Lifestyle Changes in Relation to Initiation of Antihypertensive and Lipid-Lowering Medication: A Cohort Study.

Background Lifestyle modification is a key component of cardiovascular disease prevention before and concurrently with pharmacologic interventions. We evaluated whether lifestyle factors change in relation to the initiation of antihypertensive or lipid-lowering medication (statins). Methods and Results The study population comprised 41 225 participants of the FPS (Finnish Public Sector) study aged ≥40 years who were free of cardiovascular disease at baseline and responded to ≥2 consecutive surveys administered in 4-year intervals in 2000-2013. Medication use was ascertained through pharmacy-claims data. Using a series of pre-post data sets, we compared changes in body mass index, physical activity, alcohol consumption, and smoking between 8837 initiators and 46 021 noninitiators of antihypertensive medications or statins. In participants who initiated medication use, body mass index increased more (difference in change 0.19; 95% CI, 0.16-0.22) and physical activity declined (-0.09 metabolic equivalent of task hour/day; 95% CI, -0.16 to -0.02) compared with noninitiators. The likelihood of becoming obese (odds ratio: 1.82; 95% CI, 1.63-2.03) and physically inactive (odds ratio: 1.08; 95% CI, 1.01-1.17) was higher in initiators. However, medication initiation was associated with greater decline in average alcohol consumption (-1.85 g/week; 95% CI, -3.67 to -0.14) and higher odds of quitting smoking (odds ratio for current smoking in the second survey: 0.74; 95% CI, 0.64-0.85). Conclusions These findings suggest that initiation of antihypertensive and statin medication is associated with lifestyle changes, some favorable and others unfavorable. Weight management and physical activity should be encouraged in individuals prescribed these medications.

Clinical Perspective

What Is New?

Few longitudinal studies have examined whether initiation of preventive medication affects lifestyle, either increasing or decreasing the likelihood of unhealthy behaviors.

What Are the Clinical Implications?

Because initiation of antihypertensive or statin therapy appears to be associated with some negative lifestyle changes, expansion of pharmacologic interventions toward populations at low cardiovascular disease risk may not necessarily lead to expected benefits at the population level.

Effective measures are needed to support the recommended lifestyle change in relation to the initiation of pharmacologic interventions for primary prevention.

Introduction

Use of antihypertensive medications and statins for the primary prevention of cardiovascular disease (CVD) has expanded over the past few decades1, 2, 3 and may increase further in response to recent guideline changes.4, 5, 6 However, lifestyle modification remains a key component of CVD prevention before and concurrently with pharmacologic interventions.4, 5, 7, 8

Whether and how initiation of preventive medication affects lifestyle remains unclear. Because lifestyle counseling should precede prescription of preventive medication,8 medication users may have better perception of their CVD risk9, 10 and a healthier lifestyle compared with nonusers.11, 12 The perceived effectiveness of pharmacotherapy may provide an incentive to adhere to a lifestyle that also prevents other diseases.13, 14 The counterargument is that individuals may substitute medication for a healthy lifestyle and continue an unhealthy lifestyle or even engage in it,14 which may reduce the effectiveness of pharmacotherapy.15, 16

The diagnosis of CVD or diabetes mellitus has been found to trigger lifestyle changes, including weight loss,17 increase in physical activity,18 decrease in alcohol consumption,18 and smoking cessation.17, 19 Participants of the Danish Inter99 trial with hypertension or hyperlipidemia who initiated medication for these conditions were more likely to engage in favorable dietary changes compared with noninitiators during 5‐year follow‐up.20 Statin initiators also increased their physical activity. Recently, however, studies have reported lower physical activity among people who initiate21 or use statins14, 22 or antihypertensive medications10 compared with nonusers. According to the US NHANES (National Health and Nutrition Examination Survey) data,23 in the early 2000s statin users had lower caloric and fat intake compared with nonusers. The difference disappeared by the mid‐2000s and was reversed by 2010. Most recent findings suggested statin users consumed more calories and fat than nonusers. Simultaneously, the prevalence of obesity increased faster among users.22, 23

