please try to align this.

We thank the reviewer for identifying misalignments in the presentation of odds ratios. We have gone through the manuscript to align them to the format of (OR = ##.##, 95% CI ##.## to ##.##) where possible. There is one instance where the study only provided a p-value for the OR, so we deviate from the format in that case and present the p-value in place of the confidence interval.

Comment #3

You assess the quality of each study; however, you do not mention this when describing the results or when you draw conclusions. It would be nice to see some considerations of the quality – do the good quality studies weigh more when drawing conclusions? Or do you weigh the results from each paper equally?

With respect to the conclusions drawn, we do not find that the quality changes the uncertainty that we have about the evidence. Findings from the weak studies are corroborated by findings from some of the moderate studies but not by others. For example, Feng et al (a study we rated as weak) found that men who retired experienced weight gain that was statistically significantly more than men who kept working. Godard (a study we rated as strong) and Morris et al (moderate) also found men gained weight, yet three moderate quality studies did not find this to be the case.

To bring some of this to the attention of the reader, we have added the above example into our manuscript. It can be found on page 25:

“Of the three studies that found a statistically significant effect, one was rated as weak in quality (Feng et al [54]), one was rated moderate (Morris et al [16]) and one was rated as strong (Godard [56]), while the remainder of the three studies were all rated as moderate.”

Comment #4

Introduction page 3: I do not understand the sentence: Coupled with the rise in labour force participation of middle-aged and older workers are unique challenges such as lower incidence of continued education and skills training. Perhaps you could elaborate why this is relevant.

We thank the reviewer for giving us an opportunity to clarify. We wanted to set the stage for why we should give attention to employment transitions in the age group (adults middle-aged and older) that this review targets. Despite the rise in labour participation of middle-aged and older workers, there are barriers for them to remain in the labour force. We do this by highlighting a few barriers that this age group encounters in the labour market.

The barrier that has been highlighted here is related to continuing education and skills development. Given the rapid advancement of technology, there is a need for workers to constantly maintain and even upgrade their skills over their working lives. Generally, middle-aged and older workers have lower participation rates in continued training compared to workers who are younger. There is also a disparity between employed and unemployed middle-aged and older workers where employed individuals have a higher participation rate – likely attributable to employer-sponsored training programs. This puts middle-aged and older workers at risk for job loss initially and at a slower rate of re-entry into the labour market compared to young age groups.

Matteo Picchio, 2015. "Is training effective for older workers?," IZA World of Labor, Institute of Labor Economics (IZA), pages 121-121, January.

Organisation for Economic Co-operation and Development. Employment Outlook 1998. Paris; 1998 Jun. Available: https://doi.org/10.1787/empl_outlook-1998-en

Comment #5

Methods page 5: The sentence: Screening, assessment, and inclusion of studies published in English, French, and Chinese. I feel like there is something missing to this sentence?

We have corrected this incomplete sentence. The sentence now reads:

“Screening, assessment, and inclusion were limited to studies published in English, French, and Chinese.”

Comment #6

Results page 7: you write body weight index, but it should be body mass index.

Same page: The sentence: odds or risk of becoming overweight or obese. There is a consensus within the research field of obesity that obesity is not something you are or become – it is a condition that you have. Like a disease. Whatever disease a person may have, it may not define them as persons or individuals. For example, having a chronic disease such as cancer does not make a person identify as a cancerous person. I would rewrite the sentence: odds or risk of overweight or obesity. Please go through the manuscript to make sure you use person-first language throughout.

We agree with the review to use first-person language and we have reviewed our manuscript to ensure this is the case throughout. Thank you for noting this.

The BMI typographical error has been corrected.

Comment #7

Table 2: It should be outcome(s) measures instead of outcome(s) measured. Sometimes you use a full stop and sometimes you don´t.

We changed this to Outcome measures since double plural is redundant.

We have removed the use of full stops and corrected inconsistent capitalization throughout Table 3 (formerly Table 2).

Comment #8

A few times (e.g. on the bottom of page 19) you describe women who continued employment or were not otherwise retired/unemployed. Please elaborate what you mean by not otherwise retired/unemployed.

We thank the reviewer for giving us an opportunity to clarify. The comparator group differed slightly between studies. While most studies only had continuously employed individuals serving as their comparator, Pedron et al additionally included those who were unemployed, homemakers, and those unemployed due to sickness in their “not retired” group. Similarly, Zheng included unemployed in their “not retired” group. This is an excerpt from our Discussion that raises this difference and how one study conducted a sensitivity analysis to address it:

“For example, the inclusion of those who are long-term unemployed in a “not retired” group may not make it a suitable comparator as long-term unemployment may be associated with chronic illnesses that have consequences on body weight; however, Pedron et al [51] showed that their results were statistically robust even with different definitions of their comparator.”

