Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: Within-treatment analysis in the RAINBOW trial.

INTRODUCTION: The RAINBOW randomized clinical trial validated the efficacy of an integrated collaborative care intervention for obesity and depression in primary care, although the effect was modest. To inform intervention optimization, this study investigated within-treatment variability in participant engagement and progress.
METHODS: Data were collected in 2014-2017 and analyzed post hoc in 2018. Cluster analysis evaluated patterns of change in weekly self-monitored weight from week 6 up to week 52 and depression scores on the Patient Health Questionnaire-9 (PHQ-9) from up to 15 individual sessions during the 12-month intervention. Chi-square tests and ANOVA compared weight loss and depression outcomes objectively measured by blinded assessors to validate differences among categories of treatment engagement and progress defined based on cluster analysis results.
RESULTS: Among 204 intervention participants (50.9 [SD, 12.2] years, 71% female, 72% non-Hispanic White, BMI 36.7 [6.9], PHQ-9 14.1 [3.2]), 31% (n = 63) had poor engagement, on average completing self-monitored weight in <3 of 46 weeks and <5 of 15 sessions. Among them, 50 (79%) discontinued the intervention by session 6 (week 8). Engaged participants (n = 141; 69%) self-monitored weight for 11-22 weeks, attended almost all 15 sessions, but showed variable treatment progress based on patterns of change in self-monitored weight and PHQ-9 scores over 12 months. Three patterns of weight change (%) represented minimal weight loss (n = 50, linear β1 = -0.06, quadratic β2 = 0.001), moderate weight loss (n = 61, β1 = -0.28, β2 = 0.002), and substantial weight loss (n = 12, β1 = -0.53, β2 = 0.005). Three patterns of change in PHQ-9 scores represented moderate depression without treatment progress (n = 40, intercept β0 = 11.05, β1 = -0.11, β2 = 0.002), moderate depression with treatment progress (n = 20, β0 = 12.90, β1 = -0.42, β2 = 0.006), and milder depression with treatment progress (n = 81, β0 = 7.41, β1 = -0.23, β2 = 0.003). The patterns diverged within 6-8 weeks and persisted throughout the intervention. Objectively measured weight loss and depression outcomes were significantly worse among participants with poor engagement or poor progress on either weight or PHQ-9 than those showing progress on both.
CONCLUSIONS: Participants demonstrating poor engagement or poor progress could be identified early during the intervention and were more likely to fail treatment at the end of the intervention. This insight could inform individualized and timely optimization to enhance treatment efficacy. TRIAL REGISTRATION: ClinicalTrials.gov# NCT02246413.

RCT Entities:   Population Interventions Outcomes

Introduction

Obesity and depression are highly prevalent in the United States with associated high personal and societal cost. [1, 2] Currently among US adults, nearly 40% are obese [3] and 19% experience major depression over the course of their lifetime. [4] Subthreshold depression is also common, with increased burden of morbidity and disability. [5, 6] Mounting epidemiologic evidence shows a temporally reciprocal, positive relationship between obesity and depression; namely, people with obesity are more likely to develop new-onset depression or have worsening depressive symptoms, and vice versa. [7-11]

The high co-occurrence of these 2 conditions reveals a critical need for developing effective multimorbidity treatment. Randomized clinical trials (RCTs) of integrated behavior therapy for patients with obesity and depression are limited and have shown mixed results. [12-14] Recently, the Research Aimed at Improving Both Mood and Weight (RAINBOW) trial reported that an integrated collaborative care intervention, as compared with usual care, led to significantly improved weight loss and depressive symptoms through 12 months among primary care patients of both sexes who had obesity and depression. [14] Similar to prior trials showing effectiveness of behavior therapy in either of these conditions alone [15-18] or in related multiple chronic conditions—such as depression and diabetes or coronary heart disease [19]—the magnitude of treatment effects on both weight loss and depression outcomes in the trial were modest.

The modest effects may be caused by the variability in treatment engagement and progress, which is typically high in clinical settings. Examination of this variability can inform optimization—such as when and how to adapt intervention delivery or content for enhanced efficacy—of behavioral interventions. However, research on this topic is lacking, especially in multimorbidity management.

This study reports on post hoc analyses aimed to investigate variability in treatment engagement and progress during the integrated collaborative care intervention among RAINBOW patients with comorbid obesity and depression.

Materials and methods

The Institutional Review Board for Sutter Health, Northern California, approved the study. All participants provided written informed consent. The trial protocol was published previously. [20] The co-primary efficacy outcomes were changes in body mass index (BMI) and Depression Symptom Checklist 20-item (SCL-20) [21, 22] scores objectively obtained by blinded outcome assessors. A total of 409 participants who had both BMI ≥30 (≥27 if Asian) and Patient Health Questionnaire 9-item (PHQ-9) scores ≥10, and no exclusions per protocol, were enrolled in the trial. Participants were randomly assigned to the 12-month I-CARE (ntegrated oching fo Btter Mood and Weight) intervention group (n = 204) or the usual care control group (n = 205). This study analyzed participant data only within the intervention group.

Intervention

The I-CARE intervention integrated a self-directed Group Lifestyle Balance™ (GLB) program for weight loss [23-25] and the Program to Encourage Active, Rewarding Lives for Seniors (PEARLS) program [26, 27] for collaborative stepped depression care. The GLB program [25] was adapted from the Diabetes Prevention Program [28] and provided videos for patient self-study. The PEARLS program used Problem-Solving Therapy (PST) combined with behavioral activation strategies as the first-line approach and, if indicated, therapy was intensified through stepwise increases in doses and number of antidepressant medications. The intervention had a 6-month intensive treatment phase comprising 9 one-on-one in-person visits of 60 minutes each, 11 home-viewed GLB videos of 20 to 30 minutes each, and digital self-monitoring activities; and a 6-month maintenance phase comprising 6 phone calls of 15 to 30 minutes each and continued self-monitoring. Participants met with a health coach weekly for the first 4 sessions, every 2 weeks for the next 2 sessions, and every month for the last 3 sessions; the maintenance phase included only monthly phone calls. Scheduling deviations were permissible to accommodate participant availability and preferences.

Participants received the PEARLS program for depression starting with the first in-person visit and were instructed to initiate the GLB video program after it was formally introduced during the fifth intervention session. The intervention outline is provided in S1 Appendix. A trained bachelor’s-level health coach delivered the intervention, and a supervising master’s-level registered dietitian oversaw fidelity assurance. They both met every 1 to 2 weeks with a psychiatrist and a primary care physician to review patient progress and discuss new and difficult cases. Additional detail on the intervention format, structure, and content and fidelity assurance procedures is provided in the published protocol. [20]

Measures

Process data were collected to evaluate participants’ progress over the year-long intervention during 2014–2017. After the GLB program was formally introduced in Week 6, participants were asked to manually enter their weight and minutes of leisure-time physical activity at least weekly using MyFitnessPal website or app. Also, participants were asked to wear a study-provided Fitbit pedometer that interfaced with a personal computer or the Fitbit app on a mobile device to automatically upload daily steps into the participant’s Fitbit account. The health coach was able to review the person’s self-tracked data, monitor their progress, and use it to facilitate individualized coaching during intervention sessions. In addition, the health coach administered the PHQ-9 at the beginning of each in-person or phone session. [29] Each participant could have up to 46 weeks with self-monitored weight, minutes of physical activity data, or steps as expected (from week 6 to week 52) and a maximum of 15 sessions (or 15 PHQ-9 scores). Indices of behavioral adherence to the intervention included the number of intervention sessions attended and the number of weeks with self-monitored weight, self-reported physical activity minutes, and FitBit steps separately.

Weight loss and depression outcomes used to validate the treatment engagement and progress categories in this study included weight loss and depression related primary and secondary outcomes objectively measured by blinded outcome assessors at baseline, 6, and 12 months in the RAINBOW trial. As primary outcomes, BMI was calculated as weight (kg) divided by height squared (m2); and depression severity was measured with the SCL-20 scores, ranging 0 (best) to 4 (worst). [22] Secondary outcome measures included ≥5% decrease in weight from baseline, [30] depression treatment response (i.e., ≥50% decrease in SCL-20 scores from baseline), [19, 26, 27] and complete depression remission (i.e., SCL-20 scores<0.5). [26, 27] Of 204 intervention participants, 196 and 183 had objectively-measured weight data at 6 and 12 months, respectively; and 175 and 169 had SCL-20 data at 6 and 12 months, respectively.

