Validity of clinical body weight measures as substitutes for missing data in a randomized trial

doi:10.1016/j.orcp.2008.09.002

Obesity Research & Clinical Practice

Volume 2, Issue 4, December 2008, Pages 277-281

https://doi.org/10.1016/j.orcp.2008.09.002 Get rights and content

Summary

Background

Long-term follow-up of weight loss interventions is essential, but collecting weights can be difficult, and self-reports inaccurate. We examined the relationship between weight measures obtained in the context of a weight loss trial and in routine clinical care.

Methods

Body weight data from a trial of behavioral obesity treatment among 88 obese women and 203 women age 40–65 years with comorbid obesity and depression were compared against weight data entered into an electronic medical record (EMR) during routine clinical care. Study and EMR weights and weight changes were then compared at 6 and 12 months using scatterplots, Pearson’s correlations, and t-tests.

Results

The 12-month follow-up rate for this trial was 77%. Among the 224 12-month completers, 142 women (63%) had an EMR weight within 90 days of their 12-month study weight. Study and EMR weights were highly correlated (0.99), with a mean difference of 0.1 kg. The correlation between two measures of 12-month weight change using study and EMR weights was 0.96. These results were robust to sensitivity analyses that explored the impact of different-sized windows for matching clinical weights with study weights. Among the 67 women who were missing study weights at 12 months, 33 (49%) had an EMR weight available within 90 days of their missed follow-up appointment.

Conclusions

Weight measures routinely obtained in clinical care are highly correlated with those obtained by trained research staff and may be used, without statistical correction, to achieve higher rates of long-term follow-up in weight loss studies.

Introduction

Obesity is now one of the leading health concerns in the United States. There is a strong demand for clinical research studies to meet the needs of an increasingly obese population. However, the area of obesity research has long been troubled by high rates of loss to follow-up, which approach 40% in many long-term clinical trials of obesity interventions. Participant attrition reduces the validity of the study results, because any subsequent analysis of weight data is vulnerable to bias resulting from systematic differences between participants who completed the trial and those who did not [1], [2]. Self-reported follow-up weights are poor surrogates for measured weights because of recall bias and embarrassment [3]. As a result, considerable effort has been placed on statistical methods for handling missing weight measures in obesity trials [4]. However, none of the available imputation methods can ensure a valid result when rates of attrition are high.

One potential strategy for reducing missing data in clinical trials of obesity interventions is to use weight measures that are collected during routine clinical care. This option is particularly attractive when participant recruitment occurs in the setting of an integrated health care delivery system with an electronic medical record (EMR). Little is known about the accuracy of weights obtained during routine clinical care. Two previous studies found close agreement between research study weights and routine clinical weights among 85 patients admitted for elective surgery [5] and 64 patients enrolled in a managed care system’s behavioral weight loss program [6], respectively. Here we report on the availability and accuracy of weights collected during routine outpatient care, examining the agreement between research weights and clinical weights for both absolute weight and weight change over 12 months for a total of 291 women.

Section snippets

Research methods and procedures

We conducted this study at Group Health, a mixed-model prepaid health plan and delivery system serving approximately 500,000 members in Washington State and northern Idaho. The Group Health institutional review board reviewed and approved all study procedures.

Results

Table 1 shows the demographic characteristics of the 291 women in our study population. Baseline study weights were available on all participants. Their mean age was 52.2 years, 81% were white, and 54% were married. Their mean body mass index (BMI) was 38.5 kg/m², with 32% having a BMI of 40 kg/m² or greater.

Follow-up study weights were obtained by research staff on 238 participants at 6 months (82% 6-month completers) and 224 participants at 12 months (77% 12-month completers) (Table 2). Among

Discussion

We found that weight measures obtained by nursing staff during routine clinical care were highly correlated with those obtained by trained research staff in the context of a prospective trial of behavioral treatment for obesity and comorbid depression. Our findings are consistent with the results of two previous studies that found close agreement between weights obtained by research and clinical staff among 85 patients admitted for elective surgery [5] and among 64 patients enrolled in a

Acknowledgments

This project was supported by NIH Research Grant #MH68127 funded by the National Institute of Mental Health and the Office of Behavioral Social Sciences Research.

References (7)

M.F. Kuczmarski et al.
Effects of age on validity of self-reported height, weight, and body mass index: findings from the Third National Health and Nutrition Examination Survey, 1988–1994
J Am Diet Assoc
(2001)
V.J. Stevens et al.
Validity and usefulness of medical chart weights in the long-term evaluation of weight loss programs
Addict Behav
(1988)
J.A. Linde et al.
Relation of body mass index to depression and weighing frequency in overweight women
Prev Med
(2007)

There are more references available in the full text version of this article.