We used repeat observational data to assess the extent to which initiation of antihypertensive or lipid‐lowering (statin) therapy predicts changes in lifestyle factors, such as body mass index (BMI), leisure‐time physical activity, alcohol consumption, and smoking, in a large cohort of Finnish adults. We sought to determine whether initiation of preventive medication is more likely to complement a healthy lifestyle change or substitute for it. To obtain robust estimates, we analyzed data as a series of pre–post studies.24, 25

Methods

Study Population and Design

Participants were drawn from the FPS (Finnish Public Sector) study cohort of employees of 10 towns and 6 hospital districts.25, 26 Data were sourced from questionnaire surveys administered to FPS subcohorts in 4‐year intervals from 2000 to 2013 (average response rate 70%),25 employer records, and national health registers. We included participants who responded to ≥2 consecutive surveys (waves 1 and 2) and were aged ≥40 years and free of CVD at wave 1. Individual‐level survey data could not be made publicly available, but information on the data and analyses are available upon request to the corresponding author.

We constructed a long‐form data set using subsets of the original data and organized them into 3 pre–post sets with separate baselines (2000–2002, 2004–2005, or 2008–2009) and corresponding follow‐ups (2004–2005, 2008–2009, or 2012–2013, respectively). In all, 41 225 participants belonged to these data sets: 15 880 to 1 data set, 10 143 to 2 data sets, and 15 202 to 3 data sets.

According to Finnish law, written consent is not required for register‐based and survey research as long as the participation is voluntary. The FPS study participants were informed about the aims of the study and the record linkage. The ethics committees of Helsinki University Hospital and the Finnish Institute of Occupational Health approved the FPS study protocol.

Initiator, Prevalent User, and Nonuser Cohorts of Preventive Medication

We used the participants’ personal identification numbers (a unique number assigned to each Finnish citizen) to link participants to the electronic records of the Social Insurance Institution (SII) of Finland covering the years 1994–2011. Records of pharmacy claims were used to ascertain medication use. We identified antihypertensive medications with Anatomical Therapeutic Chemical (ATC) codes C02 (antihypertensives), C03 (diuretics), C07 (β‐blockers), C08 (calcium channel blockers), C09 (renin–angiotensin system blockers), and C10AA and C10BA02 (both for statins).

We defined initiators as those who filled ≥1 prescription for any preventive medication for the first time after their baseline response date but before their wave 2 response date. Participant observations with no prescription fills by wave 2 were defined as noninitiators. We focused on comparing initiators of either antihypertensive medications or statins with those who remained noninitiators of both. For secondary comparisons, we included prevalent users, defined as those participant observations for which ≥1 prescription for any preventive medication was filled before or on the baseline response date.

Lifestyle Factors

Height, weight, physical activity, average weekly alcohol consumption, and smoking status were assessed using standard questionnaires. Based on self‐reported height and weight, we calculated BMI (kg/m2) and defined BMI 25 to <30 as overweight and BMI ≥30 as obesity. Physical activity was assessed using the following questions26: “How much did you exercise per week on average during the past year?” and “Estimate whether the level of intensity of the exercise corresponded to walking, vigorous walking, jogging, or running.” The number of hours per week spent on activity at each intensity level was multiplied by the average energy expenditure of each activity, expressed in metabolic equivalent of task (MET) hours. We used the sum of MET hours to identify active (>4 h/day), moderately active (2–4 h/day), and inactive participants (<2 h/day).27 Drinking habits were requested as follows28: “Have you ever consumed at least one glass of some alcoholic drink?” and “How much do you consume of the following alcoholic drinks on average (beer per week; wine or other mild drinks per week; spirits per month)?” We combined alcohol consumption into 1 measure of average weekly consumption (grams of pure alcohol). High alcohol consumption equated to consuming >240 g/week in men and >160 g/week in women.29 In addition, we considered answering “yes” to the following question as an indicator of high consumption: “Have you ‘passed out’ due to drinking during the past year?” Any other consumption was considered moderate. Smoking status was assessed with the following questions30: “Do you smoke or have you previously smoked regularly, that is daily or nearly daily?” and “Do you still smoke regularly?” Those responding “yes” to the second question were categorized as current smokers at each survey wave. Those who responded “yes” to the first and “no” to the second questions were categorized as former smokers, and those who responded “no” to both of the questions were categorized as never smokers. To depict clustering of lifestyle‐related risks, we calculated the number of lifestyles at the unhealthiest level (obesity, physical inactivity, high alcohol consumption, current smoking).31

The outcome measures included changes in continuous BMI, physical activity, alcohol consumption, and number of cigarettes smoked per day between the waves and the presence of obesity, physical inactivity, high alcohol consumption, current smoking, or their clustering (3 or 4 unhealthy lifestyles) at wave 2.