Statistical analysis

Cluster analysis on patterns of percent weight change and PHQ-9 score change

Patterns of change in 2 variables—percent weight change and PHQ-9 scores—were assessed separately using a method similar to the one by Babbin et al. [31] Both variables had direct relevance to treatment progress monitoring. The 1-year intervention period was examined in 4 quarters, and only participants who had any data in a quarter for at least 3 or all 4 quarters were included in the cluster analyses (n = 123/60% for self-monitored weight and n = 141/69% for PHQ-9, respectively). This approach was applied to enhance the reliability of change patterns during the yearlong intervention and reduce the influence of participants with missing data in 2 or more quarters. For either percent weight change or PHQ-9 scores, participants with no data in at least 3 quarters were classified as “cluster 0.” Cluster analyses for both percent weight change and PHQ9 scores followed the same 3 steps. First, the k-means method in the SAS FASTCLUS procedure without pre-specification of the number of clusters was used to group participants who had at least 1 measurement in each of the 4 quarters into clusters of individuals with similar patterns of change over time based on their 4 quarterly means. This step produced different numbers of clusters (range 2–6). Second, the optimal number of clusters was determined using a combination of criteria, including Pseudo F statistic (a relatively large value), R-squared value (a peak that flattens with additional clusters), Cubic Clustering Criterion (≥2), and cluster size (≥10 participants). [32] The optimal number of clusters was 3 for both percent weight change and PHQ-9 scores. Third, participants with percent weight change data in any 3 of the 4 quarters were assigned to their closest cluster defined by the smallest of the Euclidean distances between a participant’s 3 available quarterly means and each cluster’s means in the corresponding quarters. Using the same method, participants with PHQ-9 scores in any 3 of the 4 quarters were assigned to their closest cluster.

Internal consistency and sensitivity analysis

To compare individual trajectories within the resulting clusters, the polynomial regression was used to fit the trajectory of each participant’s available data on percent weight change and PHQ-9 scores over the course of the intervention. Then, analysis of variance (ANOVA) was used to compare intercept (for PHQ-9 only), linear, and quadratic coefficients of the individual trajectories among the 3 clusters for percent weight change and PHQ-9 separately. We also tested whether the polynomial model for each cluster fit the data well using the significance of polynomial terms, adjusted R2, and the Bayesian information criterion (BIC). For both percent weight change and PHQ-9 score change, the polynomial regression models with a quadratic term fit the data better than the ones without given the significance of the quadratic terms, higher adjusted R2, and lower BIC. Hence, the final models included both linear and quadratic terms.

Additionally, we tested whether the clusters derived separately for percent weight change and PHQ-9 were concordant with the clusters derived jointly for both variables because the integrated intervention addressed both obesity and depression. To do this, a sensitivity analysis was conducted using participants who had data on both variables in all 4 quarters (n = 88). The k-means method was applied in the separate and joint cluster analyses of the 88 participants.

Categorization and validation of treatment engagement and progress

Based on a cross tabulation of clusters 0 to 3 of percent weight change and PHQ-9 scores separately, all 204 intervention participants were grouped into 3 categories of treatment engagement and progress: poor engagement, poor progress, and progress. The poor engagement category included participants who had poor session attendance (i.e., cluster 0 for PHQ-9). The poor progress category included participants who had minimal improvement in self-monitored weight or PHQ-9 (i.e., cluster 1 for either) or had poor self-monitoring of weight despite attending sessions (i.e., cluster 0 for weight and cluster 1, 2, or 3 for PHQ-9). The progress category included participants who had improvements in both self-monitored weight and PHQ-9 (i.e., cluster 2 or 3 for both). For validation, intervention adherence indices—such as the number of sessions attended and the number of weeks with self-monitoring data as well as objectively measured BMI and SCL-20 at 6 and 12 months—were compared among the treatment engagement and progress categories. ANOVA was used for continuous variables and the chi-square test was used for categorical variables.

All analyses were conducted in 2018 using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina), except for sensitivity cluster analysis, which was conducted in kml and kml3d R packages. [33] Statistical significance was defined as P<0.05 (2-sided).

Results

Baseline characteristics

Baseline participant characteristics were previously published. [14] The 204 intervention participants were primarily middle aged (mean 50.9 [SD 12.2] years), female (71%), non-Hispanic White (72%), and at least college educated (70%) (Table 1). They had moderately severe obesity (BMI, mean 36.7 [SD 6.9]) and depression (PHQ-9, 14.1 [3.2]; SCL-20, 1.5 [0.5]).

Table 1

Baseline characteristics of RAINBOW intervention participants (n = 204).

Characteristic	All I-CARE participants (n = 204)	Participants included in cluster analysis of percent weight change (n = 123)	Participants included in cluster analysis of PHQ-9 scores (n = 141)
Age, year	50.9 (12.2)	52 (11.6)	51.2 (11.9)
Female, No. (%)	144 (71)	84 (68)	100 (71)
Race/Ethnicity, No. (%)
Non-Hispanic White	147 (72)	94 (76)	105 (74)
Minority	57 (28)	29 (24)	36 (26)
Education, No. (%)
High school to some college	61 (30)	31 (25)	39 (28)
College graduate	78 (38)	55 (45)	61 (43)
Post college	65 (32)	37 (30)	41 (29)
Income, No. (%), n = 176
<$100,000	66 (38)	37 (35)	46 (38)
$100,000- <$150,000	34 (19)	21 (20)	25 (21)
≥$150,000	76 (43)	47 (45)	50 (41)
Marital status, No. (%), n = 203
Married/living with a partner	123 (61)	80 (66)	84 (60)
Single/separated/divorced/widowed	80 (39)	42 (34)	56 (40)
Household size, No. (%), n = 203
< 2	40 (20)	20 (16)	27 (19)
= 2	74 (36)	48 (39)	53 (38)
3+	89 (44)	55 (45)	61 (43)
BMI, kg/m2	36.7 (6.9)	36.7 (7.0)	36.9 (6.9)
PHQ-9	14.1 (3.2)	13.7 (3.2)	13.9 (3.2)
SCL-20	1.5 (0.5)	1.4 (0.5)	1.4 (0.5)

Abbreviations: BMI, body mass index; PHQ-9, Patient Health Questionnaire-9; SCL20, Symptom Checklist-20.

Values are mean (SD) unless otherwise noted.

Clusters of percent weight change and PHQ-9 scores separately

Participants with self-monitored weight data in at least 3 quarters of the 12-month intervention period (n = 123) had similar baseline characteristics as the entire intervention group (Table 1). Among the 123 participants, the 3 clusters of percent weight change trajectories were as follows: (1) minimal weight loss (n = 50; β1 = -0.06, β2 = 0.001), (2) moderate weight loss (n = 61; β1 = -0.28, β2 = 0.002), and (3) most weight loss (n = 12; β1 = -0.53, β2 = 0.005) (Fig 1 and S2 Appendix for plots of individual trajectories within each cluster). Both the linear (β1) and quadratic terms (β2) were statistically different from zero for these clusters (all P<0.001). Pairwise comparisons of β1 and β2 coefficients of the individual trajectories across the 3 clusters showed that β1s within cluster 3 were significantly lower than those within cluster 2, which were significantly lower than those within cluster 1; and β2s within cluster 3 was not significantly different from those within cluster 2 but both were significantly higher than those within cluster 1 (S3 Appendix). Fig 1 shows that separation of the clusters began within 6 to 8 weeks and grew over time; mean weight loss reached 5% within 12 weeks in cluster 3, but not until beyond 20 weeks in cluster 2, and never in cluster 1.

Fig 1

Percent weight change trajectories among intervention participants with self-monitored weight data in at least 3 quarters of the 12-month intervention perioda,b.

β1, linear coefficient; β2, quadratic coefficient. ***P < .001. a123 participants, or 60% of the intervention group (n = 204), had self-monitored weight data in at least 3 quarters of the 12-month intervention period. bLight gray lines show individual participant trajectories within each cluster.

