The Obesity Paradox and Mortality in Older Adults: A Systematic Review

1Obesity, usually defined by the body mass index (BMI), is considered a public health problem, and is associated with many diseases. The prevalence of obesity is high in younger adults but also in older people, and evidence suggests that prevalence of obesity will continue to increase. The term “obesity paradox” is used to describe the counterintuitive finding that aged overweight and obese people with a particular disease may have better outcomes than their normal weight or underweight counterparts. However, there is wide heterogeneity between studies regarding the association between obesity and mortality in older adults, depending on the diseases concerned, the presence or absence of a particular disease, or the BMI level considered. In aged people, body composition tends to change, and body weight tends to decrease, and some authors have suggested that fatness could be healthy. Thus, it is important to confirm whether an “obesity paradox” truly exists, with a view to adapting management policies for overweight or obese old people.

2In this context, the objective of the study was to summarize the publications in the literature relating to the obesity paradox in older adults, to enhance our understanding of this phenomenon.

3A preliminary check was made in PubMed©, Scopus©, Embase©, Prospero©, and the Cochrane Library© to ensure that no systematic reviews had previously been conducted on this specific topic.

4A literature search was performed using PubMed©, Embase©, and Scopus© to cover all publications up to March 20, 2022. The search terms defined by the two researchers (LG, MD) included the following keywords in the title and/or the abstract: (“obesity paradox” OR “reverse epidemiology” OR “body mass index” OR BMI OR overweight OR obesity) AND (mortality OR death OR survival)). The search included studies in the French or English language and studies on human subjects, and excluded the following publication types: reviews, editorials, correspondence, and case reports and case series. A manual check was performed for potential additional studies. This systematic review was based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. This study was registered with PROSPERO (an International prospective register of systematic reviews) (number CRD42021289015), available at https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021289015, accessed on 20 March 2023.

5Study eligibility criteria were defined a priori by the two researchers (LG, MD) within the PICOS framework. Studies were eligible if they reported data on “obesity paradox” (using body mass index as a nutritional indicator). The population was restricted to studies that included persons 65 years or older, whatever their sex, ethnicity, or living place. The intervention (exposure) was a presence of overweight or obesity as defined by the baseline BMI value. The control was those who were underweight or a normal weight. The outcome was death, whatever the timepoint. When the study was not specifically conducted in older adults, only data concerning those aged 65 years or over were taken into account (provided that the information was available). Correspondence, editorials, reviews, basic science articles, and case reports and case series were excluded.

6The Covidence systematic review software© (Veritas Health Innovation, Melbourne, Australia), available at www.covidence.org, was used to perform data analysis. After elimination of duplicates, the two researchers (LG, MD) made a blind review of titles and abstracts of all articles. When there was disagreement about whether or not to include an article, they discussed the case until consensus was reached. Overlap between studies was verified. Data extraction was realised independently by the two researchers (LG, MD), using the same extraction form. The following data were extracted: publication year, study setting, medical condition, study design, sample size, age (mean or median and their statistical dispersion parameters, when available), and outcome(s). To check whether the obesity paradox was present or not, the following information was collected: outcome(s), BMI classes, type of analysis (whether multivariable or not), statistical estimates (Hazard ratio, Odds ratio, Rate ratio, Rates) and their respective 95% confidence intervals (95% CI), and the level of significance (p-values).

7The Newcastle–Ottawa Scale (NOS) was used to assess the quality of included studies. This scale is composed of three quality criteria: selection (4 points), comparability (2 points), and outcome assessment (3 points). This gives a total of between 0 and 9 points. Scores of 7 or more are considered high quality studies, scores of 5–6 as moderate quality, and scores below 5 as low quality. Disagreements in scoring were resolved by a joint review of the manuscript to reach consensus.

8Where possible and appropriate, some parameters were calculated from available data (e.g., mean age and/or standard deviation, rate ratio, odds ratio, etc.).