Covariates

Covariates were assessed at baseline. Information on sex, age, employment, and occupational status (upper grade nonmanual workers, lower grade nonmanual workers, clerks, or manual workers) came from employer records. Information on marital status and, for those who had left the FPS organizations, employment status came from survey responses. Nationwide hospital discharge and SII special reimbursement registers were used to ascertain history of CVD (acute myocardial infarction, chronic coronary heart disease, chronic heart failure, and stroke) at baseline and between the waves. Information on special reimbursements and prescription fills (A10) were used to ascertain history of diabetes mellitus.

Statistical Analyses

The results from the data sets were pooled. We modeled each lifestyle factor as a separate outcome with a generalized estimating equation approach accounting for within‐person correlation across data sets and applied bootstrapping to obtain stable 95% CIs.

We estimated the associations between medication initiation and continuous outcomes using linear regression. Associations with unhealthy lifestyles at wave 2 were estimated with logistic regression, stratifying by presence of the respective unhealthy lifestyle at baseline. To account for the changes in clinical guidelines and prescribing and the general lifestyle trends, all models included baseline period and time interval between the survey waves, as this interval influences the potential for change in lifestyles. Fully adjusted models included age; sex; marital, occupational, and employment status; diabetes mellitus; and other lifestyle factors measured at baseline. Participant observations that were missing data on any covariate were excluded from the multivariable models.

Sensitivity and Subgroup Analyses

First, we excluded participant observations with diabetes mellitus at baseline and those diagnosed with diabetes mellitus or CVD between waves 1 and 2 because our focus was on primary prevention. Second, to identify true medication initiators, we redefined initiation as refilling a prescription for preventive medication at least once within a year since initiation. Third, to confirm the temporal order between medication initiation and lifestyle change, we restricted our cohort to respondents of 3 consecutive surveys and compared the presence of unhealthy lifestyles at wave 3 between initiators and noninitiators who maintained the same lifestyle at waves 1 and 2.25 Fourth, we repeated the main analyses separately for antihypertensive medications and statins to see whether the results were specific to a therapeutic class. Last, we stratified the analyses by baseline period and sex. We tested statistical significance of the interactions by including user‐group‐by‐period or user‐group‐by‐sex interaction terms in the outcome models, respectively. Analyses were performed using SAS v9.4 (SAS Institute).

Results

Of a total of 81 772 participant observations based on the 41 225 FPS participants, 84% were female, with mean age of 52±7.2 years (Table 1); and 8837 (10.8%) presented as initiators of preventive medication, 26 914 (32.9%) as prevalent users, and 46 021 (56.3%) as noninitiators between waves 1 and 2 (mean interval: 4.0 years; range: 2.0–5.4 years). Medication initiation took place a median of 1.7 years (interquartile range: 0.9–2.7 years) after baseline and a median of 2.4 years (interquartile range: 1.3–3.3 years) before the second survey wave. Between waves, 318 (3.6%) initiators, 477 (1.8%) prevalent users, and 48 (0.1%) noninitiators were newly diagnosed with CVD.

