Utility of Self-rated Adherence and Assessment of Participant Feedback of the Healthy Diet and Lifestyle Study Pilot

Holly O’Reilly Technological University Dublin Chloe Panizza (  cpanizza@hawaii.edu ) University of Hawai'i Cancer Center https://orcid.org/0000-0003-4278-1602 Unhee Lim University of Hawai'i Cancer Center Kim M Yonemori University of Hawai'i Cancer Center Lynne R Wilkens University of Hawai'i Cancer Center Yurii B Shvetsov University of Hawai'i Cancer Center Michelle N Harvie Manchester University NHS Foundation Trust John Shepherd University of Hawai'i Cancer Center Fengqing Maggie Zhu Purdue University Loïc Le Marchand University of Hawai'i Cancer Center Carol J Boushey University of Hawai'i Cancer Center Kevin D Cassel University of Hawai'i Cancer Center

Research on lifestyle interventions aimed at reducing VAT have primarily been quantitative in nature (15)(16)(17)(18)(19). However, a mixed-methods approach, using both quantitative and qualitative data, would help provide a more complete picture of study effectiveness, including participants' perspectives (20). This is especially important in nutrition interventions aimed at changing behaviors (20), and may assist in identifying factors in uencing study adherence.
Previously our team reported the primary results of the randomized Healthy Diet and Lifestyle study (HDLS) pilot (21), aimed at reducing VAT among East Asian American adults. VAT and weight decreased in both study arms, but signi cantly more in the IER + MED group than the DASH group at 12 weeks (21). This secondary analysis aims to build on the primary analysis by assessing the utility of a self-rated scale used to monitor participants' dietary and physical activity compliance and facilitate goal setting during HDLS. We also aim to evaluate HDLS by analyzing participants' qualitative feedback collected upon completion and at 6-months post intervention.

Study design
The HDLS pilot study was a 12-week randomized trial conducted at the University of Hawaii Cancer Center (UHCC) between September 2016 and October 2017 to demonstrate feasibility and e cacy of following the IER+MED diet as an approach to reduce VAT. Extensive details of the intervention are provided in a previous publication (21). Brie y, eligibility included being of East Asian ancestry (Japanese, Chinese or Korean), residing in Honolulu County, BMI between 25 and 40 kg/m 2 , ages 35 to 55 years, no pregnancy, and no serious health issues. Inclusion criteria included normal blood count and biochemistry pro le and whole-body dual-energy x-ray absorptiometry (DXA)-derived VAT at L4-L5 ≥ 90 cm 2 for men and ≥ 80 cm 2 for women as determined at the eligibility clinic visit.
HDLS included baseline and Week 12 measurements of anthropometry and DXA (21). For the current analysis, outcomes of interest include body weight and VAT. One participant was excluded from analyses as s/he dropped out during the rst week of HDLS and, thus, had no self-rated adherence data.
The study protocol [NCT03639350] was approved by the institutional review board at the University of Hawaii at Manoa. Study volunteers provided written informed consent. Figure 1 is a CONSORT diagram and the CONSORT and TIDieR checklists are provided as supplementary material (Additional le 1 and 2, respectively).

Randomization and masking
Once eligibility was con rmed, strati ed and blocked randomization was used to allot participants into either the IER+MED (n=30) or DASH (n=30) group, to ensure balance between groups. Randomization strata were de ned by sex and high (80 cm 2 for women or 90 cm 2 for men to <150 cm 2 ) or very high (≥150 cm 2 ) VAT. The assigned diets were labeled as either Diet A or Diet 1 to blind participants, recruitment, and clinic staff to IER+MED or DASH group allocation. Research dietitians, who were responsible for intervention activities, were blinded to participant measures except diet and body weight.

Procedures and outcomes
Dietary and physical activity prescriptions Diet and physical activity prescriptions in HDLS have been reported previously (21). Brie y, the intervention group was assigned a IER+MED diet, and the active comparator group was assigned a euenergetic DASH diet for 12 weeks. Dietary materials for IER+MED were originally developed and tested for use among white women in the United Kingdom (15) and were modi ed to include examples of foods and beverages more readily available in Hawaii (21). Both groups were advised to walk one hour daily, ve days per week, to reduce confounding due to physical activity. Participants in the IER+MED group were advised to exercise only on their MED days.
During an in-person dietary consultation (~45-60 minutes) participants were provided with a personalized, group-speci c, diet booklet (depicting serving sizes within food groups and examples of foods), individualized food lists and menus, and trackers to encourage compliance to prescriptions. Dietitians called participants once a week between Weeks 1 and 4, and at Week 6, 8 and 10, and conducted an in person follow-up at Week 12, to assess participants' adherence to their intervention plans (22) and provide guidance using motivational interviewing principles (23;24).