Participants with PHQ-9 scores in at least 3 quarters (n = 141) had similar baseline characteristics as the entire intervention group (Table 1). Among these participants, the 3 clusters of PHQ-9 trajectories were as follows: (1) moderate depression without treatment progress (n = 40; β0 = 11.05, β1 = -0.11, β2 = 0.002), (2) moderate depression with treatment progress (n = 20; β0 = 12.90, β1 = -0.42, β2 = 0.006), and 3) milder depression with treatment progress (n = 81; β0 = 7.41, β1 = -0.23, β2 = 0.003) (Fig 2 and S4 Appendix for plots of individual trajectories within each cluster). The intercept (β0), linear term (β1), and quadratic terms (β2) were statistically significant from zero for these clusters (all P<0.001, except for cluster 1 β1, P<0.01 and β2, P<0.05). Additionally, pairwise comparisons of β0, β1, and β2 coefficients of the individual trajectories across the 3 clusters showed that β0s within cluster 2 were significantly higher than those within cluster 1, which were significantly higher than those within cluster 3; β1s within cluster 1 were significantly higher than those within cluster 3, which were significantly higher than those within cluster 2; and β2s were not significantly different among the 3 clusters (S5 Appendix).

Fig 2

PHQ-9 trajectories among intervention participants with PHQ-9 data in at least 3 quarters of the 12-month intervention perioda,b.

β0 = intercept; β1 = linear coefficient; β2 = quadratic coefficient. *P < .05; **P < .01; ***P < .001. a141 participants, or 69% of the intervention group (n = 204), had PHQ-9 data in at least 3 quarters of the 12-month intervention period. bLight gray lines show individual participant trajectories within each cluster.

Interaction of percent weight change and PHQ-9 score clusters

The cross-classification of participants according to the clusters of percent weight change and PHQ-9 scores and participants with insufficient data (cluster 0) is shown in Table 2. Three categories were identified. The poor treatment engagement (n = 63) had both poor session attendance and all but 2 participants had poor self-monitoring of weight and consequently inadequate data to be included in cluster analysis. The poor treatment progress (n = 80) had minimal improvement in self-monitored weight or PHQ-9 or had poor self-monitoring of weight despite attending sessions. The progress category (n = 61) showed overall positive treatment progress for both self-monitored weight and PHQ-9. There were minimal differences in baseline characteristics among these categories (S6 Appendix).

Table 2

Categories of treatment engagement and progress based on percent weight change and PHQ-9 clusters.

	Frequency Percent Row Percent Column Percent	PHQ-9 trajectory cluster				Total
		0-No PHQ-9 cluster (i.e., Poor session attendance) (n = 63)	1-Moderate depression without treatment progress (n = 40)	2-Moderate depression with treatment progress (n = 20)	3-Milder depression with treatment progress (n = 81)
Percent weight change trajectory cluster	0-No weight cluster (i.e., Poor self-monitoring) (n = 81)	61	11	1	8	81
		29.9	5.4	0.5	3.9	39.7
		75.3	13.6	1.2	9.9
		96.8	27.5	5.0	9.9
	1-Minimal weight loss (n = 50)	1	18	9	22	50
		0.5	8.8	4.4	10.8	24.5
		2.0	36.0	18.0	44.0
		1.6	45.0	45.0	27.2
	2-Moderate weight loss (n = 61)	1	10	8	42	61
		0.5	4.9	3.9	20.6	29.9
		1.6	16.4	13.1	68.9
		1.6	25.0	40.0	51.9
	3-Most weight loss (n = 12)	0	1	2	9	12
		0.0	0.5	1.0	4.4	5.9
		0.0	8.3	16.7	75.0
		0.0	2.5	10.0	11.1
	Total	63	40	20	81	204
		30.9	19.6	9.8	39.7	100.0

Different shades indicate the 3 categories of treatment engagement and progress: (1) light gray: the poor engagement category (n = 63), (2) gray: the poor progress category (n = 80), and (3) dark gray: the progress category (n = 61).

Validation of treatment engagement and progress categories

These categories differed significantly in the indices of behavioral adherence to the intervention and objectively measured weight loss and depression outcomes at 6 and 12 months (Table 3). Participants with poor engagement attended fewer than 5 out of 15 sessions (SD 2.6) and provided self-monitored data in a minimal number of weeks either actively (manual entries of weight or minutes of leisure-time physical activity in <3 weeks) or passively (FitBit steps uploaded automatically in <9 weeks). Among them, 39 (62%) discontinued the intervention by session 5 (week 6) and another 11 (17%) dropped out at session 6 (week 8). Relatedly, these participants also had minimal improvements in both weight loss and depression outcomes at 6 and 12 months. Participants with poor treatment progress attended almost all 15 sessions and showed intermediate levels of self-monitoring. However, the objectively measured weight loss and depression outcomes among these participants were comparable to participants with poor engagement and worse than participants with treatment progress. The last category had perfect attendance and good active (weight or physical activity minutes monitored for 21 to 22 weeks) and passive self-monitoring (FitBit steps uploaded in 36 weeks). Mean (SD) reductions were -2.0 (1.3) in BMI and -0.6 (0.6) in SCL-20 at 6 months, which sustained at 12 months. These categories also differed significantly in the number of days since Session 1 for each subsequent in-person session, possibly reflecting different degrees of scheduling difficulties, disinterest, or lack of commitment (S7 Appendix).

Table 3

Comparisons of adherence behaviors and outcomes by category of treatment engagement and progress.

	All intervention participants (n = 204)	Poor engagement (n = 63; 31%)	Poor progress (n = 80; 39%)	Progress (n = 61; 30%)	P value
Adherence behaviors
No. of sessions attended	11.4 (5.1)	4.5 (2.6)a	14.2 (2.3)b	15.0 (0.1)c	<0.001
No. of weeks with self-monitored weight	11.5 (11.5)	2.6 (4.3)a	11.2 (8.5)b	21.0 (12.6)c	<0.001
No. of weeks with self-reported minutes of leisure-time physical activity	11.5 (12.0)	1.5 (2.3)a	11.6 (8.5)b	21.8 (13.2)c	<0.001
No. of weeks with FitBit steps	24.5 (17.4)	8.7 (9.2)a	28.0 (15.8)b	36.1 (13.8)c	<0.001
Weight loss and depression outcomes
BMI change from baseline
6 months, n = 196	-0.7 (1.7)	0.0 (1.8)a	-0.2 (1.4)a	-2.0 (1.3)b	<0.001
12 months, n = 183	-0.7 (2.2)	0.1 (2.4)a	-0.0 (1.6)a	-2.2 (1.9)b	<0.001
≥5% weight loss from baseline, No. (%)
6 months, n = 196	48 (24.5)	8 (14.0)a	8 (10.1)a	32 (53.3)b	<0.001
12 months, n = 183	51 (27.9)	8 (16.7)a	10 (13.2)a	33 (55.9)b	<0.001
SCL-20 change from baseline
6 months, n = 175	-0.3 (0.7)	0.0 (0.7)a	-0.3 (0.7)b	-0.6 (0.6)b	<0.001
12 months, n = 169	-0.3 (0.7)	0.3 (0.8)a	-0.3 (0.7)b	-0.6 (0.6)b	<0.001
Depression response (≥50% decrease in SCL-20 scores from baseline), No. (%)
6 months, n = 175	55 (31.4)	5 (13.5)a	19 (24.4)a	31 (51.7)b	<0.001
12 months, n = 169	49 (29.0)	4 (12.1)a	18 (23.4)a	27 (45.8)b	0.001
Depression remission (SCL-20 scores < 0.5), No. (%)
6 months, n = 175	31 (17.7)	1 (2.7)a	11 (14.1)a	19 (31.7)b	<0.001
12 months, n = 169	30 (17.8)	3 (9.1)a	10 (13.0)a	17 (28.8)b	0.02

Values are mean (SD) unless otherwise noted. P values are obtained from ANOVA comparing 3 categories for continuous variables or from the chi-square test comparing 3 categories for categorical variables.

a, b, c Different superscripts denote statistically significant differences between categories.