9As shown in Figure 1, 2226 studies were identified by the literature search. Among these, 1285 duplicates were found and excluded. After checking titles and abstracts of the remaining 942 studies, 273 articles were included for full-text assessment. After full-text examination of these 273 studies, 215 were excluded for at least one of the following reasons: lack of relevant information, overlapping data, or inappropriate age of the study population. Thus, 58 studies were retained in this review.

10Table 1 summarizes the characteristics of the studies included in the review. All studies were observational cohorts; 41 were prospective and 17 were retrospective.

11As shown in Table 2, 20 of the 58 studies included in this review did not find any evidence of an obesity paradox. Of these 20 studies, 16 involved patients with no specific medical condition. One involved patients with chronic diseases, and two involved patients with type 2 diabetes mellitus. Of the 58 studies, 34 used the threshold of BMI ≥ 25.0 kg/m2. A further 10 studies used a threshold different from 25 kg/m2 and found evidence of the obesity paradox. Regarding the time points, 9 studies looked at short-term mortality (less than 12-month mortality, ICU mortality, hospital mortality). All of these, except Yamamoto et al. and Kananen et al., found evidence of the obesity paradox. Of the 28 studies that examined longer-term mortality (time point ≥ 5 years), 15 (54%) found evidence of the obesity paradox. In the studies that were conducted in people with a particular medical condition (n = 24), the obesity paradox appeared in 18 (75%) cases. In the studies that were carried out among people with no specific medical condition (n = 34), the obesity paradox appeared in 17 (50%) cases.

12An appendix provides detailed information of the analyses and results of the relationship between BMI and mortality in aged adults. Of the analyses tested for the existence of an obesity paradox, 48 were adjusted for confounders, and 10 were unadjusted analyses (see Supplementary Materials).

13The quality of the included studies, as assessed using the NOS, was considered high for all 58 studies (Table 3).

14In this systematic review of studies exploring the relationship between BMI and mortality in patients aged 65 years or older, 28 out of the 58 studies included observed longer survival in patients with a BMI ≥ 25 kg/m2 (the so-called obesity paradox). Among these 28 studies, 16 involved patients with a specific or acute medical condition. Seven studies found improved survival in overweight and obese older people when focussing on short-term mortality. One showed increased survival only in the oldest patients. Two showed increased survival only in men. Of the 23 studies that did not observe an obesity paradox, 7 involved populations selected according to the presence of a particular medical condition.

15Nearly two-thirds of the studies included in this work report better survival in overweight or obese older people. Several factors may influence the relationship between obesity and survival in the older population, including age, degree of obesity, presence or absence of comorbidities, and occurrence of an acute event.

16Regarding age, the studies in this review that failed to show better survival in overweight or obese individuals included populations that were, on average, younger than those demonstrating an obesity paradox. Wu et al., in their study of the impact of age on the association between BMI and all-cause mortality in patients with atrial fibrillation, found better survival in overweight or obese patients aged 75 years or older but not in patients aged between 65 and 74 years. Observations made in older populations must therefore take into account the intrinsic characteristics of the survivors. For the same BMI, patient profiles can be different, and this profile can influence survival. For instance, body composition may differ due to ethnicity, sex, or advancing age. BMI does not provide information on body composition, and is less correlated with percentage of body mass or fat mass index, especially in younger people. Abdominal obesity has direct metabolic consequences (adipose tissue inflammation, dysglycaemia, alteration of blood pressure regulation, etc.). Conversely, subcutaneous fat accumulation in the hips, for example, appears to have benign effects on cardiovascular risk. Other indicators, such as waist circumference or waist-to-hip ratio, are strongly associated with higher mortality risk. Taking only BMI into account does not make it possible to differentiate between these situations. In all studies included in this work, BMI was defined as an obesity index. If obesity is defined by “body adiposity”, BMI level is probably not the best criterion. The term “BMI paradox” may be more appropriate than “obesity paradox”, as suggested by Antonopoulos et al..

