Our results illustrate that normal weight in COPD is associated with increased HRQoL among DMP participants in GOLD grades 1–3, but obese participants in grade 4 have higher HRQoL than their normal weight peers. The findings are robust for several control variables including sociodemographic data and disease status. Moreover, we confirmed findings from other studies [
31,
32] that reduced BMI is associated with high emphysema rates. A recent review by Sun et al. [
33] indicates that reduced BMI is also associated with increased lung function decline. Consequently, patients with low or normal weight in GOLD 4 may be on a worse disease trajectory than their higher BMI counterparts who have slower lung function decline. Moreover, GOLD grade 4 resembles a special case as it is the last severity stage and patients are close to running out of reserves. Based on the observational nature of this study and the available variables, more advanced clinical research is needed to identify the actual causes of the shifted association between BMI and HRQoL in GOLD grade 4. Additional confounding and colliding also have to be accounted for.
Potential confounders
Regarding the impact of BMI on mortality, there is a rich debate on confounders in the context of the obesity paradox in COPD. For potential confounders in our study, we refer to this debate. Spelta et al. [
34] name several possible explanations for the obesity paradox: higher BMI may be associated with physiological changes that restrict but do not obstruct lung function and, hence, lead to worse GOLD classification for higher weight patients. In these cases, obesity and not COPD is the driver behind lower FEV1. Figure
1 illustrates the median BMI of the study population, stratified by GOLD severity. In our study, median BMI declined with severity grade, especially between GOLD grades 2 and 4. Mean BMI (not shown here) also declined with increasing GOLD grade. This indicates that there is no substantial correlation between higher weight and reduced lung function in our study population.
Spelta et al. 2018 also refer to evidence that higher BMI is associated with less hyperinflation. Owing to lack of respective data, it is not possible to evaluate the distribution of hyperinflation among our study population. Yet other studies show that, although hyperinflation increases mortality risk [
35], only static but not the more common dynamic hyperinflation seems to be associated with reduced HRQoL [
36].
Emphysema correlates with a loss of body fat as well as skeletal muscles [
37]. Emphysema is also associated with reduced HRQoL in COPD patients [
38] and is more prevalent among higher GOLD severity grades [
39]. Thus, emphysema may explain lower HRQoL and lower weight among COPD patients with GOLD grade 4. However, we controlled for emphysema by including the respective ICD-10 code, and no significant influence on the overall findings was observed.
BMI has been criticized for not indicating body composition or body fat percentage [
40]. Instruments that include waist adiposity, such as “A Body Shape Index” [
41], may therefore be better predictors of mortality [
42]. It was not possible to account for this issue, but it seems unlikely that patients with severe airway restriction (GOLD 4) have better body composition than their controls in lower severity grades.
Smoking is also a potential confounder and may lead to reverse causation [
43], as it reduces hunger/calorie intake and increases energy consumption. Like emphysema, smoking status was controlled for in this study, but no significant effect on results was observed.
Another potential confounder is increased surveillance and better care for obese patients. They may receive more intense care and may stay in hospital for longer periods of time [
8]. Thus, hospital days in the year before the questionnaire were controlled for, but the findings did not change.
Income is an indicator of socioeconomic status. Patients with lower income may also have lower HRQoL and higher BMI. So, income was accounted for.
Another issue, collider bias, is a form of selection bias, where exposure and outcome cause a third variable [
44]. Colliders should be removed from models, as they distort estimations. Some studies state that collider bias is responsible for and invalidates the obesity paradox [
45]. An analysis of why collider bias seems to be no valid explanation for our findings can be found in the Additional file
1 (see section on collider bias).
In sum, this study has been able to control for most of the important confounders of BMI impact discussed in the context of COPD, or at least hints have been found that potential bias of factors not controlled or controlled for (collider bias) might be low.