Assessment of adherence to prescribed diet and physical activity prescriptions
During follow-up calls conducted by study dietitians and the in-person dietitian consultation at Week 12, all participants were asked: "How well have you been following your diet plan? On a scale of zero to ten with zero being not at all, four being somewhat, and ten being following the plan very well, where would you place yourself?". Similarly, the same questions were asked for adherence to physical activity prescriptions. Self-rated adherence scores assisted with identifying participant's barriers to change and setting personal diet and physical activity goals to achieve by the next follow-up call (23;25).
These assessment questions were adapted from those used to assess motivation and con dence to change dietary behaviors used by Resnicow et al. (23;24;26). For example, in the Body & Soul Study, participants were asked "on a scale of 1 to 10 (with 10 being the highest), how motivated or interested are you in increasing your fruit and vegetable consumption?" Resnicow et al. found that this time and costeffective assessment technique assisted with increasing fruit and vegetable intake among African Americans in a church setting.
Responses to self-rated adherence scores across the 12 week HDLS were averaged for each person, and participants were divided into a high or low level of adherence, split by median score (7.3 for diet and 7.1 for physical activity).

Qualitative Feedback
During the clinic visit at Week 12, participants completed a self-administered exit questionnaire. The two questions relevant to this analysis were if he/she would have been interested in cooking classes or demonstrations and whether the materials provided were su cient to be able to follow the diet plan. Participants recorded their responses using a 5-point Likert scale (strongly agree/agree/neither agree or disagree/disagree/strongly disagree).
At 6-months post-intervention, recruitment staff performed a follow-up, by telephone, of participants who completed the HDLS. Participants were interviewed using a standardized questionnaire tailored to the study objectives. Questions pertinent to this analysis included "What type or types of exercise have you been doing since you completed the study?"; "Are you still trying to follow the diet you were assigned to during the study?"; "If no, can you describe the type of diet that you have been following since you completed the study?"; and "Do you have any comments or suggestions as to how the study could be improved?".

Statistical analysis
Analyses were limited to participants with complete Week 1 study data (n=59). Multiple imputation was used to replace missing values of the outcome and exposure variables and generate ve imputed datasets. Less than 5% of data values were missing. Mean changes in VAT and weight loss, from baseline to 12 weeks, were computed by self-reported adherence level, overall and by randomization arm, and compared between adherence levels using a t-test. Responses to 5-point Likert scale questions are reported as frequencies and percentages. Quantitative analyses were performed using IBM SPSS Statistics version 26 (IBM Corp., Armonk, NY, USA) and statistical signi cance was de ned as p<0.05.
Qualitative methods were used to evaluate responses to the open-ended 6-month post-intervention telephone interview questions (27). Responses to open-ended questions from each participant were transcribed separately by staff members not involved in the study. Co-investigators (KC, HO) and another staff member independently coded each response transcript. Each coder identi ed and nominated common themes and preliminary codes, and all codes were reviewed and discussed by the team until a nal consensus was reached. A codebook was then developed for analysis using NVIVO Version 11 (QSR International, Melbourne, Australia) and used to drive a subsequent thematic analysis of all interview transcript data (27).

Characteristics
Of the 60 participants enrolled in HDLS, six (10%) dropped out (IER+MED:4; DASH:2), with only one participant withdrawing from the study before the Week 1 follow-up call. Of the 54 participants completing HDLS, all 54 participants completed the Exit Questionnaire and 48 participants responded to the six-month post-intervention telephone interview. Of participants included in the current analysis, 29 (48.2%) were randomized to IER+MED and 30 (50.8%) to DASH ( Table 1). The study participants were mostly women (69.5%), participants with high VAT (67.8%), and of Japanese ancestry (62.7%). Baseline characteristics were similar for the 59 participants included in this analysis, the 54 participants who completed the Week 12 visit and the 48 who completed the 6-month post-intervention survey.
Splitting data by median adherence, baseline characteristics were similar between dietary adherence groups and between physical activity adherence groups ( Table 1). The largest differences in adherence were seen between ethnic groups. For dietary adherence, 71.4% (n=5) participants of Korean or Mixed Asian ancestry were in the low category. For physical activity adherence, 100% (n=7) of participants with Korean ancestry were in the low group, and 75% (n=6) of participants of Chinese ancestry were in the high adherence group. 1 Self-rated dietary adherence ranging from zero being not at all to ten being following the plan very well, split by median adherence (7.3). 2 Imputed values for missing adherence data. 3 Self-rated physical activity adherence ranging from zero being not at all to ten being following the plan very well, split by median adherence (7.1). 4 Column percentages for overall column and row percentages for data by self-rated adherence level. 5 Intermittent energy restriction combined with a Mediterranean diet (IER+MED). 6 Euenergetic Dietary Approaches to Stop Hypertension diet (DASH). 7 Women at ≥80 cm 2 and men at ≥90 cm 2 .