17Obesity is a factor associated with higher mortality in younger populations, but it is also associated with an increased risk of developing and dying from a number of diseases, such as cancer, Some authors point to the obesity-related cellular and immune changes that make obese people more vulnerable, including an increased risk of infections. Older obese people could be considered constitutionally more robust as they have survived the risk factor of obesity into adulthood. The degree of obesity could also be a factor. In this review, not all authors differentiated between different classes of obesity. However, the positive effect on survival in cases of overweight and obesity was not found for morbid obesity (BMI ≥ 35.0 kg/m²) in 5 studies. Furthermore, weight is not a reflection of body composition, in particular the muscle mass/fat mass ratio. Loss of muscle mass and strength (sarcopenia) is a factor associated with an increased risk of death. Tian et al. reported that obese people with sarcopenia have a higher risk of death than obese people without sarcopenia. Obese people may be less frequently sarcopenic than non-obese people. In 1493 subjects aged 65 years or more (median age 74 ± 11 years), Sousa-Santos et al. found a prevalence of 0.8% of obese sarcopenic individuals versus 11.6% of sarcopenic individuals of all BMI status.

18The presence of a chronic pathology or an acute event may also influence survival. In this review, 20 studies of the 38 which found a favourable effect of overweight or obesity on survival involved patients with a particular chronic condition or facing a specific medical event. This finding suggests that even moderately overweight older individuals with chronic disease or acute medical events have better survival. Obesity in older people with a chronic disease could be a sign of greater robustness or higher reserves (better appetite, less risk of undernutrition). Overweight or obese older subjects would be less undernourished than the general older population. Cereda et al., in their meta-analysis of the prevalence of undernutrition in an older population, found a prevalence of undernutrition ranging from 3.1 to 29.4%, depending on the setting. Sousa-Santos et al. showed that 6% of obese elderly subjects (BMI ≥ 30 kg/m2) were also undernourished or at risk of undernutrition. In the event of an acute event, obese elderly people may have a better chance of survival, particularly because of their greater functional reserves. This observation is also made in younger obese or overweight subjects. Akinnusi et al. show in their meta-analysis of patients admitted to intensive care that obese subjects have a similar mortality to non-obese subjects. In 2013, the meta-analysis by Flegal et al. confirmed in a population without any particular pathology that overweight people (BMI > 25 kg/m²) (all types of obesity and all ages) had a higher overall mortality rate, whatever the cause. However, mildly overweight people (BMI ≥ 25 and <30 kg/m²) had lower all-cause mortality than normal weight people (BMI < 25 kg/m²). Thus, this advantage was found regardless of age.

19Several mechanisms could explain “obesity paradox”. Probably, there are “good adipose tissues” in elderly subjects. In the literature, overweight or obesity, defined by high level of BMI, is shown to have positive influence on prothrombotic factors, production of certain cytokines, or NT-proBNP levels. Adipokine produced by adipose tissue seems to be cardioprotective. Obesity could have a protective effect against progression or consequences of some chronic diseases. High BMI could also reflect better nutritional status and adequate muscle reserves. Casas-Vara et al. showed better nutritional status in overweight or obese elderly people with heart failure.

20Our systematic review has limitations. Although the WHO has proposed thresholds for BMI, the authors used different thresholds in their respective studies. In addition, the outcomes were also different between the studies. This made it difficult to compare the studies, and precluded meta-analysis. The age variable was missing in 14.0% of cases (8/57).

21However, this work covers a large number of studies, totalling more than 1,120,000 people aged 65 years or over, with varying medical conditions and in different settings. The follow-up time of the studies ranged from 30 days to 156 months (even though the majority of studies have a long-term follow-up). These differences in follow-up time may make comparison difficult. In addition, there is no information on BMI variation over time, especially for studies with long-term follow-up. Weight loss or gain between baseline measurement and death could have a significant impact. The fact that only studies conducted in subjects aged 65 years or older were selected gives a certain homogeneity to this systematic review in terms of population. Finally, all studies were evaluated for methodological quality using the NOS, and were found to be of high quality.

22The findings of this systematic review are in favour of the existence of an obesity paradox, which could more specifically concern older subjects with a comorbidity and/or experiencing an acute event. Nevertheless, because BMI does not reflect body composition, the term “BMI paradox” would be more appropriate. The influence of the level of BMI remains unclear. These findings should help guide strategies for nutritional counselling in the older population.