Introduction

Functional disability refers to a reduced ability to carry out activities of daily living (ADLs), which can lead to significant adverse health outcomes. As of 2021, an estimated 1.3 billion individuals, representing 16% of the global population, were living with a significant disability [1]. Although the increase in global life expectancy has contributed to this high proportion, the disease burden has remained constant from 1990 to 2019 [2]. While obesity or underweight status being key risk factors for disability in older people, reports on desirable body mass index (BMI) ranges have been inconsistent [3,4,5,6].

ADLs are commonly used to quantify the level of functional impairment. They are assessed using several standardized tools, most commonly the Katz Index and the Barthel Index. The Katz Index evaluates six basic functions, while the Barthel Index assesses ten self-care and mobility tasks. There are other assessments, including Japan’s Long-Term Care Insurance (LTCI) certification for need of care, the Groningen Activity Restriction Scale (GARS), the Korean Activities of Daily Living (K-ADL), and Resident Assessment Instrument-Minimum Data Set (RAI-MDS).

Existing studies evaluating the association between BMI and ADL impairment have also yielded inconsistent results. Studies have identified a positive association [7], an inverse association [8], and no association [9] between underweight BMI and ADL. Similarly, for overweight BMI, both a direct association [10], a negative association [11], and no association [12] with ADL have been reported. Despite these varying findings, no systematic review or meta-analysis including both longitudinal and cross-sectional studies in adults (aged 18 years and above) has yet been conducted. Such a study would yield robust results, enhance statistical power, and uncover relationships that might not be evident in individual studies. Furthermore, the findings may substantiate the need to shift health promotion strategies from their prior focus on preventing obesity in middle age, towards a comprehensive approach that emphasizes maintaining a healthy BMI throughout one’s life.

In the present report, we conducted a systematic review and meta-analysis to determine the association between BMI and ADL impairment among adults, thereby identifying the optimal BMI range for maximizing disability-free lifespan in adulthood.

Methods

Registration and search strategy

We carried out this systematic review and meta-analysis using standardized methods and adhered to the reporting requirements of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [13]. The protocol was pre-registered before the systematic review commenced in the PROSPERO (PROSPERO ID: CRD42022357046).

The search strategy was developed and executed by MRM through discussion with the content experts [HY, KMSUR, and RO] and review teams. Keywords and medical subject heading (Mesh) terms were combined using Boolean operators. The comprehensive search strategy developed for PubMed has been provided in Supplementary Text S1. We conducted search in PubMed, Web of Science (core collection), Cochrane Library, Scopus, Embase, and ICHUSHI (Japanese database). The search was initially conducted from inception to September 27, 2022, and updated on August 12, 2024.

Eligibility criteria

Studies were selected in compliance with the following inclusion criteria:

Any article describing the association between BMI and functional disability in relation to the ADLs was included. There was no time limit for the study period. Articles published in English, Japanese and Chinese were considered only. No geographic restriction was applied. Randomized controlled trials, quasi-experimental trial studies, case-control studies, cohort studies, cross-sectional studies, and comparative cross-sectional studies were considered. Articles only published in peer-reviewed journals were considered.

We excluded articles for the following exclusion criteria:

  • Articles not describing the association between BMI and functional disability in relation to the ADL.

  • Study participants were below 18 years.

  • Articles published in languages other than English, Japanese or Chinese.

  • Qualitative study, case report, review, systematic review, scoping review, editorial, commentary, book chapter, view paper/opinion paper, letter to editor, and conference abstract were excluded.

Data collection and analysis

Software

We used Rayan software [14] for the deduplication and screening of title/abstract and full texts. We used Rayyan solely for the screening process. No AI-assisted screening or automated tools within Rayyan were employed for either title/abstract or full-text screening. All screening decisions were made by the reviewers. Meta-analysis was performed using metaanalysisonline, a web-based meta-analytic platform.

Screening

The screening of citations was conducted in two phases. After removing duplicates, two review authors [GN and TA] independently screened the titles and abstracts of the citations to determine eligibility. Any disagreements were resolved in consultation with a third review author [HY]. Two independent review authors [MRM and ZS] screened full-text articles, and any discrepancies were resolved through discussion with a third review author [HY]. Articles obtained from the ICHUSHI (Japanese database) were independently screened by two Japanese review authors [YH and YY], and any disagreements were resolved by a third review author [YN] during both the title/abstract and full-text screening phases. No formal inter-rater reliability statistics were calculated, as consensus resolution was used to finalize inclusion decisions.