Adherence
Overall, mean ± SE, dietary adherence over 12 weeks was 6.0 ± 0.2 and 8.2 ± 0.1, for the low and high adherence groups, respectively. Range of dietary adherence scores were 2.9-7.7 and 7.1-9.9 respectively.
For physical activity adherence, mean scores were 5.9 ± 0.2 and 8.5 ± 0.2 for the low and high adherence groups, and ranged from 3.0-7.2 and 7.0-10.0, respectively. Compared to participants with low self-rated adherence to dietary prescriptions, those with high adherence lost signi cantly more VAT (22.9 ± 3.7 cm 2 vs. 11.7 ± 3.9 cm 2 ; p=0.044) and weight at Week 12 (5.4 ± 0.8 kg vs. 3.5 ± 0.6 kg; p=0.045) ( Table 2). For physical activity, compared to participants with low adherence, those with high adherence lost signi cantly more VAT (22.3 ± 3.7 cm 2 vs. 11.6 ± 3.6 cm 2 ; p=0.035) ( Table 3). Weight loss was also greater for those with high vs. low adherence to physical activity prescriptions (5.0 ± 0.7 kg vs. 3.7 ± 0.7 kg); however, these differences were not signi cant. Within study arm comparisons, high dietary adherence and high physical activity adherence had greater VAT and weight loss than their counterpart low adherence groups, but these differences were not signi cant (Tables 2 and 3). Repeating analyses using % change in VAT and % change in weight instead of absolute change in VAT and weight produced similar results.  Qualitative Feedback Answering the question, "materials provided were su cient to be able to follow the diet plan", 44 (81%) participants rated that they strongly agree or agree, 8 (15%) that they neither agree or disagree, and 2 (4%) participants did not respond. For the question "Would in-house classes or cooking demonstrations have been of interest?" 36 (67%) participants answered yes, 12 (22%) reported no, 2 (4%) were undecided, and 4 (7%) did not answer the question. Women tended to be more interested in the in-house classes or cooking demonstrations than men (75% and 47%, respectively).
The thematic structure identi ed for the open-ended questions as part of the 6-month post-intervention telephone interview followed the topics of the survey questions including 1) Exercise; 2) Diets; 3) Comments; and 4) Suggestions (27). Participant responses were summarized based on this thematic structure.
Exercise. Almost all (n=46, 96%) participants reported continuing exercise at 6-months post-intervention. Overall, respondents reported 60 types of exercise regimens they adopted or continued after the study, with 31 (65%) reporting "walking" as a primary form of exercise. Other popular exercise included running, swimming, paddling, tennis, golf, cycling, weightlifting, shing, Zumba, Step Aerobics, Aqua Aerobics, high intensity interval training, calisthenics, stair climbing, and use of a gymnasium.
Diets. Approximately half of the respondents reported "yes", they were still trying to follow their prescribed diet at 6-months post-intervention (IER+MED, 66.7%; DASH, 44.0%). Of those participants not following the diet at 6-months post-study, many (n=8) reported focusing on a healthy diet without mentioning the speci c dietary changes. For example, "Your study increased my awareness regarding foods and I am eating healthier, and use the basics of the study as a guide." Also, although no longer following their prescribed diet, participants reported substituting nutrient poor items for nutrient dense choices (n=4), reducing their portion size (n=9), decreasing intake of sugar, meat or carbohydrates (n=7), and increasing intake of fruit and vegetables (n=14). An example being, "gave up some snack foods and unhealthy food. More aware of healthier options." In addition, participants reported adopting a modi ed version of the study diet with the re-adoption of non-study foods (n=6), for example, "following protocol, but added dark chocolate." Comments. Under this theme, praise for the support of the intervention dietitians was the most frequent comment (n=14). For example, "Helpful to have dietitian suggestions and accountability." Unfamiliarity with the foods prescribed during the study, and preparation of these foods (n=2) was the second most frequent comment.
Suggestions. The most common response, on how to improve the study, was to provide cooking education (n=8) in the form of classes delivered either in-person or online; and the creation of a cookbook to support the study. For example, "would really appreciate cooking classes." The Identi cation of more varied foods to support the study was also suggested, including the identi cation of premade foods (n=3) that would be acceptable for use, and longer study duration (n=1).
Discussion Among all participants, higher self-rated adherence to dietary or physical activity prescriptions was associated with signi cantly greater loss of VAT on completion of the 12-week HDLS pilot. This is the rst known randomized trial to report an association between self-rated adherence and VAT loss. The association found between dietary adherence and weight loss supports previous ndings, with higher adherence resulting in greater loss (10;14).
The single-item questions used to assess adherence in HDLS were adapted from questions used by Resnicow et al (23;24). Resnicow et al. demonstrated that these single-item questions, based on motivational interviewing, were effective at evaluating con dence and motivation to change, and for eliciting motivational messages and barriers to change (23;24;26). Similarly, Dansinger et al used a 0-10 scale to assess participants self-rated adherence and the effectiveness of 4 popular diets for weight loss (10). In both the current analysis and the Dansigner et al study, greater dietary adherence was associated with weight loss. These results support a single-item question can be used to assess intervention compliance and assist with facilitating behavior change during an intervention.