Cited by (36)

Parental concerns about picky eating and undereating, feeding practices, and child's weight
2022, Obesity Research and Clinical Practice
Citation Excerpt :
The parent’s height and weight was measured by research staff using the same wall-mounted stadiometer and mechanical beam scale. Clinical measurements have been shown to have good accuracy compared to height and weight obtained for research purposes [33]. Height and weight were used to calculate BMI.
Parents’ concerns about their child’s feeding may lead parents to pressure their child to eat, which may lead to a greater risk for obesity. We aimed to assess if parental concerns for picky eating and undereating are associated with pressure to eat and increased child BMI z-score (BMIz).
We performed a cross-sectional study of 328 parents of healthy preschoolers assessing parent concerns about picky eating (Child Eating Behavior Questionnaire) and child undereating (“Are you concerned …doesn’t eat enough?”), parent pressure to eat (Child Feeding Questionnaire), and covariates. Dyads’ heights and weights were measured. Structural equation modeling (SEM) was performed to examine the relationships between parental concerns, pressure to eat, and child BMIz. Measurement models were tested and refined, and the structural model was tested. Model fit was determined using multiple goodness-of-fit indices.
Dyads were racially and socioeconomically diverse. The SEM model demonstrated good goodness-of-fit. Children who were perceived as not eating enough had significantly higher picky eating scores (β 0.756; p < 0.001). Parents had higher pressure to eat scores if children were more picky (β 0.148; p = 0.02) or were perceived as not eating enough (β 0.654; p < 0.001). Parental pressure to eat was not associated with the child’s BMIz.
In a cohort of diverse preschoolers, parent concerns about eating were associated with increased pressure to eat, but pressure to eat was not associated with BMIz. Identifying these relationships is important to develop effective interventions to improve feeding practices in young children.
Weight loss prior to pregnancy and subsequent gestational weight gain: Prepare, a randomized clinical trial
2021, American Journal of Obstetrics and Gynecology
Women with elevated body mass index are encouraged to lose weight before pregnancy, but no trials have tested the effects of prepregnancy weight loss on health outcomes.
This study aimed to determine whether prepregnancy weight loss reduces gestational weight gain and improves pregnancy outcomes.
Pragmatic randomized clinical trial was conducted between May 2015 and October 2019 at Kaiser Permanente Northwest, an integrated health system. Data collection was blind to condition assignment. Eligible participants were women aged 18 to 40 years with a body mass index of ≥27 kg/m² who were planning pregnancy within 2 years. Recruitment contacts were sent to 27,665 health system members who met age and body mass index criteria; 329 women attended screening visits, and 326 were randomized. They were randomized to either a behavioral weight loss intervention or usual care control. The intervention consisted of health coaching phone sessions weekly for 6 months and then monthly for 18 months or until end of pregnancy. We used logistic regression to examine the a priori primary hypothesis that participants in the intervention would be less likely to exceed National Academy of Medicine guidelines for gestational weight gain during each trimester and overall. Secondary and exploratory outcomes included absolute weight gain before and during pregnancy and perinatal and newborn outcomes.
Of the 326 participants, 169 had singleton pregnancies lasting ≥14 weeks (analytical cohort: intervention, 89; control, 80). At baseline, mean age was 31.3±3.5 years, and body mass index was 34.8±5.8 kg/m². Participants in the intervention group lost more weight before pregnancy than those in the control group (−0.25±0.51 vs −0.03±0.21 kg/wk; P<.001). However, participants in the intervention group gained more weight than those in the control group in the second trimester (0.42±0.26 vs 0.33±0.28 kg/wk; P=.04) and third trimester (0.56±0.37 vs 0.43±0.33 kg/wk; P=.02) and overall (13.2±8.20 vs 10.3±7.41 kg; P=.03). Nevertheless, arms did not differ in rates of exceeding gestational weight gain guidelines at any time point. Spontaneous pregnancy loss was less common in the intervention arm than in the control arm (8 [4.9%] vs 19 [11.8%]; odds ratio, 0.39 [0.16–0.92]), but we found no other differences in the secondary or exploratory outcomes.
Participation in the prepregnancy weight loss intervention had no effect on women’s likelihood of exceeding gestational weight gain guidelines. Although the intervention group successfully lost weight before conception, the intervention group was associated with greater weight gain in late pregnancy. To effectively reduce weight throughout pregnancy and improve maternal and child outcomes, prepregnancy weight loss interventions may need to be combined with intensive weight management that continues throughout delivery.
Weight loss after bariatric surgery and periodontal changes: a 12-month prospective study
2017, Surgery for Obesity and Related Diseases
Citation Excerpt :