Data extraction

Two review authors [GN and TA] independently conducted data extraction using a pre-defined template. Any disagreements were discussed and subsequently resolved in consultation with a third reviewer [HY]. The extracted data included publication year, author name, types of study, sample size, BMI category, ADL assessment methods, and outcomes.

Risk of bias assessment

No randomized controlled trials were included in this review. The risk of bias of the included studies was assessed using the RoBANS tool [15]. The RoBANS tool consists of six domains: selection of participants, confounding variables, measurement of exposure, blinding of outcome assessments, incomplete outcome data, and selective outcome reporting. Each domain was classified as low, high, or unclear risk. Two review authors (MRM and ZS) independently assessed the quality of the included studies, and any disagreements were resolved by a third reviewer (KMSUR). Risk of bias assessment of included Japanese studies was performed independently by two Japanese review authors (YH and YY) and resolved discrepancies through discussion with a third review author (YN).

Data synthesis

We performed meta-analysis on the effect of BMI on ADL impairment. Meta-analysis was performed using metaanalysisonline, a web-based meta-analytic platform based on inverse-variance weighting of log-transformed effect sizes. We performed a random-effects model meta-analysis considering the heterogeneity. Between-study variance (τ²) was estimated using the platform’s default estimator (DerSimonian–Laird, DL). Results were presented as pooled OR/HR with 95% confidence intervals. Statistical heterogeneity was assessed using the Q-test and I2 statistic. Based on the risk of bias assessment, studies rated as high risk were excluded from all meta-analyses. In the meta-analysis, the normal BMI category (BMI 18.5–24.9) served as the reference group, and comparisons with other BMI categories were analyzed separately. We conducted separate meta-analyses for BMI categories indicating underweight, overweight and obesity. Specifically, in studies that used multiple exposure categories (e.g., underweight, overweight, and obesity), we extracted and included all relevant estimates in the meta-analysis. For example, we extracted the ORs for ADL impairment comparing underweight (BMI < 18.5) overweight (BMI 25–29.9), and obesity (BMI ≥ 30) with normal weight, respectively, and incorporated each estimate into the separate analyses.

Due to heterogeneity in ADL assessments, the meta-analyses were conducted stratified by the ADL assessment methods. For studies using the Katz Index, we conducted separate meta-analyses for different BMI categories where BMI cut-offs were comparable. Studies using various ADL assessments, including the Katz Index, Barthel Index, and others, were combined in separate meta-analyses for different BMI categories under the approach described as ‘irrespective of ADL measurement method,’ where BMI cut-offs were comparable. Separate meta-analyses were performed for separate effect estimates. Subgroup analyses were conducted with the following subgroups: study regions (Europe, North America, Asia, Australia, Africa); and measurement method of ADL. A narrative synthesis was performed for the rest of the included studies. We also conducted subgroup analyses by study design.

Results

Literature search

The study selection process is presented in the PRISMA 2020 flow diagram (Fig. 1). We initially identified 25,212 articles, and after deduplication, title and abstract screening, and full-text screening, 132 studies were included in this systematic review.

Fig. 1: PRISMA flow diagram of study selection.
figure 1

Flow diagram illustrating the identification, screening, eligibility assessment, and inclusion of studies.

Full size image

Characteristics of studies (study design, location, age group)

Details and characteristics of studies are described in Supplementary Table S1. Out of 132 studies, 77 were longitudinal studies and 55 were cross-sectional studies. Majority of the studies were from USA (n = 34) and Japan (n = 22). Our review also included studies were from Australia, Bangladesh, Brazil, China, England, Finland, France, Germany, Ghana, India, Indonesia, Ireland, Italy, Mexico, Netherlands, New Zealand, Poland, South Africa, South Korea, Spain, Switzerland, Taiwan, Thailand, and Turkey. Heterogeneous age groups were considered for inclusion in included studies ranges from ≥18 years to ≥90 years (Supplementary Table S1). The majority of the studies included participants who were 65 years of age and above.

BMI categories used in included studies

The BMI cut-off values differed greatly across studies, with a range from <18.5 kg/m2 to ≥40 kg/m2. Our meta-analysis was conducted based on the BMI categories reported in each study. Additionally, the reference group used in the included studies varied substantially. The presence of multiple BMI cut-off values and reference groups caused considerable challenges in conducting the meta-analysis.