Previous studies have indicated an association between underweight and higher mortality in COPD [
46,
47]. This corresponds with our finding that underweight patients have more hospital days, more exacerbations, and decreased lung function. Another study evaluated data from a double-blind randomized controlled trial carried out in 1368 centers in 43 countries and concluded that BMI over 40, in patients with moderate COPD and increased cardiovascular risk, does not reduce mortality, whereas less severely obese patients (BMI 30–35) show a clear survival benefit compared with BMI 20–25 [
48]. This somewhat resembles our study finding where BMI from 35 to 40 was associated with higher HRQoL, whereas BMI scores > 40 were associated with HRQoL declines in GOLD 4 patients. A study by Landbo et al. [
5] also found that COPD grade had a significant influence on the association of BMI and mortality. For less severe COPD, normal weight appears to be optimal, but for severe COPD, overweight appears to be beneficial [
5]. A narrative review by Spelta et al. [
34] evaluated the nine most important studies on the obesity paradox and concluded that the “protective” effect of obesity is most evident for the severe grades of COPD but not for mild to medium airway restriction. Our study findings also point in this direction. The stratification by severity grade could explain why previous, unstratified evaluations [
49,
50] reported conflicting results on the association between obesity and HRQoL in COPD.
Clinical relevance
Minimal important differences (MIDs) define which changes in HRQoL are meaningful for patients. For example, the MID for EQ-5D-5L VAS in COPD was estimated to be around 6.9 in a study from 2016 [
26]. To achieve this MID, patients in GOLD stage 2 would need to lose around 14 BMI points. A reduction of this magnitude is difficult to achieve. Although intervention options can be ordered systematically from lifestyle changes to pharmacotherapy to surgery [
17], only the latter seems to be effective enough in achieving comparable long-term BMI reductions [
51] with acceptable cost-effectiveness [
52], when patients are not willing to rigorously change their lifestyle. Surgery has also been found to lower COPD impact [
53]. Only doctors and patients can make the respective decisions, and they should always account for possible side effects or unforeseen consequences. However, to avoid the need for surgical intervention or rigorous lifestyle changes, prevention of obesity would certainly be preferable. To improve prevention, COPD patients should be put under special surveillance either when they gain weight quickly or when they reach certain BMI thresholds. Another important but more critical aspect is the avoidance of reaching HRQoL decrements that are relevant to the patient. For example, in GOLD grade 2, patients above BMI 42 would need a HRQoL improvement > MID to reach the optimum HRQoL in their stratum. Reaching a BMI below 42 for patients above this threshold would take them out of the described risk group.
The advantage of obesity treatment, contrary to the treatment of other diseases, is the underlying principle. To lose weight, calorie input has to be smaller than calorie expenditure. Both factors can be directly influenced and lead to immediate improvement when accounted for. Many other diseases are based on more complex coherencies, where cause and effect have not been identified and available treatments are far less effective.
Limitations and strengths of this study
One limitation of this study is the lack of adjustment for lung hyperinflation and body composition. However, as stated in the discussion section, the influence of lung hyperinflation is unlikely to change study outcomes significantly, as the more prevalent dynamic form apparently has little influence on HRQoL. BMI fails to reflect body composition and does not allow detection of the increased mortality risk associated with sarcopenia. Supplementary measures such as hip–waist ratio are needed to acquire a more comprehensive picture of the risks associated with patient weight. As only DMP participants from AOK Bayern were included in this study and the response rate was 30%, selection bias is likely. However, on account of the relatively large number of observations, the findings regarding the association of BMI and HRQoL should still be valid for broader populations of COPD patients. Moreover, as DMP participation is associated with increased quality of care [
54], the gap in HRQoL could be even higher in a non-DMP setting. Based on study design, a more representative approach was not possible.
A clear strength of this study is the large number of observations and the linkage of claims and survey data that enabled the evaluation of HRQoL. In contrast to mortality, HRQoL allows the guiding of patient management along the course of chronic disease and represents no binary outcome. Other strengths of this study are the stratification by severity grade, the number of included control variables, and the sensitivity analysis including an unadjusted model. To our knowledge, this study is the first to evaluate the association between BMI, HRQoL, and COPD severity grade based on a large number of observations from claims as well as survey data. Although this study could not solve all the issues associated with the connection of BMI and HRQoL in COPD, it certainly provides much needed and differentiated evidence to improve patient management.