Sensitivity analysis

Among participants with both self-monitored weight and PHQ-9 score in all 4 quarters (n = 88), joint cluster analysis resulted in 3 clusters: (A) no treatment progress in either percent weight change or PHQ-9 (n = 19); (B) treatment progress in PHQ-9 only (n = 33); and (C) treatment progress in both (n = 36) (S8 Appendix). Separate cluster analysis resulted in 2 clusters for percentage weight change: without (n = 48) and with (n = 40) weight loss; and 2 clusters for PHQ-9: without (n = 26) and with (n = 62) treatment progress (S8 Appendix). The number of participants in clusters resulting from joint and separate cluster analyses showed high concordance. For example, 52 participants with minimal weight loss (i.e., clusters A and B) in the joint cluster analysis compared to 48 (cluster 1) in the cluster analysis on weight only. Also, 69 participants with PHQ-9 response (i.e., clusters B and C) in the joint cluster analysis compared to 62 (cluster 2) in the cluster analysis on PHQ-9 only. In addition, the number of participants in clusters resulting from the joint cluster analysis also showed concordance with the number of participants in Table 2. For example, the number of 33 participants with treatment progress in PHQ-9 only resulting from the joint cluster analysis (i.e., group B in S8 Appendix) was concordant with the number of 31 participants who had depression treatment progress (i.e., PHQ-9 cluster 2 and 3) but minimal weight loss (i.e., weight change cluster 1) in Table 2.

Discussion

This study showed that even in the context of an efficacious intervention for obesity and depression, participants varied in treatment engagement and progress. Poor treatment engagement manifested as low adherence to session attendance and/or self-monitoring affected >30% of participants. Among those engaged patterns of treatment progress differentiated for weight loss: (1) minimal weight loss, (2) moderate weight loss, and (3) substantial weight loss; and for depression: (1) initial moderate depressive symptoms without progress, (2) initial moderate depressive symptoms with progress, and (3) initial milder depressive symptoms with progress. These patterns were not only significantly associated with intervention adherence behaviors—such as session attendance and self-monitoring—but also with objectively assessed efficacy outcomes.

Evidence on this topic is scarce due to limited data assessment points in conventional clinical trials. Only a few prior studies have investigated the dynamic trajectories of weight loss in behavioral interventions and, to a lesser extent, the trajectories of depression symptoms. The 3 weight loss patterns identified in the current study were similar to those found in previous weight loss studies. [34, 35] The current study also identified 3 depression symptom patterns, similar to—although not identical to—the 2 prior studies investigating dynamic trajectories of depression symptoms. These prior studies identified 2 patterns, gradual/slower responders and rapid responders (in an antidepressant only or antidepressant plus psychotherapy study). [36-38] one study [36] found that higher baseline depression severity was associated with the gradual/slower responder trajectory; whereas the current study found that among participants with higher baseline depression severity there were 2 distinct subgroups, those with treatment progress and those without, although this sample had comorbid obesity.

To our knowledge, the present study is the first to examine the temporal patterns of change in both weight and depression severity in response to an integrated collaborative care intervention for comorbid obesity and depression. Separation of the weight loss and PHQ-9 clusters was evident by 6 to 8 weeks of treatment and persisted throughout the 12-month intervention. Additionally, this study identified subgroups of treatment engagement and progress levels that were significantly associated with objectively-measured weight loss and depression outcomes at the end of the intervention. These findings suggest that evaluation of dynamic treatment engagement and progress early in the course of intervention might provide important information regarding how an individual will respond by the intervention endpoint. Consistent with this study, earlier studies on weight loss demonstrated that initial weight loss at 1 to 2 months was significantly associated with 1-year and even longer-term weight loss up to 8 years. [39-41] This study found that participants with poor engagement or poor progress showed minimal differences in baseline characteristics from those with progress; however, they differed significantly in intervention adherence behaviors. This has practical implications because poor adherence behaviors—such as low rates of session attendance and self-monitoring—are easy to detect and respond to early in the course of an intervention. Treatment strategies could be adjusted for these individuals to optimize treatment outcomes. For example, participants who show early signs of nonengagement such as poor session attendance and/or self-monitoring or poor progress such as not reaching interim intervention goals may benefit from an augmented intervention with motivational interviewing strategies, thereby minimizing the risk of treatment failure. Similarly, a recent study of a community-based intervention for chronic disease management in participants with two or more diseases (i.e., diabetes, obesity, hypertension, and tobacco dependence) reported that treatment response was predicted by participants’ reactions to the challenges and failures they faced during the intervention instead of their baseline characteristics; the authors concluded that behavioral interventions could be modified to help non-responders face the challenges and failures. [42]

This study has limitations. First, because of the post hoc nature of the analyses the findings need to be replicated in future studies. Second, the weight change and PHQ-9 score clusters may be specific to the study data. Therefore, future studies of independent samples are needed to verify the external validity of the results. In addition, given that participants might be less severely depressed in RCTs, [43] this may reduce generalizability of the findings to the clinical population with more severe depression. Third, this study only evaluated trajectories of weight change and depression symptoms over 12 months; consequently, it does not provide insight into trajectories of long-term outcomes.

Conclusions

This study carefully examined heterogeneity in treatment engagement and progress over the course of an efficacious, yearlong integrated collaborative care intervention for obesity and depression and identified subgroups of patients who were more or less likely to engage in or benefit from this type of treatment. Signs of poor engagement or progress manifested early in the intervention, persisted, and correlated significantly with treatment efficacy outcomes. Identifying patients with likely treatment failure using engagement and progress data early in the intervention could enable individualized optimization to enhance efficacy.

Intervention outlinea,b.

a In-between session support as needed via EHR secure email, between weeks 1–52. b Co-located psychiatric and medical supervision during weekly intervention management team meeting, between weeks 1–52. c The 9 one-on-one I-CARE sessions will occur primarily in the clinic, but video conferences (as the second option) or phone sessions (as the last option and for visits 1–5, phone session is only an option upon PI and intervention manager approval) throughout the intensive phase will be an option for participants with considerable constraints. d I-CARE Mood is the PEARLS program; I-CARE Lifestyle is the GLB program. e Participants receive Fitbit, MyFitnessPal, and My Health Online instructions via mail or e-mail prior to first session.

(DOCX)

Click here for additional data file.

Individual participant trajectories within each cluster of percent weight change.

(DOCX)

Click here for additional data file.

Mean (±SD) beta coefficients of individual trajectories within each cluster of percent weight change.

β1 = Linear coefficient; β2 = Quadratic coefficient. abcDifferent letters indicate significant difference.

(DOCX)

Click here for additional data file.

Individual participant trajectories within each cluster of PHQ-9 change.

(DOCX)

Click here for additional data file.

Mean (±SD) beta coefficients of individual trajectories within each cluster of PHQ-9 change.

β0 = Intercept; β1 = Linear coefficient; β2 = Quadratic coefficient. abcDifferent letters indicate significant difference.

(DOCX)

Click here for additional data file.

Comparisons of baseline characteristics by category of treatment engagement and progress.

Abbreviations: BMI, body mass index; PHQ-9, Patient Health Questionnaire-9; SCL20, Symptom Checklist-20. Values are mean (SD) unless otherwise noted.

(DOCX)

Click here for additional data file.

Mean (±SD) number of days from session 1 by category of treatment engagement and progress.

Abbreviations: NA, not applicable (SD is NA because only 1 participant attended session 9 in the poor engagement category). a, b, c Different superscripts denote statistically significant differences between categories.

(DOCX)

Click here for additional data file.

Joint and separate cluster analysis of weight and PHQ-9 trajectories among intervention participants who had at least one self-monitored weight measure and one PHQ-9 score in all 4 quarters (n = 88).

(DOCX)

Click here for additional data file.

18 Sep 2019

PONE-D-19-15164

Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: within-treatment analysis in the RAINBOW trial

PLOS ONE

Dear Dr. Ma,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The study is timely. However, one reviewer has recommended a rejection. In my opinion, the data are interesting, and manuscript can be published if all questions/comments are addressed and presentation is revised. Presently it is impossible to understand for an “outsider”.

Overall, it appears that on the one hand it is understandable for any physician that a patient who is not taking medications, i.e. not coming to sessions or does not do required assignment (checking weight), i.e. “not engaged” is usually not expected to improve in any disease for any outcome. So, a priori cluster 0, n=61 is not expected to improve and it appears that this is your conclusion: “Participants with failed treatment outcomes demonstrated poor engagement or progress early in the intervention. This insight could inform individualized optimization to enhance efficacy.” Perhaps it can be acceptable if other components of the paper are improved as requested by the reviewers. If this is your major conclusion you have to incorporate this “intuitive” thinking somewhere. It is obvious but still valuable since published papers do not commonly talk about this.