Several other studies have veri ed the agreement between self-evaluation and behavior. A cross-sectional study by Adjoian et al. assessed the validity of self-rated overall diet quality compared to Healthy Eating Index-2010 (HEI-2010) scores among a multiethnic adult population in New York City (NYC) (28). Those with lower self-rated diet quality had signi cantly lower HEI-2010 scores. In regards to physical activity, a single-item questionnaire, assessing participants activity in the previous week, has been validated against accelerometer data (29). The question asks ''In the past week, on how many days have you done a total of 30 min or more of physical activity, which was enough to raise your breathing rate? This may include sport, exercise and brisk walking or cycling for recreation or to get to and from places, but should not include housework or physical activity that may be part of your job". These studies add to the value of short self-report questionnaires, where participant responses matched direct objective measures of study outcomes. However, the single-item questions utilized in HDLS are unique in that the questions are broad; therefore, theoretically, they could be applied to assess self-rated compliance to a range of intervention prescriptions.
Key strategies for dietary adherence include designing weight loss studies that limit hunger, are tailored to participants dietary preferences, and promote self-monitoring of dietary intake (9). All these strategies were employed in HDLS (21), which may explain why both low and high dietary adherence groups successfully lost VAT and weight, mean dietary adherence for both study arms was relatively high (IER + MED, 7.3; DASH, 6.8), and participants' positive re ections of the study. However, modi cations to the design of HDLS may help to further improve dietary adherence. For example, responses to both the Exit Questionnaire and 6-month post-study telephone interview indicated that most participants (~ 2/3) were interested in cooking classes and demonstrations. The integration of these classes may help to further personalize dietary prescriptions, reduce any unfamiliarity of foods prescribed, increase participants con dence in preparing meals (30), and ultimately improve study compliance, long-term adherence and adoption.
HDLS participants did not provide any comments or suggestions on how to improve physical activity prescriptions. However, ethnicity appeared to be a large driver for physical activity adherence, with all participants of Korean ancestry (n = 7) being in the low adherence category. A review on adherence to lifestyle modi cations programs for weight management highlights strategies which may help improve physical activity compliance for those in the low adherence category (9). The authors identi ed psychosocial factors, socio-demographic factors, behavioral factors, and physical factors as in uencing adherence. For example, psychological factors in uencing adherence include self-e cacy, depression, motivation, stress, body shape concern, quality of life and stage of change. Socio-demographic factors include age, gender, employment status and education (9). Focusing on reasons for ethnic differences in physical activity adherence and additional predictors of adherence, may help to improve physical activity compliance scores in future trials.
The strengths of the current study include the evidence-based strategies implemented to ensure participant engagement and compliance. These include the adoption of dietary protocols by Harvie et al (15), and behavior change strategies from Body and Soul (23;24). A limitation was the small sample size which may have reduced statistical power to show signi cant associations between adherence and changes in VAT and body weight by study arms. Also, the association seen between weight loss and adherence level may have been due to reverse causation. Participants with greater weight loss may have thought they were adhering to the dietary and physical activity prescriptions more closely, and rated their adherence higher. However, change in VAT is harder for participants to self-monitor and change in VAT between adherence arms was proportionately larger than change in weight. Therefore, it is unlikely VAT results were due to reverse causation. Another limitation is change in VAT and weight were used as proxy measurements for adherence to intervention prescriptions. To assess the validity of self-rated compliance to monitor intervention adherence, future studies should compare self-rated adherence results to more direct objective measurements as opposed to proxy methods (e.g., compare self-rated physical activity adherence to accelerometer data).

Conclusions
Results support the utility of self-rated compliance as a method to monitor dietary and physical activity adherence and facilitate participant goal setting during a behavioral modi cation intervention aimed at reducing VAT and weight. Overall, participant feedback on HDLS was positive, and incorporation of cooking classes and demonstrations into future larger trials may help to further improve dietary adherence.