Assessment of ADL in included studies

In the included studies, various methods for assessing ADL impairment were identified. Katz Index was used in thirty-three studies. Five studies used modified Katz Index. Eighteen studies used Barthel index. Six studies used LTCI. Namely, LTCI certification of need of care was defined as LTCI-ADL impairment in these studies. Forty-six studies did not specify scale or checklist name used in assessing ADL, but they did mention the questions they asked instead. Other studies used multiple scales and checklists including the GARS, K-ADL, RAI-MDS-ADL long-form scale, Stanford Health Assessment Questionnaire, Modified Stanford Health Assessment Questionnaire, Brazilian Multidimensional Functional Assessment Questionnaire, Brazilian Portuguese version of the Kihon Checklist (KCL), SAGE, Physical self-maintenance scale (PSMS).

Risk of bias in included studies

We assessed the risk of bias in all included studies, with detailed findings presented in Supplementary Table S2. Most studies exhibited a low risk of bias in participant selection, except for ten studies which had a high risk due to small sample size. In the domain of confounding variables, seven studies were assessed as having a high risk of bias, while the majority were considered low risk. For the measurement of exposure, four studies were rated as high risk due to reliance on self-reported data, and seven studies had unclear risk due to insufficient reporting. The blinding of outcome assessment, incomplete outcome data, and selective outcome reporting domains were assessed as a low risk of bias for all studies.

Obese group (BMI ≥ 30) and ADL impairment

Figure 2A shows the pooled OR for the association between the obese group (BMI ≥ 30) and ADL impairment measured by Katz Index. Combined results through the random-effects model revealed that obesity (BMI ≥ 30) had a higher risk of ADL impairment (OR: 1.77, 95% CI: 1.54–2.02). Cross-sectional studies showed a pooled OR of 1.61 (95% CI: 1.23–2.10), while longitudinal studies showed a pooled OR of 1.85 (95% CI: 1.58–2.16). Figure 2B shows pooled OR for the association between the obese group (BMI ≥ 30) and ADL impairment (irrespective of ADL measurement methods). The random-effects meta-analysis showed that obesity (BMI ≥ 30) was associated with a higher risk of ADL impairment (OR: 1.67, 95% CI: 1.52–1.84). Cross-sectional studies showed a pooled OR of 1.56 (95% CI: 1.38–1.76), while longitudinal studies showed a pooled OR of 1.77 (95% CI: 1.54–2.03).

Fig. 2: Association between the obese group (BMI ≥ 30) and ADL impairment, in comparison to the normal-weight group.
figure 2

A Individual and pooled odds ratios for the association between the obese group and ADL impairment measured by Katz Index. B Individual and pooled odds ratios for the association between the obese group and ADL impairment (irrespective of ADL measurement methods).

Full size image

Overweight group (BMI 25–29.9 and BMI ≥ 25) and ADL impairment

Figure 3A shows the pooled OR for the association between the overweight group (BMI 25–29.9) and ADL impairment measured by Katz Index. Combined results through the random-effects model revealed that the overweight group had a higher risk of ADL impairment (OR: 1.19, 95% CI: 1.10–1.29). Cross-sectional studies showed a pooled OR of 1.06 (95% CI: 0.95–1.18), while longitudinal studies demonstrated a higher pooled OR of 1.26 (95% CI: 1.15–1.39). Figure 3B shows pooled OR for the association the overweight group (BMI 25–29.9) and ADL impairment (irrespective of ADL measurement methods). Pooled analysis shows increased risk of ADL impairment (OR: 1.14, 95% CI: 1.07–1.22). Pooled analysis in cross-sectional studies shows OR: 1.06, 95% CI: 0.95–1.18, while longitudinal studies demonstrated a higher pooled OR of 1.19 (95% CI: 1.09–1.30).

Fig. 3: Association between the overweight group (BMI 25–29.9) and ADL impairment, in comparison to the normal-weight group.
figure 3

A Individual and pooled odds ratios for the association between the overweight group and ADL impairment measured by Katz Index. B Individual and pooled odds ratios for the association between the overweight group and ADL impairment (irrespective of ADL measurement methods).

Full size image

Underweight group (BMI < 18.5) and ADL impairment

Figure 4A shows the pooled OR of the association between the underweight group (BMI < 18.5) and ADL impairment measured by the Katz Index, with a pooled OR of 1.09 (95% CI: 0.89–1.34). Cross-sectional studies showed a pooled OR of 1.31 (95% CI: 0.99–1.73), while longitudinal studies demonstrated a pooled OR of 1.02 (95% CI: 0.80–1.31). Figure 4B shows pooled OR for the association between the underweight group (BMI < 18.5) and ADL impairment (irrespective of ADL measurement methods). Pooled analysis shows increased risk of ADL impairment (OR: 1.33, 95% CI: 1.19–1.49). Pooled analysis in cross-sectional studies shows OR: 1.46, 95% CI: 1.24–1.72, while longitudinal studies demonstrated pooled OR of 1.20 (95% CI: 1.02–1.42). Figure 4C, D show pooled HR for the association between BMI < 18.5 and ADL impairment in longitudinal studies, measured by LTCI (HR 1.29, 95% CI: 1.00–1.66) and irrespective of measurement method (HR 1.29, 95% CI: 1.14–1.47), respectively.