On the other hand, Table 3 shows that if you compare “Weight % change” in “poor progress” and “Progress” there is significant weight change differences. The Table shows that both groups attended practically the same number of sessions, n=14 and n=15, respectively (yet statistically significant?). Also, although self-monitoring (for weight?) appears different 11.2 vs. 21 weeks, it is not statistically different. In addition, for “weight % change”, n=196 or n=183. What does this mean? Analysis included 123 participants for weight according to Methods. All this is confusing. Each of these points has to be addressed.

There are multiple “small” inconsistencies that will need to be addressed after major problems are resolved. For example:

1. the definition of “engagement” only appears in Results, when it is expected in Methods.

2. Table 1 shows 204 participants. However, this sub-analysis includes only 123 or 141 participants, so data for these are expected.

3. Is this “post hoc” analysis of previous data? Was “Engagement” assessed after study was completed?

4. Did you try to answer a question: Who failed the treatment?

5. Use n=123 and n=141 in all analyses and in Tables.

6. Since weight loss and depression are assessed separately, present the outcomes separately to simplify understanding. Use n=141 for all data on depression and n=123 for all data on weight loss. Then address interaction between these outcomes.

==============================

We would appreciate receiving your revised manuscript by Nov 02 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Elena Barengolts, MD

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 http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2.  We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. Please ensure you have included the registration number for the clinical trial referenced in the manuscript

4. Thank you for stating the following in the Competing Interests section:

'The authors have declared that no competing interests exist.'

We note that one or more of the authors are employed by a commercial company: One Medical.

Please provide an amended Funding Statement declaring this commercial affiliation, as well as a statement regarding the Role of Funders in your study. If the funding organization did not play a role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript and only provided financial support in the form of authors' salaries and/or research materials, please review your statements relating to the author contributions, and ensure you have specifically and accurately indicated the role(s) that these authors had in your study. You can update author roles in the Author Contributions section of the online submission form.

Please also include the following statement within your amended Funding Statement.

“The funder provided support in the form of salaries for authors [insert relevant initials], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.”

If your commercial affiliation did play a role in your study, please state and explain this role within your updated Funding Statement.

2. Please also provide an updated Competing Interests Statement declaring this commercial affiliation along with any other relevant declarations relating to employment, consultancy, patents, products in development, or marketed products, etc.

Within your Competing Interests Statement, please confirm that this commercial affiliation does not alter your adherence to all PLOS ONE policies on sharing data and materials by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests) . If this adherence statement is not accurate and  there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include both an updated Funding Statement and Competing Interests Statement in your cover letter. We will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

Additional Editor Comments (if provided):

[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: Partly

Reviewer #2: Yes

**********

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

Reviewer #1: No

Reviewer #2: Yes

**********

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: No

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: ABSTRACT

I do not understand the regression weights in the abstract. I also do not understand how these are related to outcome. Rather give effect sizes that are interpretable and clearly describe the variables please. Furthermore, if you want to show differences between groups in efficacy - what I think that you want to shoe, you should present interaction effects toon differences between sub-groups (and not main effects in sub-groups).

METHOD

I also do not understand the method sections and hence also not the results. Perhaps its me I don’t know. But based on what I read and what I understand from it, I do not recommend publishing this paper

Reviewer #2: This study investigates the combined variation in body weight changes and depression scores among patients undergoing an integrated care programme. This is a very worthwhile objective, in order to describe and understand the spectrum of reality behind mean and standard deviation summary values. The authors successfully describe the variation by means of cluster analysis. However, although attempts were made to verify the results using sensitivity analyses, it remains unclear to what extent the clusters were distinct and whether the chosen ‚optimal‘ number of clusters is a characteristic of the data or a construct of the analysis method.

Further, the conclusion line in the abstract ‘Participants with failed treatment outcomes demonstrated poor engagement or progress early in the intervention’ does not, to me, seem to be clearly demonstrated. Most of those with poor engagement had no data on either weight loss or depression. And particularly for depression scores, many of those in cluster 1 never-the-less had improvements in the early period. The authors need to argue this point more exactly.

Specific points:

1. The sensitivity analysis based jointly on weight and PHQ scores is particularly interesting since it yields a cluster whose members responded in PHQ but not in weight. This result is not evident (or at least, not emphasized) in the main cluster analysis. The concordance of this clustering result with the main analysis and the display in Table 3 might be informative.

2. I suggest using, as a sensitivity analysis, clustering based not on the quarterly means but on an alternative averaging over time (e.g. 4-monthly). Further, an alternative clustering method without pre-specification of the number of clusters might help to test the robustness of the results.

3. Figs. 1 and 2 display the individual trajectories but they are so small that it is difficult to get the picture. Could larger plots be shown?

4. The regression coefficient values given in appendix tables S2 and S3 differ from those in the text (pp. 9 and 10).

5. Many the ‘observed’ correlations and ‘significant’ differences are tautologous, stemming from the same or related data. For instance, the clusters were defined based on the weight and depression score changes, so it is almost inevitable that the clusters differ significantly with regard to the regression coefficients for these changes. It is similarly to be expected that the categories of treatment engagement and progress differ significantly with respect to adherence behaviours, since these behaviours were used to define the ‚no weight‘ and ‚no PHQ-9‘ clusters and thus the first category.

6. Did the authors test whether the polynomial models of trajectories fitted the data well, either to the individual trajectories or to the cluster means?

7. It would be interesting to investigate the possible chronological and causal relationships between weight change and depression score change: were they simultaneous or did one tend to follow the other? Or maybe no regularity can be discerned?

**********

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: Yes: Marc Molendijk

Reviewer #2: Yes: Jeremy Franklin

[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 to be viewed.]

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 us at figures@plos.org. Please note that Supporting Information files do not need this step.

19 Nov 2019

We thank the editor and reviewers for their constructive comments on our originally submitted manuscript, titled “Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: within-treatment analysis in the RAINBOW trial.” Below please find our point-to-point responses.

Editor’s comment:

1. Overall, it appears that on the one hand it is understandable for any physician that a patient who is not taking medications, i.e. not coming to sessions or does not do required assignment (checking weight), i.e. “not engaged” is usually not expected to improve in any disease for any outcome. So, a priori cluster 0, n=61 is not expected to improve and it appears that this is your conclusion: “Participants with failed treatment outcomes demonstrated poor engagement or progress early in the intervention. This insight could inform individualized optimization to enhance efficacy.” Perhaps it can be acceptable if other components of the paper are improved as requested by the reviewers. If this is your major conclusion you have to incorporate this “intuitive” thinking somewhere. It is obvious but still valuable since published papers do not commonly talk about this.

Authors’ response: We agree with the intuitiveness of our finding that patients with poor treatment engagement had poor outcomes. In addition to stating this obvious finding (which we also agree is nonetheless valuable), we underscored its potential clinical implications in the third paragraph of the discussion section: The percent weight change and PHQ-9 clusters diverged at the beginning of the intervention, and the clusters correlated significantly with adherence behaviors that are easier to monitor early in an intervention in the real world. These findings suggest that the adherence behaviors and early signs of poor progress (within weeks) based on self-monitored weight and PHQ-9 scores could be used to identify participants who would not achieve clinically significant improvements in BMI or SCL-20 at the end of the 12-month treatment. In doing so, alternative strategies may be provided to augment the treatment in order to help people move from the poor engagement or poor progress group to the progress group. As detailed below, we responded to all the editor’s and reviewers’ other comments and suggestions, and we believe that the revised manuscript has improved substantially as a result.

2. On the other hand, Table 3 shows that if you compare “Weight % change” in “poor progress” and “Progress” there is significant weight change differences. The Table shows that both groups attended practically the same number of sessions, n=14 and n=15, respectively (yet statistically significant?). Also, although self-monitoring (for weight?) appears different 11.2 vs. 21 weeks, it is not statistically different. In addition, for “weight % change”, n=196 or n=183. What does this mean? Analysis included 123 participants for weight according to Methods. All this is confusing. Each of these points has to be addressed.