Third, anthropometric measurements were not purposefully collected for this study, but extracted from medical records. Although this may raise concerns about increased measurement error, there is evidence that weights documented in medical records are convertible with weights recorded in a research setting [25,26], particularly among women, who represented 88% of our study group [27,28]. Fourth, our regression models did not include a control variable for socioeconomic position, which is strongly related to both obesity [29] and periodontal disease [30].
Several longitudinal studies have explored the association of obesity and weight gain with periodontal disease. However, the effect of weight loss on periodontal tissues remains unclear.
To explore whether weight loss after bariatric surgery was associated with changes in periodontal measures over 12 months.
Two public hospitals in São Paulo, Brazil.
We used data from 110 morbidly obese patients (body mass index [BMI]>40 kg/m² or≥35 kg/m² with co-morbid conditions) who underwent bariatric surgery between April 2011 and March 2013. Data on demographic factors, BMI, smoking habits, and glucose levels were extracted from medical records preoperatively and after 6 and 12 months postsurgery. A full-mouth periodontal examination was conducted by trained examiners to assess probing pocket depth, clinical attachment loss, and bleeding on probing (BOP) at baseline and 6 and 12 months after surgery. Data were analyzed using linear mixed-effects models.
BMI was not significantly related to the proportion of sites with BOP at baseline, but it was negatively associated with the rate of change in the proportion of sites with BOP. The greater the BMI loss, the higher the proportion of sites with BOP, particularly 6 months after surgery. However, BMI was not associated with baseline probing pocket depth and clinical attachment loss or rates of changes in these periodontal outcomes.
The findings suggest that weight loss was associated with increased gingival bleeding, showing a peak at 6 months after bariatric surgery. Periodontal pocketing and attachment loss remained unchanged during the study period.
Prepare, a randomized trial to promote and evaluate weight loss among overweight and obese women planning pregnancy: Study design and rationale
2016, Contemporary Clinical Trials
Citation Excerpt :
We carefully considered the best way to collect outcomes given limited resources and our recruitment population of young, busy women [72]. Because we and others have previously documented the high accuracy of EMR-recorded weights in pregnant and nonpregnant populations [63–65,73] and because of the high retention rate in our health system, we feel we can maximize resources by using EMR-recorded weights. We also believe that we can recruit and retain a more diverse population by using EMR instead of requiring follow-up measurements be done in the research center.
Women who are overweight or have obesity at pregnancy onset, and those who gain excessive weight during pregnancy, are at increased risk of pregnancy-related complications and large for gestational age infants.
This report describes methodology for the Prepare study, a randomized, controlled clinical trial testing a preconception and pregnancy weight management program for women who are overweight or have obesity (BMI ≥ 27 kg/m²).
This trial examines multiple pregnancy and neonatal outcomes, with the primary outcome being gestational weight gain (GWG). Secondary outcomes include change in weight before conception, offspring birth weight adjusted for gestational age, offspring weight for length, and pregnancy diet quality and physical activity level.
Nonpregnant women who anticipate becoming pregnant in the next 2 years are randomly assigned to an intervention program or a usual care control condition. Intervention participants receive weight management counseling by telephone before and during pregnancy, with weekly contacts during the first 6 months and monthly contacts for the next 18 months. Intervention participants also have unlimited access to a study website that provides self-management tools. All participants who become pregnant are contacted at 20 weeks' gestation to assess physical activity levels and dietary habits. All other outcome data are obtained from medical records. Intervention satisfaction is assessed via questionnaire.
This clinical trial tests the efficacy of an intervention program designed to help overweight and obese women achieve healthy lifestyle changes that will result in a healthy weight prior to pregnancy and appropriate weight gain during pregnancy.
Comparison of weight captured via electronic health record and cellular scales to the gold-standard clinical method
2023, Obesity Science and Practice
A Pilot Study Examining the Association of Parental Stress and Household Food Insecurity with Dietary Quality in Pre-School-Aged Children
2023, Nutrients

View all citing articles on Scopus

View full text

Validity of clinical body weight measures as substitutes for missing data in a randomized trial

Summary

Background

Methods

Results

Conclusions

Introduction

Section snippets

Research methods and procedures

Results

Discussion

Acknowledgments

J Am Diet Assoc

Addict Behav

Prev Med