Comment #9

Results page 24: Suggestion for rephrasing the sentence: suggests that there may be a mediating role that sleep has in the impact of employment transitions on body weight change… to: suggests that sleep may play a mediating role in the relationship between employment transitions and body weight change…

Same page: you write: dietas outcomes. I have never come across this word before. Is it a typo or is it a real phrase?

We have updated the manuscript with the suggested rewrite of that sentence. The sentence now reads:

“Despite a lack of a formal analysis of mediation, Monsivais et al [12] suggests that sleep may play a mediating role in the relationship between employment transitions and body weight change for future research to consider.”

We have also corrected the typographical error identified from “dietas” to “diet as”.

Comment #10

Discussion page 26: you write: However, this linear rise in BMI may not be true for other countries. Perhaps you should mention here where the study is conducted, as this is not fresh in the readers memory.

We have updated the manuscript to specify the country that the Zheng study was conducted in. The sentence now reads:

“However, this linear rise in BMI may not be true for settings outside of the USA.”

Comment #11

Discussion page 27: you mention that operationalization of weight gain as a categorial or continuous variable has influenced the results in the studies by Zheng and by Forman-Hoffman et al. It seems that you only describe changes in BMI as a continuous variable and do not mention the categorical changes? I do not see from your discussion how categorical or continuous weight change has influenced the results.

Additionally, in the following section, you mention a study examining weight change as both categorical and continuous (Paige et al.); however, you only mention the results for the continuous variable.

I am a bit confused what you wish to discuss with these two paragraphs, and I do not think that the message is coming across very clear. Perhaps you should consider rewriting these paragraphs.

We thank the reviewer for identifying passages in our manuscript where we can improve the clarity of the writing. To clarify what we intended to convey, Zheng analyzed BMI change on a continuous scale while Foreman-Hoffman et al used a categorical outcome variable based on thresholds of 5% gain, 5% loss, or no change (between -5% and +5%). Both studies used a very similar sample from the Health and Retirement Study from the US.

In Zheng’s study, they had found men who retired from active occupations gained a statistically significant amount of weight; however, the actual gain is very small. In the Forman-Hoffman et al paper, they did not find that retired men have greater odds of weight gain of 5% of more compared to men who continued to work. We wished to summarize these two findings in a single sentence originally, but we acknowledge that it may have been written in an unclear manner originally:

“as was the case with Zheng [14] and Forman-Hoffman et al [17] where men experienced increases in BMI but not beyond a threshold of a 5% change in BMI as their weight outcome”.

With that sentence, we wished to highlight that differences in how outcomes are defined, will impact whether a statistically significant effect is detected or not, and even if one is detected, it may not be clinically meaningful. Neither paper operationalized their outcome in the other way, which is why we included some discussion based on a paper in another topic area that does try to capture change in a few different ways (Paige et al).

The Paige et al paper defined weight change on a continuous scale (kg) and as categories (1kg gain, 1kg loss, “no change”). Their finding on the continuous scale was that education was not associated with weight change. Their finding using multinomial logistic regression was that a higher educational attainment was associated with a lower likelihood of weight gain of 1kg and weight loss of 1kg (both ORs smaller than 1, with the “no change” group being the reference). This is what we wished to convey with this original sentence:

“The authors found no statistically significant or clinically meaningful change in body weight when the continuous variables were modeled [65]; however, they found that those with higher educational attainment had lower risk of both weight gain and loss of more than 1kg compared to those with no school certifications [65].”

To improve the clarity of our messaging, we have rewritten parts of these passages to be much more explicit in describing the definition of outcomes in both studies, and to minimize conflating results into single sentences.

“Operationalization of weight gain as a categorical or as a continuous variable, however, does seem to influence whether a statistically significant finding is detected, as was the case with Zheng [14] and Forman-Hoffman et al [17]. Change in BMI was measured on a continuous scale in Zheng [14]while change was categorized as 5% or greater gain, 5% or greater loss, or no change in Forman-Hoffman et al [17]. A statistically significant increase in BMI was found among retired men in Zheng [14], but the odds ratio for 5% or greater gain was not significant in Forman-Hoffman et al [17], suggesting men may gain weight following retirement that does not exceed clinical thresholds. Indeed, many of the findings of weight gain measured on a continuous scale are below clinically significant thresholds.”

Submitted filename: Response to Reviewers May 22 2022.docx

Click here for additional data file.

5 Aug 2022

A systematic review of evidence on employment transitions and weight change by gender in ageing populations

PONE-D-21-39118R1

Dear Dr. Tam,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

George Vousden

Staff Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: Yes: Maja Bramming

**********

10 Aug 2022

PONE-D-21-39118R1

A systematic review of evidence on employment transitions and weight change by gender in ageing populations

Dear Dr. Tam:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. George Vousden

Staff Editor

PLOS ONE