Table 1

Baseline Characteristics of the 81 772 Participant Observations

Characteristic	All Observations (n=81 772)	Initiators (n=8837)	Prevalent Users (n=26 914)	Noninitiators (n=46 021)
Baseline year, %
2000–2002	25.8	27.2	18.3	29.9
2004–2005	33.4	42.0	31.5	32.8
2008–2009	40.8	30.8	50.2	37.4
Age, y, mean (SD)	52.2 (7.24)	52.4 (6.61)	55.1 (7.24)	50.4 (6.78)
40–49	41.7	38.0	25.7	51.7
50–59	43.1	49.0	48.4	38.8
60–78	15.3	13.0	25.9	9.5
Women, %	83.9	83.8	82.6	84.3
Married/cohabiting, %	76.3	76.5	75.6	76.6
Occupational status, %
Upper grade nonmanual workers	32.1	29.8	30.6	33.3
Lower grade nonmanual workers	26.5	26.1	25.4	27.2
Clerks	7.4	7.2	8.3	6.9
Manual workers	34.1	36.9	35.7	32.6
Employed, %	89.4	92.3	81.0	93.7
Body weight, %
Normal	49.8	43.0	37.2	58.3
Overweight	35.5	39.4	39.8	32.3
Obese	14.7	17.6	23.0	9.4
Physical activity, %
Active	40.8	38.1	34.5	45.0
Moderate	33.6	33.5	34.7	33.0
Inactive	25.6	28.4	30.9	22.0
Alcohol consumption, %
Moderate	71.7	70.9	69.5	73.2
None	14.9	14.3	16.5	14.1
High	13.4	14.8	14.1	12.7
Smoking, %
Never	68.2	65.1	68.5	68.6
Former	18.3	19.7	19.5	17.4
Current	13.4	15.1	12.0	14.0
No. of unhealthy lifestyles,a %
0	52.0	47.0	45.4	56.8
1–2	45.2	49.3	50.7	41.3
3–4	2.8	3.6	3.9	2.0
History of antihypertensive medication use, %	29.3	NA	89.1	NA
History of statin use, %	9.6	NA	29.1	NA
Diabetes mellitus, %	2.7	2.0	6.7	0.6

The observations are based on the 41 225 participants of the FPS (Finnish Public Sector) study. Numbers are percentages unless otherwise specified. Of all participant observations included, 79 889 (97.7%) had data available on body mass index, 80 988 (99.0%) on physical activity, 81 337 (99.5%) on alcohol consumption, and 79 675 (97.4%) on smoking at baseline. NA indicates not applicable.

Obesity, physical inactivity, high alcohol consumption, and current smoking.

Change in Lifestyle

Average BMI increased among both initiators and noninitiators (Table 2), but in the fully adjusted model, the mean increase in BMI was larger (difference in change: 0.19; 95% CI, 0.16–0.22) among initiators. Among wave 1 obese participants, the odds of obesity at wave 2 were 1.37 times greater (95% CI, 1.15–1.65) for initiators compared with noninitiators (Figure). Among wave 1 nonobese participants, the odds of obesity at wave 2 were 1.82 times greater (95% CI, 1.63–2.03) for initiators versus noninitiators. Furthermore, average MET h/day declined among initiators versus noninitiators (−0.09; 95% CI, −0.16 to −0.02; Table 2). Initiators had greater odds of physical inactivity at wave 2 regardless of baseline activity (Figure). Table S1 shows the shifts of participant observations between levels of lifestyle factors within user groups.

Table 2

Changes in BMI, Physical Activity, and Alcohol Consumption Between the Survey Waves in Initiators and Prevalent Users of Preventive Medication Versus Noninitiators

Lifestyle Factor/User Group	Outcome
	Mean (SDa)		Mean Change (P Valuea)	β Value (95% CI)b
	Baseline	Second Survey		Unadjusted Differencec	Adjusted Differenced
BMI, kg/m2
Initiators (n=8425)	26.3 (4.2)	26.8 (4.5)	0.51 (<0.001)	0.13 (0.08–0.17)	0.19 (0.16–0.22)
Prevalent users (n=25 348)	27.0 (4.6)	27.3 (4.7)	0.31 (<0.001)	−0.04 (−0.07 to 0.01)	0.08 (0.06–0.10)
Noninitiators (n=44 091)	24.9 (3.7)	25.2 (3.9)	0.37 (<0.001)	Reference	Reference
Physical activity (MET h/d)
Initiators (n=8658)	4.15 (3.77)	3.97 (3.60)	−0.18 (<0.001)	−0.13 (−0.22 to −0.04)	−0.09 (−0.16 to −0.02)
Prevalent users (n=26 368)	3.78 (3.44)	3.69 (3.38)	−0.09 (<0.001)	−0.12 (−0.17 to −0.06)	−0.08 (−0.13 to −0.04)
Noninitiators (n=45 312)	4.71 (4.00)	4.70 (4.03)	−0.01 (0.47)	Reference	Reference
Alcohol consumption, g/wk
Initiators (n=8721)	69.0 (108)	66.4 (110)	−2.64 (0.011)	−2.57 (−4.75 to −0.38)	−1.85 (−3.67 to −0.14)
Prevalent users (n=26 546)	66.7 (109)	63.9 (111)	−2.38 (<0.001)	−1.73 (−3.13 to 0.34)	−0.72 (−1.92 to 0.49)
Noninitiators (n=45 488)	61.1 (93.2)	60.6 (95.6)	−0.44 (0.26)	Reference	Reference
No. of cigarettes/daye
Initiators (n=876)	13.1 (7.05)	12.5 (6.83)	−0.60 (<0.001)	−0.37 (−0.63 to −0.11)	−0.34 (−0.60 to −0.08)
Prevalent users (n=2251)	12.3 (6.65)	12.1 (6.71)	−0.23 (0.016)	−0.02 (−0.20 to 0.16)	0.05 (−0.13 to 0.24)
Noninitiators (n=4532)	11.9 (6.60)	11.7 (6.45)	−0.19 (0.002)	Reference	Reference

