Probability of an Obese Person Attaining Normal Body Weight
Probability of an Obese Person Attaining Normal Body Weight
We conducted a population-based cohort study using primary care electronic health records from family practices in the United Kingdom. We selected a cohort of adults aged 20 years and older from the UK Clinical Practice Research Datalink (CPRD), an anonymized database of longitudinal patient electronic medical records from primary care. The CPRD is the world's largest primary care database and contains more than 7% of the UK population, with nearly 700 general practices contributing data that meets quality standards for research. The CPRD data are considered to be broadly representative of the UK population in terms of patient demographic characteristics and the size and distribution of practices. Our research was part of a larger study to evaluate the use of bariatric surgery. We aimed to estimate body weight transitions in the absence of bariatric surgery; therefore, we excluded participants who received bariatric surgery.
There were 2,006,296 patients registered in the CPRD between November 1, 2004, and October 31, 2014, who were aged 20 years or older and had 3 or more BMI records. A minimum of 3 BMI records per patient was required to estimate weight changes, including weight regain following weight loss. The annual count of the CPRD registered population aged 20 years and older peaked at 3.7 million during this period, with a total of 7.1 million participants aged 20 years or older registered at any time during the period.
We classified participants according to the BMI value of their first record into 6 categories: 18.5 to 24.9 (normal weight), 25.0 to 29.9 (overweight), 30.0 to 34.9 (simple obesity), 35.0 to 35.9 (severe obesity), 40.0 to 44.9 (morbid obesity), and 45.0 or greater (superobesity) kilograms per meters squared. We selected a random sample of up to 30,000 participants, using the sample command in Stata version 13 (StataCorp LP, College Station, TX), from each category of BMI and gender, resulting in 314,477 participants. There were fewer than 30,000 women with a BMI of 45 or greater kilograms per meters squared and fewer than 30,000 men with a BMI of either 40 to 45 or 45 or greater kilograms per meters squared.
We then extracted full CPRD records for this sample. We analyzed data for research quality records for each participant. The start was the latest of November 1, 2004, the participant registration date, or the general practice CPRD start date. The end date was the earliest of October 31, 2014, the date death or end of registration, or the last data collection date for the general practice. We excluded the 2738 (1%) participants who had bariatric surgery and the 32,757 (10%) participants who had fewer than 3 BMI values recorded between November 1, 2004 and October 31, 2014; this left 278,982 participants for further analysis.
We conducted a longitudinal analysis of BMI records. The start date for each participant was the later of November 1, 2004 or the beginning of the patient's CPRD record. The end date was the earlier of October 31, 2014 or the end of the patient's CPRD record. We used the first BMI record after the participant start date as the index BMI, and we used the date of this record as the index date. We evaluated the number of BMI records for each BMI category, and we calculated the number of records showing either an increase or a decrease in BMI category. For patients who showed a decrease in BMI category, we evaluated whether subsequent changes in BMI category were increases or further decreases.
We analyzed data in a time-to-event framework to evaluate the proportion of patients from each starting BMI category who attained either, first, a normal body weight or, second, a 5% reduction in body weight during the 9-year follow-up. We did not envisage a 5% reduction in body weight in the original study protocol, but we added it as a minor amendment because this is a widely recommended target for body weight reduction. In the first analysis, we estimated the annual probability of attaining normal body weight. We used the number of events (BMI category recorded as < 25 kg/m) and the person-years of follow-up to estimate the annual rates and their confidence intervals (CIs), which we converted to annual probabilities using the formula 1 – exp.
Among participants who reduced their BMI category, we evaluated the direction of the next change in BMI category. In the second analysis, to examine the proportion of participants who lost 5% of their body weight, we also evaluated the development of a body weight that was more than 95% of the initial body weight in a time-to-event framework. We conducted analyses in Stata version 13 using the stset, sts list, and stcox commands.
Methods
We conducted a population-based cohort study using primary care electronic health records from family practices in the United Kingdom. We selected a cohort of adults aged 20 years and older from the UK Clinical Practice Research Datalink (CPRD), an anonymized database of longitudinal patient electronic medical records from primary care. The CPRD is the world's largest primary care database and contains more than 7% of the UK population, with nearly 700 general practices contributing data that meets quality standards for research. The CPRD data are considered to be broadly representative of the UK population in terms of patient demographic characteristics and the size and distribution of practices. Our research was part of a larger study to evaluate the use of bariatric surgery. We aimed to estimate body weight transitions in the absence of bariatric surgery; therefore, we excluded participants who received bariatric surgery.
Sample Selection
There were 2,006,296 patients registered in the CPRD between November 1, 2004, and October 31, 2014, who were aged 20 years or older and had 3 or more BMI records. A minimum of 3 BMI records per patient was required to estimate weight changes, including weight regain following weight loss. The annual count of the CPRD registered population aged 20 years and older peaked at 3.7 million during this period, with a total of 7.1 million participants aged 20 years or older registered at any time during the period.
We classified participants according to the BMI value of their first record into 6 categories: 18.5 to 24.9 (normal weight), 25.0 to 29.9 (overweight), 30.0 to 34.9 (simple obesity), 35.0 to 35.9 (severe obesity), 40.0 to 44.9 (morbid obesity), and 45.0 or greater (superobesity) kilograms per meters squared. We selected a random sample of up to 30,000 participants, using the sample command in Stata version 13 (StataCorp LP, College Station, TX), from each category of BMI and gender, resulting in 314,477 participants. There were fewer than 30,000 women with a BMI of 45 or greater kilograms per meters squared and fewer than 30,000 men with a BMI of either 40 to 45 or 45 or greater kilograms per meters squared.
We then extracted full CPRD records for this sample. We analyzed data for research quality records for each participant. The start was the latest of November 1, 2004, the participant registration date, or the general practice CPRD start date. The end date was the earliest of October 31, 2014, the date death or end of registration, or the last data collection date for the general practice. We excluded the 2738 (1%) participants who had bariatric surgery and the 32,757 (10%) participants who had fewer than 3 BMI values recorded between November 1, 2004 and October 31, 2014; this left 278,982 participants for further analysis.
Analysis
We conducted a longitudinal analysis of BMI records. The start date for each participant was the later of November 1, 2004 or the beginning of the patient's CPRD record. The end date was the earlier of October 31, 2014 or the end of the patient's CPRD record. We used the first BMI record after the participant start date as the index BMI, and we used the date of this record as the index date. We evaluated the number of BMI records for each BMI category, and we calculated the number of records showing either an increase or a decrease in BMI category. For patients who showed a decrease in BMI category, we evaluated whether subsequent changes in BMI category were increases or further decreases.
We analyzed data in a time-to-event framework to evaluate the proportion of patients from each starting BMI category who attained either, first, a normal body weight or, second, a 5% reduction in body weight during the 9-year follow-up. We did not envisage a 5% reduction in body weight in the original study protocol, but we added it as a minor amendment because this is a widely recommended target for body weight reduction. In the first analysis, we estimated the annual probability of attaining normal body weight. We used the number of events (BMI category recorded as < 25 kg/m) and the person-years of follow-up to estimate the annual rates and their confidence intervals (CIs), which we converted to annual probabilities using the formula 1 – exp.
Among participants who reduced their BMI category, we evaluated the direction of the next change in BMI category. In the second analysis, to examine the proportion of participants who lost 5% of their body weight, we also evaluated the development of a body weight that was more than 95% of the initial body weight in a time-to-event framework. We conducted analyses in Stata version 13 using the stset, sts list, and stcox commands.
Source...