Authors’ response: Owing to formatting issues, some numbers in Table 3 were truncated; this problem is now fixed. As indicated in Table 3, the P values were obtained from ANOVA or chi-square tests comparing 3 groups, “poor engagement,” “poor progress,” and “progress” while significant pairwise differences were denoted by superscripts. The different numbers of participants with weight data (n=123, 196, 183) were correct and were attributed to differences in the sources of weight measurements. Weight data used in the cluster analysis were self-monitored weights provided by participants throughout the intervention, and 123 was the number of participants included in the cluster analysis for weight change over time because they met the criterion of having self-monitored weight data in at least 3 quarters during the one-year intervention. This criterion was applied to enhance the reliability of weight change patterns and reduce the influence of participants with missing data in 2 or more quarters. Weight data used to validate the results from cluster analyses as reported in Table 3 were weights objectively measured at baseline, 6, and 12 months by research staff blinded to treatment assignment. Of the 204 intervention participants, the number of participants with objectively-measured weights was 196 at 6 months and 183 at 12 months. We have clarified the different sources of weight data in the measures section (see pages 5-7 lines 128-161).

3. The definition of “engagement” only appears in Results, when it is expected in Methods.

Authors’ response: We have moved the definitions of the 3 categories of treatment engagement and progress to the methods section: “The poor engagement category included participants who had poor session attendance (i.e., cluster 0 for PHQ-9). The poor progress category included participants who had minimal improvement in self-monitored weight or PHQ-9 (i.e., cluster 1 for either) or had poor self-monitoring of weight despite attending sessions (i.e., cluster 0 for weight and cluster 1, 2, or 3 for PHQ-9). The progress category included participants who had improvements in both self-monitored weight and PHQ-9 (i.e., cluster 2 or 3 for both)” (see page 9 lines 208-214).

4. Table 1 shows 204 participants. However, this sub-analysis includes only 123 or 141 participants, so data for these are expected.

Authors’ response: The total number of participants in the intervention was 204. In Table 1 and the results section, we have added baseline characteristics of the subgroups of participants with data in at least 3 quarters during the one-year intervention to be included in cluster analyses of percent weight change (n=123) and PHQ-9 scores (n=141).

5. Is this “post hoc” analysis of previous data? Was “Engagement” assessed after study was completed?

Authors’ response: Yes, this is a “post hoc” analysis of the data collected in the RAINBOW trial. We have clarified this in the methods of the abstract (page 2 line 38) and in the last paragraph of the introduction (page 4 line 90). In addition, we acknowledge the post hoc nature of the study as a limitation in the discussion, which was in the original submission as well. It is also correct that engagement was assessed after the study was completed. In the statistical analysis section under the “categorization and validation of treatment engagement and progress” subheading, we defined the 3 categories of treatment engagement and progress (i.e., poor engagement, poor progress, and progress) (page 9 lines 208-214).

6. Did you try to answer a question: Who failed the treatment?

Authors’ response: Our analyses identifying participants with poor engagement or poor progress address the question “Who failed the treatment?” We have added in the results (page 13 lines 297-298) with a new Appendix table (S6) to clarify that there were minimal differences in baseline characteristics among the 3 categories (poor engagement, poor progress, and progress). However, as reported in our original submission, these categories differed significantly in multiple indices of behavioral adherence to the intervention (Table 3). In the discussion, we have added the following with citation of a recent study (Ed et al., 2018): “This study found that participants with poor engagement or poor progress showed minimal differences in baseline characteristics from those with progress; however, they differed significantly in intervention adherence behaviors. …… Similarly, a recent study of a community-based intervention for chronic disease management in participants with two or more diseases (i.e., diabetes, obesity, hypertension, and tobacco dependence) reported that treatment response was predicted by participants’ reactions to the challenges and failures they faced during the intervention instead of their baseline characteristics; the authors concluded that behavioral interventions could be modified to help non-responders face the challenges and failures.”

7. Use n=123 and n=141 in all analyses and in Tables.

Authors’ response: By responding to comment #4 above, we have added n=123 and n=141 in Table 1 to show demographics for these two subgroups. The subgroup of 123 participants included those who met the criteria to be included in cluster analysis for percent weight change. The subgroup of 141 participants included those who met the criteria to be included in cluster analysis for PHQ-9 scores. In Table 2, we have added the numbers of participants in cluster 0, 1, 2, and 3 for percent weight change and PHQ-9. The sum of clusters 1, 2, and 3 for percent weight change was 123, and the sum of clusters 1, 2, and 3 for PHQ-9 was 141. Cluster 0 included participants who did not have adequate data to be included in the cluster analyses. The four clusters combined capture all 204 intervention participants. Table 3 shows adherence behaviors and objectively measured weight and depression outcomes by category of treatment engagement and progress, which was created based on combinations of the four percent weight change clusters and the four PHQ-9 clusters as displayed in Table 2. Therefore, it is impossible to identify n=123 and n=141 in Table 3.

8. Since weight loss and depression are assessed separately, present the outcomes separately to simplify understanding. Use n=141 for all data on depression and n=123 for all data on weight loss. Then address interaction between these outcomes.

Authors’ response: We indeed presented the results as recommended. Figures 1 and 2 present cluster analysis results for the trajectories of percent weight change and PHQ-9 scores separately. Then, Tables 2 and 3 present results combining the clusters for each outcome to examine their interactions. To further clarify this order of presentation, we have added subheadings in the results section.

Journal requirements:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming.

Authors’ response: Done.

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

Authors’ response: We have provided the information in the cover letter: “The IRB-approved study protocol and written informed consent forms provided by participants in the study do not permit sharing of de-identified data without a formal data sharing and use agreement. For data requests, the University of Illinois at Chicago (UIC) IRB can be reached at 1-(312) 996-1711 or uicirb@uic.edu.”

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

Authors’ response: Please see our response above.

3. Please ensure you have included the registration number for the clinical trial referenced in the manuscript

Authors’ response: We did include the registration number (ClinicalTrials.gov#NCT02246413) below abstract in the original submission and have kept it there in this resubmission.

4. Thank you for stating the following in the Competing Interests section:

'The authors have declared that no competing interests exist.'

We note that one or more of the authors are employed by a commercial company: One Medical.

1. Please provide an amended Funding Statement declaring this commercial affiliation, as well as a statement regarding the Role of Funders in your study. If the funding organization did not play a role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript and only provided financial support in the form of authors' salaries and/or research materials, please review your statements relating to the author contributions, and ensure you have specifically and accurately indicated the role(s) that these authors had in your study. You can update author roles in the Author Contributions section of the online submission form.

Please also include the following statement within your amended Funding Statement.

“The funder provided support in the form of salaries for authors [insert relevant initials], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.”

If your commercial affiliation did play a role in your study, please state and explain this role within your updated Funding Statement.

Authors’ response: The NIH was the sole funder of the reported research, and there was no commercial affiliation. Our Finding Statement in the cover letter is as follows: “Research reported in this publication was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health (URL: https://www.nhlbi.nih.gov/) under Award Number R01HL119453 (Recipient: JM). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. There were no other funding sources.”

2. Please also provide an updated Competing Interests Statement declaring this commercial affiliation along with any other relevant declarations relating to employment, consultancy, patents, products in development, or marketed products, etc.

Within your Competing Interests Statement, please confirm that this commercial affiliation does not alter your adherence to all PLOS ONE policies on sharing data and materials by including the following statement: "This does not alter our adherence to PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests) . If this adherence statement is not accurate and there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include both an updated Funding Statement and Competing Interests Statement in your cover letter. We will change the online submission form on your behalf.

Authors’ response: It is accurate that the authors had no competing interests. Our Competing Interests Statement in the cover letter is as follows: “The authors declare that no competing interests existed for the research as reported. Dr. Lenard Lesser’s affiliation with 1Life Healthcare/One Medical constituted no competing interests. Dr. Lesser had originally supported on the NIH grant for his role as a study physician on this study while he was an employee at the Palo Alto Medical Foundation Research Institute (PAMFRI) where the study was conducted. Starting from 07/2016 till 10/2017, Dr. Lesser transitioned from PAMFRI to 1Life Healthcare, Inc/One Medical. At that time, his continuous involvement in the study was compensated through a research contract as an independent consultant with PAMFRI specifically for the NIH grant supporting the study. 1Life Healthcare/One Medical provided no support of any form for the study; and had no role in the study design, data collection/analysis, decision to publish, or preparation of the manuscript. This does not alter our adherence to PLOS ONE policies on sharing data and materials.”