Only participant‐observations who had data available on all baseline covariates were included in adjusted models (n=74 291 or 95.4% of those with data on body mass index at both surveys, n=75 613 or 94.1% of those with data on physical activity at both surveys, n=75 781 or 93.8% of those with data on alcohol consumption at both surveys, n=7339 or 95.8% of those reporting number of cigarettes smoked at both surveys). BMI indicates body mass index; MET, metabolic equivalent of task.

SDs and P values reported here do not take into account that each individual may have contributed to >1 data set.

Derived from generalized estimating equations.

Adjusted for baseline year and time interval between the survey waves.

Adjusted for baseline year; time interval; age; sex; marital, occupational, and employment status; diabetes mellitus; smoking status; and other lifestyle factors.

Among those who were current smokers at both waves.

Figure 1

Adjusted odds ratios (ORs) for unhealthy lifestyles at the second survey for users of preventive medication (vs noninitiators) by presence of the respective unhealthy lifestyle at baseline. Number of participant observations and prevalence of unhealthy lifestyle at the second survey in parentheses.

Average weekly alcohol consumption declined more among initiators compared with noninitiators (−1.85 g/week; 95% CI, −3.67 to −0.14; Table 2). When the comparison was restricted to those initiators (n=7014) and noninitiators (n=37 297) who reported any alcohol consumption at both waves, the adjusted difference in change in weekly consumption was −1.67 g (95% CI, −3.72 to 0.39). No difference in the odds of high alcohol consumption at wave 2 was observed between user groups regardless of baseline drinking (Figure).

The prevalence of current smoking tended to decrease in all subgroups (Table 1, Table S1). However, baseline smokers who initiated medication use were more likely to quit smoking compared with smokers who remained untreated (adjusted odds ratio [OR] for current smoking among initiators versus noninitiators: 0.74; 95% CI, 0.64–0.85; Figure). Furthermore, initiators who were current smokers at both waves decreased smoking more than their noninitiator counterparts (adjusted difference in change in the number of cigarettes per day: −0.34; 95% CI: −0.60 to −0.08; Table 2).

Among those with 0–2 unhealthy lifestyles at baseline, the adjusted OR of a cluster of 3 to 4 unhealthy lifestyles at wave 2 was 1.66 (95% CI, 1.40–1.98) for initiators compared with noninitiators (Figure). The relative difference between user groups was smaller in the subgroup that already had 3 to 4 unhealthy lifestyles at baseline. In this subgroup (n=1231), BMI increased (adjusted difference: 0.32; 95% CI, 0.06–0.59) and MET h/day declined (−0.33; 95% CI, −0.57 to −0.09) more among initiators compared with noninitiators, whereas there was no significant difference in change in average alcohol consumption (0.09 g/week; 95% CI, −17.6 to 17.8) or in the odds of current smoking at wave 2 (adjusted OR: 0.92; 95% CI, 0.69–1.24).

Prevalent users’ lifestyle changed in the same direction as initiators’ lifestyle, although the average changes in BMI and weekly alcohol consumption were smaller in comparison with noninitiators (Table 2). Prevalent users were more likely to develop obesity, physical inactivity, and high alcohol consumption than were noninitiators (Figure). Within the subgroup with 0 to 2 unhealthy lifestyles at baseline, in particular, prevalent users had twice the odds of having 3 to 4 unhealthy lifestyles at wave 2 compared with noninitiators (adjusted OR: 2.02; 95% CI, 1.78–2.28).