Fig. 4: Association between the underweight group (BMI < 18.5) and ADL impairment, in comparison to the normal-weight group.
figure 4

A Individual and pooled odds ratios for the association between the underweight group and ADL impairment measured by Katz Index. B Individual and pooled odds ratios for the association between the underweight group and ADL impairment (irrespective of ADL measurement methods). C Individual and pooled hazard ratios for the association between the underweight group and ADL impairment measured by LTCI in longitudinal studies. D Individual and pooled hazard ratios for the association between the underweight group and ADL impairment (irrespective of ADL measurement methods) in longitudinal studies.

Full size image

Obese group (BMI 30–<35 and BMI > 35) and ADL impairment

Pooled OR for the association between the BMI 30–<35 group and ADL (irrespective of ADL measurement methods) in longitudinal studies shows higher risk of ADL impairment (OR: 1.55, 95% CI: 1.28–1.87) (Fig. 5A). Pooled OR for the association between the BMI > 35 group and ADL impairment (irrespective of ADL measurement methods) in longitudinal studies shows heightened risk of ADL impairment (OR: 2.44, 95% CI: 2.03–2.94) (Fig. 5B). Figure 5C shows the pooled HR for the association between the BMI 30–<35 group and ADL impairment in longitudinal studies (HR: 1.31, 95% CI: 1.21–1.43).

Fig. 5: Association between high-BMI categories and ADL impairment, in comparison to the normal-weight group.
figure 5

A Individual and pooled odds ratios for the association between the BMI 30–<35 group and ADL impairment (irrespective of ADL measurement methods) in longitudinal studies. B Individual and pooled odds ratios for the association between the BMI > 35 group and ADL impairment (irrespective of ADL measurement methods) in longitudinal studies. C Individual and pooled hazard ratios for the association between the BMI 30–<35 group and ADL impairment in longitudinal studies. D Individual and pooled hazard ratios for the association between the BMI ≥ 25 group and ADL impairment measured by LTCI in longitudinal studies.

Full size image

Pooled hazard ratios for the association between the BMI ≥ 25 group and ADL impairment measured by LTCI in longitudinal studies shows HR: 1.54, 95% CI: 1.05–2.27 (Fig. 5D). Additionally, we conducted a pooled analysis for association between the different the BMI groups (BMI ≥ 30, BMI 25–29.9, and BMI < 18.5) on ADL (irrespective of scale) based on geographical region (Supplementary Fig. S1: A, B, and C, respectively). The number of studies in each region was low except North America region. In North America region, the obese group revealed a positive association between BMI and ADL impairment.

Narrative synthesis

A total of 80 studies were not included in the meta-analysis due to the heterogeneity in BMI reference category and outcome measurement. Moreover, definition of obese, overweight and underweight groups were heterogeneous. In this case we have considered obesity, overweight and underweight as mentioned in the respective study. Among the longitudinal studies, the results of 43 studies were not included in the meta- analysis, among these, 20 studies [8, 9, 16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34] identified an association between the obese group and ADL impairment, four studies [8, 28, 29, 34] reported an association between the overweight group and ADL impairment, and 10 studies [8, 20, 24, 28, 31, 32, 34,35,36,37] indicated an association between the underweight group and ADL impairment and four studies reported association between BMI (without categorization) and ADL impairment [28, 38,39,40] (Supplementary Table S1). Among the cross-sectional studies, the results of 37 studies were not included in the meta-analysis. Among these, 12 studies [41,42,43,44,45,46,47,48,49,50,51,52] reported an association between the obese group and ADL impairment, 10 studies [42, 44, 46, 48, 51, 53,54,55,56,57] found an association between the overweight group and ADL impairment, and eight studies [42, 49, 53, 56, 58,59,60,61] identified an association between the underweight group and ADL impairment and five studies [62,63,64,65,66] reported association between BMI (without categorization) and ADL impairment (Supplementary Table S1).