Reviewer 1’s comments:

ABSTRACT

I do not understand the regression weights in the abstract. I also do not understand how these are related to outcome. Rather give effect sizes that are interpretable and clearly describe the variables please. Furthermore, if you want to show differences between groups in efficacy - what I think that you want to show, you should present interaction effects on differences between sub-groups (and not main effects in sub-groups).

Authors’ response: The objective of this paper was not to show differences in efficacy outcomes between the intervention and control groups, which has been published in JAMA (Ma et al., JAMA 2019, 321(9):869-879). Instead, this paper aimed to examine different patterns of engagement and progress and the relation of these patterns to outcomes among participants within the intervention group only. In other words, the objective of this paper was to understand participants’ response or lack thereof to the intervention so that future research can be designed to address non-responders and enhance intervention efficacy. The introduction of the abstract states, “The RAINBOW randomized clinical trial validated the efficacy of an integrated collaborative care intervention for obesity and depression in primary care, although the effect was modest. To inform intervention optimization, this study investigated within-treatment variability in participant engagement and progress.” Beta coefficients are the appropriate effect estimates for the cluster analyses as reported showing patterns of weight change (%) and PHQ-9 scores over the 12-month intervention. As noted in the abstract, β0 is the intercept, β1 is the linear term, and β2 is the quadratic term.

METHOD

I also do not understand the method sections and hence also not the results. Perhaps its me I don’t know. But based on what I read and what I understand from it, I do not recommend publishing this paper.

Authors’ response: The reviewer indicated lack of understanding of the methods and results sections without specific comments that we could effectively address. We appreciate the opportunity to revise and resubmit our manuscript per the editor’s decision. We agree with the editor that the study has merit and is timely. We have indeed revised our methods and results sections by responding to the editor’s and reviewer 2’s comments, and we believe the revised manuscript is substantially improved as a result. We welcome any additional specific feedback from the editor and reviewers to further improve the manuscript for publication.

Reviewer 2’s comments:

1. This study investigates the combined variation in body weight changes and depression scores among patients undergoing an integrated care program. This is a very worthwhile objective, in order to describe and understand the spectrum of reality behind mean and standard deviation summary values. The authors successfully describe the variation by means of cluster analysis. However, although attempts were made to verify the results using sensitivity analyses, it remains unclear to what extent the clusters were distinct and whether the chosen ‘optimal’ number of clusters is a characteristic of the data or a construct of the analysis method.

Authors’ response: In the original manuscript, we described the separation of these clusters and included detailed results in Appendices. In the revised manuscript, we have further clarified the results on clusters of percent weight change: “Pairwise comparisons of β1 and β2 coefficients of the individual trajectories across the 3 clusters showed that these clusters were distinct from each other: β1 of cluster 3 was significantly lower than cluster 2, which was significantly lower than cluster 1; and β2 of cluster 3 was not significantly different from cluster 2 but both were significantly higher than cluster 1 (S3 Appendix)” (see page 11 lines 244-247). We have also further clarified the results on clusters of PHQ-9 scores: “pairwise comparisons of β0, β1, and β2 coefficients of the individual trajectories across the 3 clusters showed that these clusters were distinct from each other: β0 of cluster 2 was significantly higher than cluster 1, which was significantly higher than cluster 3; β1 of cluster 1 was significantly higher than cluster 3, which was significantly higher than cluster 2; and β2s were not significantly different among the 3 clusters (S5 Appendix)” (see page 12 lines 269-273). In addition, as explained in the methods section, we used a combination of criteria to determine the optimal number of clusters: Pseudo F statistic (a relatively large value), R-squared value (a peak that flattens with additional clusters), Cubic Clustering Criterion (≥2), and cluster size (≥10 participants). Based on these criteria, the optimal number of clusters was 3 for both percent weight change and PHQ-9 scores. This was a data-driven approach, and hence the results may be a characteristic of the data and not generalizable. We have also added a sentence in the limitation section: “Second, the weight change and PHQ-9 score clusters may be specific to the study data. Therefore, future studies are needed to replicate the results” (see page 20 lines 408-409).

2. Further, the conclusion line in the abstract ‘Participants with failed treatment outcomes demonstrated poor engagement or progress early in the intervention’ does not, to me, seem to be clearly demonstrated. Most of those with poor engagement had no data on either weight loss or depression. And particularly for depression scores, many of those in cluster 1 never-the-less had improvements in the early period. The authors need to argue this point more exactly.

Authors’ response: We defined treatment failure using weight and SCL-20 measures objectively obtained by trained research staff at baseline, 6, and 12 months (please see our response to editor’s comment #2 as well as the measures section, pages 6-7 and lines 152-161). Participants with failed treatment outcomes fell into two categories, those with poor engagement (n=63) and those with poor progress (n=80). Although those with poor engagement had insufficient data–not no data—on either self-monitored weight or PHQ-9 scores to be included in cluster analysis, these participants could be identified early in the intervention because the majority (79%) of them discontinued the intervention by session 6 (week 8). Participants with poor progress (mainly cluster 1 participants for percent weight change or PHQ-9) could also be identified early in the intervention because they showed limited progress by 6-8 weeks compared with those in clusters 2 and 3. To further strengthen our point, we have added the dropout rate by session 6 in the poor engagement category in the abstract (see page 2 lines 48-49) and manuscript text (see page 15 lines 312-313). We have also clarified our conclusion: “Participants demonstrating poor engagement or poor progress could be identified early during the intervention and were more likely to fail treatment at the end of the intervention. This insight could inform individualized and timely optimization to enhance treatment efficacy.”

Specific points:

3. The sensitivity analysis based jointly on weight and PHQ scores is particularly interesting since it yields a cluster whose members responded in PHQ but not in weight. This result is not evident (or at least, not emphasized) in the main cluster analysis. The concordance of this clustering result with the main analysis and the display in Table 3 might be informative.

Authors’ response: we have emphasized the concordance of the clustering results from the sensitivity analysis and the results from main cluster analyses and especially pointed out the group who responded in PHQ-9 but not in weight: “In addition, the number of participants in clusters resulting from the joint cluster analysis also showed concordance with the number of participants in Table 2. For example, the number of 33 participants with treatment progress in PHQ-9 only resulting from the joint cluster analysis (i.e., group B in S8 Appendix) was concordant with the number of 31 participants who had depression treatment progress (i.e., PHQ-9 cluster 2 and 3) but minimal weight loss (i.e., weight change cluster 1) in Table 2” (see pages 17-18 lines 350-355).

4. I suggest using, as a sensitivity analysis, clustering based not on the quarterly means but on an alternative averaging over time (e.g. 4-monthly). Further, an alternative clustering method without pre-specification of the number of clusters might help to test the robustness of the results.

Authors’ response: We did not pre-specify the number of clusters. As clarified in our response to this reviewer’s comment 1 above, we used a data-driven approach with multiple criteria. We have clarified this in the statistical analysis section: “First, the k-means method in the SAS FASTCLUS procedure without pre-specification of the number of clusters was used to group participants who had at least 1 measurement in each of the 4 quarters into clusters of individuals with similar patterns of change over time based on their 4 quarterly means. This step produced different numbers of clusters (range 2–6). Second, the optimal number of clusters was determined using a combination of criteria, including Pseudo F statistic (a relatively large value), R-squared value (a peak that flattens with additional clusters), Cubic Clustering Criterion (≥2), and cluster size (≥10 participants).[31] The optimal number of clusters was 3 for both percent weight change and PHQ-9 scores. Third, participants with percent weight change data in any 3 of the 4 quarters were assigned to their closest cluster defined by the smallest of the Euclidean distances between a participant’s 3 available quarterly means and each cluster’s means in the corresponding quarters” (see pages 7-8 lines 174-185).