Exclusion of participants with diabetes mellitus or new CVD (Tables S2 and S3) or initiators with no refills within the first year since initiation (Tables S4 and S5) did not materially change the main results. The analyses including initiators and noninitiators who maintained the same lifestyle at waves 1 and 2 showed that initiators were more likely to develop obesity and a cluster of unhealthy lifestyles by wave 3 (Table S6). These analyses support the temporal order between medication initiation and unhealthy lifestyle change.

The associations between antihypertensive medication use and lifestyle changes resembled those in the main analyses (Tables S7 and S8). Conversely, the associations for statin initiation were generally weaker. However, statin initiation appeared to be more strongly associated with decline in average weekly alcohol consumption (−2.38 g; 95% CI, −4.33 to −0.43) and quitting smoking (OR for current smoking: 0.63; 95% CI, 0.53–0.75). Almost half (43.9%) of initiators and 24.3% of noninitiators of statins were users of antihypertensive medication at wave 1 (Table S9). In addition, 7.6% of initiators of antihypertensive medication and 4.7% of noninitiators had used statins.

The baseline period did not appear to modify the associations between user group and changes in BMI, alcohol consumption, or physical activity or between user group and odds of any of unhealthy behaviors at wave 2 (Tables S10 and S11; P>0.05 for user group‐by‐period interactions). Sex did not modify the associations between user group and change in BMI or weekly alcohol consumption or between user group and odds of any of unhealthy behaviors at wave 2 (Tables S12 and S13; P>0.05 for user‐group‐by‐sex interactions). The average decline in physical activity was of the same magnitude in both sexes (Table S12). Among women, however, physical activity declined more among initiators than noninitiators (adjusted difference in change: −0.11 MET h/day; 95% CI, −0.19 to −0.04). Among men, no difference was found (adjusted difference in change: 0.01 MET h/day; 95% CI, −0.19 to 0.20; P=0.20 for user group‐by‐sex interaction). In the subgroup of physically inactive participants at baseline, the odds of physical inactivity associated with medication initiation were increased among women (OR: 1.20; 95% CI, 1.08–1.32) but not in men (OR: 1.01; 95% CI, 0.81–1.26; P=0.21 for user‐group‐by‐sex interaction). Among those at least moderately active at baseline, no such sex difference was observed (OR, women: 1.10 [95% CI, 1.00–1.22]; men: 1.09 [95% CI, 0.91–1.30]).

Discussion

In this large cohort study of Finnish adults free of CVD, there are 4 key findings. First, initiators experienced greater increases in BMI and were more likely to become obese than did noninitiators. This attenuates positive treatment effects and supports the substitution hypothesis. Second, the likelihood of becoming physically inactive was higher among initiators, a finding supporting the substitution hypothesis. Third, in support of the complementation hypothesis, smokers who initiated preventive medication were more likely to either quit or decrease smoking compared with untreated smokers. This is likely to amplify the effectiveness of pharmacotherapy. Fourth, although average alcohol consumption decreased more among initiators than noninitiators, there was no difference in the odds of heavy drinking.

This was the first large‐scale, prospective, cohort study on the associations between use of preventive medications and lifestyle change. Prior research on the relationships between preventive medication use and lifestyle has typically considered either use of antihypertensive medications10, 32 or statins9, 11, 14, 21, 22, 23, 33 but rarely both.12, 20 Most previous studies have been cross‐sectional9, 11, 14, 22, 23, 33 and provide no information on potential lifestyle changes during the time window around the initiation of medication use. We included both antihypertensive and statin medications and also studied them separately. Using individual‐level data with repeated measurements, we were able to determine each individual's lifestyle before and after medication initiation.

