Discussion

In this study, 92% of obese patients attended to in primary care had associated comorbidity. The probability of weight loss among patients with hypertension, anxiety and/or substance use problems was 8%, 13% and 17% lower, respectively, compared to patients not presenting with such comorbidities. In contrast, obese patients with dementia and/or diabetes seemed to be more likely to lose weight from one year to another, especially if the latter had been diagnosed in the previous year. In addition, male sex, higher age, more severe obesity and longer time-length between weight measurements were all independently associated with weight loss according to the information recorded in the electronic medical records.

Primary care databases offer great potential for capturing the constellation of morbidities affecting people in the community.17 Moreover, such data are more informative of the real-life evolution of patients’ health characteristics (such as their weight) compared to strictly designed clinical trials, offering broader opportunities to move towards patient-centred comparative effectiveness research.18 Nevertheless, the accuracy and completeness of diagnostic coding in primary care electronic health records rely on general practitioners’ (GPs) efforts and capacity that are often limited, especially regarding BMI registration.19 We could not address waist circumference, which is an important complementary parameter to assess metabolic health, given that it discriminates body fat from fat-free mass. It was not possible to include drug use data to shed light on the role of certain medications such as metformin on weight loss. Although the diagnostic information covered all prevalent and incident chronic conditions during 2010, we did not consider those new diagnoses occurring during 2011.

Obesity rarely appears as an isolated condition. However, the burden of multimorbidity in obese patients varies among studies. In a primary care-based study of adults over 30 years in the UK, between 69% and 79% of obese patients showed at least one additional comorbidity, depending on the type of obesity and sex.6 In our study, the percentage of obese patients suffering from one or more chronic comorbidities rose to 92%, probably due to differences in the number of baseline chronic diseases considered (114 diseases in our study vs 11 diseases in the UK study). Similar results were reported in a Canadian study of the general adult population of Alberta,4 which was based on a list of 14 baseline chronic conditions. In this study, a significantly higher mean number of chronic disease was reported in the obese population (1.9 diseases) compared to the non-obese population (1.0 diseases); these numbers were higher in our study (3.3 diseases and 1.8 diseases, respectively). Two recent survey-based studies5 ,20 also showed a significant stepwise increase in the prevalence of the majority of chronic conditions, especially of cardiometabolic risk factors, from the lowest to the highest BMI category, although crude prevalences were lower than those obtained in our study. These differences may be related to the existence, in our study, of the so-called BMI recording and/or measurement bias in favour of those obese patients with a greater morbidity burden who see their GP more often.11 ,21 Indeed, in a sensitivity analysis carried out comparing obese patients with and without BMI documentation (see online supplementary table S1), we found that the former were much older, suffered from a higher mean number of chronic comorbidities and had a higher prevalence for most of the studied physical conditions.

A loss of weight between both study years was recorded for only half of the obese patients studied. This finding could be in accordance with the debate regarding the inadequate recognition, documentation and management of obesity by GPs.10 GPs have a clear opportunity to early diagnose and treat this health problem in most segments of the population, and their expertise is highly regarded by patients.22 Indeed, there is evidence indicating that patients receiving adequate weight reduction advice in primary care are more likely to reduce weight than those not receiving any counselling.23

In our study, older male patients with a higher level of obesity recently diagnosed from diabetes were more frequently reported to reduce weight from one year to another. In a German study about the management of obesity in primary care,19 the presence of vascular complications associated with hypertension or diabetes seemed to slightly increase the likelihood of using weight reduction interventions. In a similar US study,9 it was also stated that management of obesity occurred more frequently among patients with weight-related comorbidities such as diabetes, lipid disorders, hypertension or cardiovascular disease. The impact of weight loss regarding remission of diabetic symptoms in obese patients has been widely discussed,24 and seemed to be a determining factor in our setting, especially if the diagnosis of diabetes was made in the previous year. However, that was not the case for the other cardiovascular comorbidities considered in our analyses. In fact, suffering from hypertension was negatively associated with reduction of weight, after adjusting for potential confounders. This finding requires further investigation. We hypothesised that the association between obesity and hypertension might not be as deep-rooted as thought, neither among physicians nor among patients. GPs may focus more on the pharmacological treatment of this risk factor, while patients could have a lower feeling of ill health compared to those suffering from a well-defined disease such as diabetes. Moreover, in patients with hypertension, nutrition counselling may be based on hyposodic rather than hypocaloric diets.

Dementia is a well-recognised source of weight loss. This may in part be explained by reduced food intake because of impaired autonomy, eating disturbances and reduced appetite.25 In addition, neurodegenerative processes during the presymptomatic stage of the disease may be responsible for involuntary weight loss.26 Involuntary weight loss is also associated with cholinesterase inhibitors in dementia.27 There is evidence for a higher risk for mental disorders, such as anxiety, in obese individuals.28 Comorbid anxiety has also been shown to negatively influence the course of weight in treated and untreated patients.29 Moreover, substance use (ie, drug and alcohol) problems may be associated with the development of other detrimental lifestyles, such as an unhealthy diet and lack of physical activity, which may increase the risk of obesity, especially during youth.30 This could also be the reason for the poorer weight reduction outcomes among substance users observed in our study.

Presenting with multiple comorbid conditions has been shown to be positively associated with receiving advice to lose weight,31–33 although only about half of these patients actually received such advice from their physician.34 In our study, when considering cardiovascular and mental comorbidities of obesity, the total number of comorbidities itself was no longer significantly associated with weight loss. This was probably because we had already included a wide range of chronic conditions in the regression model, neutralising the explanatory power of the synthetic variable, that is, number of total comorbidities.

According to Bramlage et al,19 younger overweight and/or obese patients up to grade I were recognised and diagnosed considerably less frequently than older patients and/or those with more severe obesity, which may explain the lower weight loss rates documented for these population groups in this study. The association of older age with greater weight loss may be related to muscle loss rather than fat loss, especially in the very elderly. Moreover, the larger weight loss among individuals with higher grades of obesity could also be explained by regression to the mean. According to the study aforementioned, although women had a higher probability to be diagnosed as overweight or obese than men, the latter were more likely to use weight reduction measures,19 which may also elucidate our findings. This could be explained by women's role as informal carers of family members, which may be at the expense of their own health attitudes.