We interpreted the reviewer’s comment on sensitivity analysis to suggest that we conduct a sensitivity cluster analysis using quarterly means calculated from monthly means (i.e., calculating monthly means first and then averaging the monthly means for quarterly means versus our approach which calculated quarterly means based on all available data in each quarter). Accordingly, we conducted the sensitivity analysis as suggested, and we applied the same criteria as described above for determining the optimal number of clusters. The results on percent weight change clusters were exactly the same as our original results. The results on PHQ-9 clusters still showed 3 clusters although the cluster memberships changed somewhat. Specifically 42 of 81 participants in Cluster 3 (Milder depression with treatment progress) in our method were classified in Cluster 2 (Moderate depression with treatment progress) in the sensitivity analysis—namely, the intercepts of clusters 2 and 3 were mostly affected in the sensitivity analysis. We believe this is an artifact of the “smoothing effect” of taking quarter means of monthly means in the sensitivity analysis. In accordance with the frequency of the 15 intervention sessions, PHQ-9 was administered weekly for 4 sessions, biweekly for 2 sessions, and monthly for the remaining 9 sessions. This taping-off schedule is typical of behavioral interventions. Given the session schedule, the sensitivity analysis does not differ from our analysis in calculating quarterly means for the latter 3 quarters but does affect the first quarter (and thus the intercept in particular) as it smoothed out the weekly and biweekly PHQ-9 scores in the first 2 months of the intervention and diminished the intercept and early treatment response of certain participants. As shown in the figures below, the sensitivity analysis (colored figure) moved the Cluster 2 regression line, particularly the intercept, downward. In contrast, our original method took advantage of all available PHQ-9 scores in the first quarter, more closely reflecting the real starting point and early response. Therefore, we decided to keep our method.

5. Figs. 1 and 2 display the individual trajectories but they are so small that it is difficult to get the picture. Could larger plots be shown?

Authors’ response: We have provided original-sized plots of the individual trajectories as Appendices (S2 and S4 Appendix).

6. The regression coefficient values given in appendix tables S2 and S3 differ from those in the text (pp. 9 and 10).

Authors’ response: The regression coefficient values in the text were model-based regression coefficients of each cluster (e.g., percent weight change cluster 1, 2, and 3). To verify that the resulting clusters could separate the individual trajectories within the clusters, we also provided mean (SD) of the regression coefficients of individual trajectories within each cluster in Appendices (current Appendix tables S3 and S5). ANOVA was then used to compare these regression coefficients of the individual trajectories among the 3 clusters for percent weight change and PHQ-9 separately. Therefore, the regression coefficients given in Appendix tables S3 and S5 were slightly different from those in the text and provided verification of the separation between clusters.

7. Many the ‘observed’ correlations and ‘significant’ differences are tautologous, stemming from the same or related data. For instance, the clusters were defined based on the weight and depression score changes, so it is almost inevitable that the clusters differ significantly with regard to the regression coefficients for these changes. It is similarly to be expected that the categories of treatment engagement and progress differ significantly with respect to adherence behaviors, since these behaviors were used to define the ‘no weight’ and ‘no PHQ-9’ clusters and thus the first category.

Authors’ response: We used a similar method that others have used (Babbin et al., Multivariate Behav Res. 2015, 50(1):91-108) to do cluster analysis and to verify that the derived clusters differed with regard to the regression coefficients of individual trajectories within the clusters. We have cited this paper in the statistical analysis section (page 7 line 166). We agree that percent weight cluster 0 (‘no weight’ cluster) included participants who had few weeks with self-monitored weight and PHQ-9 cluster 0 (‘no PHQ-9’ cluster) included participants who attended few sessions. In our response to the Editor’s comment 1, we underscored the potential clinical implications of these seemingly “intuitive” or “obvious” findings. To validate the distinctions of the clusters based on self-monitored weights and PHQ-9 scores over the course of the intervention, we compared weight and SCL-20 outcome data objectively obtained by trained research staff blinded to participants’ random assignment; results are shown in Table 3.

8. Did the authors test whether the polynomial models of trajectories fitted the data well, either to the individual trajectories or to the cluster means?

Authors’ response: We have clarified in the methods section that “We also tested whether the polynomial model for each cluster fit the data well using the significance of polynomial terms, adjusted R2, and the Bayesian information criterion (BIC). For both percent weight change and PHQ-9 score change, the polynomial regression models with a quadratic term fit the data better than the ones without given the significance of the quadratic terms, higher adjusted R2, and lower BIC. Hence, the final models included both linear and quadratic terms.” (see page 8 lines 194-199).

9. It would be interesting to investigate the possible chronological and causal relationships between weight change and depression score change: were they simultaneous or did one tend to follow the other? Or maybe no regularity can be discerned?

Authors’ response: We agree it would be interesting to test the chronological and causal relationships between weight change and depression score change. However, this is beyond the scope of this study because our data do not support the requisite design and analyses for examining temporally sensitive, causal relations. We have acknowledged that this is needed in future studies.

Submitted filename: Responses to reviewers_FINAL.docx

Click here for additional data file.

27 Dec 2019

PONE-D-19-15164R1

Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: within-treatment analysis in the RAINBOW trial

PLOS ONE

Dear Dr. Ma,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Feb 10 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Elena Barengolts, MD

Academic Editor

PLOS ONE

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

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: (No Response)

**********

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: Yes

**********

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: The authors have responded in detail to the comments and made appropriate additions to the manuscript.

My remaining reservation concerns the use of significance tests on differences in trajectory parameters between clusters to prove the distinctness of the clusters (points 1 and 7). The authors replied that their method of verification was the same as that used by Babbin et al (Ref. 31):

“We used a similar method that others have used (Babbin et al., Multivariate Behav Res. 2015, 50(1):91-108) to do cluster analysis and to verify that the derived clusters differed with regard to the regression coefficients of individual trajectories within the clusters.”

As I understand Babbin’s paper, no significance tests were performed on trajectory coefficients between clusters. Cluster-mean time-series parameters were displayed (Tab. 2), but without tests. The distinctness of the clusters was verified by displaying and testing cluster mean values of baseline variables which were not involved in the clustering (Tab. 3). This is an important difference to the method of the present study, in which the tested parameters concern the trajectories of the same variable used to create the clusters.

Therefore, I maintain that these significance tests do not prove the distinctness of the clusters. Even if one created purely random data from a single distribution, clusters could be found, and such clusters would show significantly different mean values of variables derived from those variables used to create the clusters. A proper assessment of distinctness should use a method described in the literature under the key-term ‘cluster validity’. In the absence of such an analysis, claims of distinctness and significance should be toned down.

**********

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: Jeremy Franklin

[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 to be viewed.]

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 us at figures@plos.org. Please note that Supporting Information files do not need this step.

28 Jan 2020

Authors’ response: As suggested by the reviewer, we toned down claims of the distinctness of clusters in the methods and results (lines 177-179, 233-237, and 257-261). Additionally, we emphasized the need for future validation studies with independent samples in the discussion. As described in the methods, cluster analysis was performed using the k-means method and the optimal number of clusters was decided using a combination of criteria, including Pseudo F statistic (a relatively large value), R-squared value (a peak that flattens with additional clusters), Cubic Clustering Criterion (≥2), and cluster size (≥10 participants). Although we currently do not have an independent sample to validate the 3 clusters of weight loss and the 3 clusters of depression symptoms, we stated in the discussion section that “The 3 weight loss patterns identified in the current study were similar to those found in previous weight loss studies.[34, 35]” Validation of the clusters of weight loss and depression symptoms is needed using independent samples in future studies. We emphasized this point in the discussion of the limitations (lines 387-390): “First, because of the post hoc nature of the analyses the findings need to be replicated in future studies. Second, the weight change and PHQ-9 score clusters may be specific to the study data. Therefore, future studies of independent samples are needed to verify the external validity of the results.”

Submitted filename: Response to reviewer comments_R2_FINAL.docx

Click here for additional data file.

31 Mar 2020

Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: within-treatment analysis in the RAINBOW trial

PONE-D-19-15164R2

Dear Dr. Ma,

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

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. 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 enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and 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.

With kind regards,

Elena Barengolts, MD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

8 Apr 2020

PONE-D-19-15164R2

Variability in engagement and progress in efficacious integrated collaborative care for primary care patients with obesity and depression: within-treatment analysis in the RAINBOW trial

Dear Dr. Ma:

I am 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 notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, 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.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Elena Barengolts

Academic Editor

PLOS ONE