Our results on greater declines in smoking and increases in obesity among initiators of preventive medication agree with those from a small Canadian cohort study involving 769 new hypertensive participants of the National Population Survey (1994–2002) who also reported use of antihypertensive medication.32 Another longitudinal study based on the Framingham Heart Study Offspring Cohort (1983–2001) showed an increase in obesity among statin users but no clear decrease in smoking.14 Furthermore, statin use was associated with an increase in moderate alcohol consumption and physical activity in men but not in women. In addition, a study including 871 participants of the Inter99 health behavior intervention trial reported an association between increased physical activity and likelihood of initiating statin use over a 5‐year period (up to 2006),20 but no association was observed with initiation of antihypertensive medication. These results are consistent with the observations in our study population, the majority of whom started preventive therapy with antihypertensive medications. Together with the preceding observations and data from repeated cross‐sectional surveys from the United States23 and Korea22 showing faster increases in fat and caloric intake, obesity, and physical inactivity among statin users versus nonusers, our study provides more support for the substitution hypothesis (unfavorable or no lifestyle change) than for the complementation hypothesis (positive lifestyle change).

Statin initiation is associated with increase in muscle complaints,34 potentially interfering with physical activity. A US study among older men reported that prevalent statin use was associated with modestly lower physical activity.21 However, a longitudinal analysis showed that physical activity declined faster among initiators than among prevalent users or noninitiators. We observed virtually no differences in change in physical activity between initiators, prevalent users, and nonusers of statins in our cohort, which was younger, predominantly female, and free of CVD.

It is well documented that smoking cessation often results in weight gain, particularly in women, that cannot be prevented by increasing physical activity.35 In our data, the average increase in BMI among quitters was 1.20 (SD: 2.05) for initiators and 1.09 (SD: 2.07) for noninitiators (data not shown). Although initiation of preventive medication was associated with quitting smoking, the difference in the increase of BMI remained approximately the same for initiators and noninitiators after excluding the quitters (adjusted difference: 0.18; 95% CI, 0.15–0.22).

Limitations

A key limitation of our study is the generalizability of the results. Our study population consisted of a relatively homogenous sample of white public sector workers, with only 16% of the participant observations related to men and the mean age being slightly >50 years. Furthermore, during the study period, intensive public health action took place in Finland36 and may have affected the lifestyles of both medication users and nonusers. For example, smoking in public indoor places and workplaces was prohibited already in the 1990s and in restaurants in 2006. The national development program for the prevention and care of diabetes mellitus was implemented in 2000–2010.37 This program aimed to increase awareness of diabetes mellitus and its risk factors (including the same lifestyle factors considered in this study) in the whole population through various activities, such as media campaigns and health fairs. Consequently, our results may not be generalizable to countries or settings with different preventive policies.

Furthermore, initiators of preventive medication have presumably received counseling from health professionals before initiation, as the clinical guidelines have recommended lifestyle change lasting for weeks or months as the sole intervention among individuals in primary prevention.29, 38 Unfortunately, we have no information on the contents and timing of this counseling or other forms of health promotion to which the medication users in this study were exposed. Because some participants may have initiated lifestyle change before baseline, we may have underestimated the change related to medication initiation. However, high‐risk behaviors were more common among initiators than noninitiators at baseline; therefore, there was more room for improvement.

We did not have information on each participant's diet (eg, fat or fruit and vegetable intake), blood pressure, or cholesterol levels. Our noninitiator population is likely to be a mixture of patients free of hypertension or dyslipidemia and those who may have been diagnosed with these conditions but did not use medications; that is, a small proportion of noninitiators may be individuals among whom lifestyle modification alone was effective. Our definition of medication use was based on dispensation data, and we could not verify whether and how the dispensed medications were consumed. Self‐reporting has been found to underestimate obesity,39 smoking,40 and alcohol consumption.41 However, this is unlikely to have changed between waves, and each lifestyle factor has been shown to be associated with increased all‐cause mortality and CVD event risk in these data, except for high alcohol consumption.26

In sensitivity analyses, we excluded participant observations linked to records indicating myocardial infarction, chronic coronary heart disease, chronic heart failure, or stroke between the survey waves and those indicating diabetes mellitus at any time before the second wave. This did not alter observed associations. However, other conditions such as respiratory events, cancers, or chronic gastrointestinal disease could also affect lifestyle factors and, if distributed unevenly between initiators and noninitiators of preventive medication, may have confounded the associations between medication initiation and lifestyle change. Furthermore, it is possible that the associations between medication initiation and increases in BMI were confounded by menopausal status (among women) or age‐related weight gain. When we analyzed these associations within 10‐year age groups, BMI seemed to increase more among initiators than noninitiators in each age group, but the association between medication initiation and change in BMI was stronger among those aged 40 to 49 years (adjusted difference in change in BMI: 0.30; 95% CI, 0.22–0.37) than among the older age groups in the whole study population (adjusted difference in change in BMI, 50–59 years: 0.11 [95% CI, 0.05–0.17]; ≥60 years: 0.13 [95% CI, 0.02–0.24). This was true for women and men (data not shown); therefore, our conclusions did not change.

Our findings support the notion that there is scope to improve management of lifestyle‐related risk factors among individuals who have initiated preventive medication. Patients’ awareness of their risk factors alone seems not to be effective in improving health behaviors.42 The most recent primary prevention guidelines in Europe8 and the United States43 emphasize cognitive‐behavioral strategies (eg, motivational interviewing, shared decision‐making), multidisciplinary and team‐based approaches, and evaluation of social and other individual‐level determinants of health when choosing the optimal strategy for each patient.

Perspectives

Initiation of antihypertensive or statin therapy appears to be associated with lifestyle changes, some positive and others negative. This means that expansion of pharmacologic interventions toward populations at low CVD risk may not necessarily lead to expected benefits at the population level. More effective measures are needed to support the recommended lifestyle change in relation to the initiation of pharmacologic interventions for primary prevention.

Sources of Funding

This work was funded by the Academy of Finland (grant no. 138255), NordForsk (the Nordic Research Program on Health and Welfare to Kivimäki), the Academy of Finland (grant no. 311492 to Kivimäki) and Helsinki Institute of Life Science (to Kivimäki), and Hospital District of Southwest Finland (to Korhonen).

Disclosures

Setoguchi received research funding from and was a consultant for Pfizer Inc. and Janssen Research & Development, LLC. The remaining authors have no disclosures to report.

Table S1. Cross‐Tabulation of Unhealthy Lifestyle Factors at Baseline and During the Second Survey Wave by Use of Preventive Medication

Table S2. Changes in Body Mass Index, Physical Activity, and Alcohol Consumption From the Baseline Survey to the Second Survey Wave Among Users of Preventive Medication Versus Noninitiators

Table S3. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Second Survey Wave for Users of Preventive Medication (Versus Noninitiators) Stratified by Presence of the Respective Unhealthy Lifestyle at Baseline

Table S4. Changes in Body Mass Index, Physical Activity, and Alcohol Consumption From the Baseline Survey to the Second Survey Wave Among Initiators With at Least 1 Refill of Preventive Medication Versus Noninitiators

Table S5. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Second Survey Wave for Initiators of Preventive Medication With at Least 1 Refill (Versus Noninitiators) Stratified by Presence of the Respective Unhealthy Lifestyle at Baseline

Table S6. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Third Survey Wave for Initiators of Preventive Medication (Versus Noninitiators) Stratified by Presence/Absence of the Respective Unhealthy Lifestyle at Baseline and Wave 2

Table S7. Changes in Body Mass Index, Physical Activity, and Alcohol Consumption From the Baseline Survey to the Second Survey Wave Among Users of Antihypertensive Medications and Statins Versus Noninitiators

Table S8. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Second Survey Wave for Users of Antihypertensive Medications and Statins (Versus Noninitiators) Stratified by Presence of the Respective Unhealthy Lifestyle at Baseline

Table S9. Baseline Characteristics of the 81 772 Participant Observations Stratified by Use of Antihypertensive Medications and Statins

Table S10. Changes in Body Mass Index, Alcohol Consumption, and Physical Activity From Baseline Survey to the Second Survey Wave Among Users of Preventive Medication Versus Noninitiators, by Baseline Year

Table S11. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Second Survey Wave for Users of Preventive Medication (Versus Noninitiators) Stratified by Presence of the Respective Unhealthy Lifestyle at Baseline, by Baseline Year

Table S12. Changes in Body Mass Index, Physical Activity, and Alcohol Consumption From the Baseline Survey to the Second Survey Wave Among Users of Preventive Medication (Versus Noninitiators), by Sex

Table S13. Odds Ratios for Obesity, Physical Inactivity, High Alcohol Consumption, and Current Smoking at the Second Survey Wave for Users of Preventive Medication (Versus Noninitiators) Stratified by Presence of the Respective Unhealthy Lifestyle at Baseline, by Sex

Click here for additional data file.