Discussion

The meta-analysis investigated the link between BMI categories and ADL impairment in both longitudinal and cross-sectional studies. The results show a positive association between obesity (BMI ≥ 30) and ADL impairment. Across different measurement methods, obese individuals consistently demonstrate higher OR for ADL limitations. Overweight individuals (BMI 25–29.9) also exhibit a moderate increase in ADL risk, particularly in longitudinal studies. However, the association weakens in cross-sectional studies. For the underweight group (BMI < 18.5), the Katz Index in both longitudinal and cross-sectional studies did not reveal an association. In contrast, an association was observed when assessed irrespective of ADL measurement methods. Interestingly, when using the LTCI for ADL measurement, the underweight group shows a higher risk. Cross-sectional analyses further support these trends, with positive associations observed irrespective of ADL measurement methods. In longitudinal studies, higher BMI categories (BMI ≥ 25, BMI 30–<35, BMI > 35) consistently demonstrate elevated risks of ADL impairment, suggesting a progressive association with increasing BMI. It is important to note that the analysis is complicated by the heterogeneity in BMI reference categories and outcome measurements, as well as the diverse definitions of obesity, overweight, and underweight groups. Due to these variations, a substantial number of studies could not be included in the meta-analysis. The complex interplay between BMI and ADL impairment may be influenced by various mechanisms. An elevated BMI often leads to increased mechanical stress on weight-bearing joints, such as the knees and hips. This can raise the risk of osteoarthritis [67], which may in turn reduce mobility and make it harder to perform ADLs. On the other hand, a low BMI is linked to muscle mass loss [68]. Having low muscle mass and strength has been found to be positively linked to dependence on ADLs [69].

The review authors did not find any systematic review explored the association between BMI and ADL impairment. In contrast, systematic review on the association between physical activity and ADL were conducted [70]. Another systematic review identified underweight and ADL as risk factors for sarcopenia [71]. In another systematic review, the authors found an association between BMI and frailty [72]. The authors reported that the underweight group had a higher risk of frailty (RR: 1.45, 95% CI: 1.10–1.90), while the overweight group had a slightly lower risk (RR: 0.93, 95% CI: 0.85–1.02), and the obese group had an increased risk (RR 1.40, 95% CI: 1.17–1.67).

In addition to our findings, other systematic reviews have examined BMI and disability, though with different focuses. One review assessed BMI and the risk of disability retirement, reporting a higher risk among underweight (HR/RR = 1.20, 95% CI: 1.02–1.41), overweight (HR/RR = 1.13, 95% CI: 1.07–1.19), and obese individuals (HR/RR = 1.52, 95% CI: 1.36–1.71) compared with normal-weight individuals, which aligns with our results [73].

Another review focused on BMI and ADL in older adults using longitudinal studies showed that underweight older adults were more likely to experience difficulties with BADL (OR = 1.33, 95% CI: 1.03–1.72). In contrast, overweight older adults had a lower likelihood of BADL difficulties (OR = 0.81, 95% CI: 0.79–0.83) [74]. The focus of these reviews differed from ours. Our review expands on previous work by including both cross-sectional and longitudinal studies and by considering all adults aged ≥18 years.

Implications

This systematic review highlights the importance of weight management and maintaining a healthy BMI to maintain independence in ADL. It also emphasizes the need for comprehensive strategies to prevent and address both obesity and underweight status across different age groups, in order to reduce the risk of ADL impairment.

The variability in how BMI categories and ADL measurement methods are defined underscores the need for standardization in future research. There is a particular need for focused efforts to standardize the assessment of ADL, by developing consistent criteria and ensuring that researchers adopt a uniform approach. By addressing these methodological variations, future studies can strengthen their findings and provide more accurate assessments of the relationship between BMI and ADL impairment.

Recent guidance from the 2025 Lancet Commission highlights important limitations of relying solely on BMI to classify obesity, noting that BMI can misclassify adiposity [73]. The Commission recommends supplementing BMI with additional anthropometric or body composition measures to more accurately identify excess adiposity and related functional limitations. This aligns with our observation that inconsistent BMI cut-offs across studies complicated the meta-analyses and underscores the limitations of relying solely on BMI when evaluating associations with ADL impairment. Future research would benefit from incorporating more comprehensive measures of body composition alongside BMI.

Strength and limitations

This review represents the first systematic review on this topic, which we conducted following standard methods and reported according to the PRISMA 2020 guidelines. The quality of the included articles was critically assessed using the RoBANs tool.

However, this systematic review also has limitations. Although a comprehensive search of electronic databases was conducted, we did not perform hand-searching of reference lists from included studies. Therefore, it is possible that some relevant studies may have been missed. A limitation of this review is that excluding individuals under 18 years of age narrowed its scope. Additionally, the review is limited by the lack of formal inter-rater reliability statistics (e.g., kappa coefficients) between reviewers during the screening stages. However, independent dual screening was performed, and disagreements were resolved in consultation with a third reviewer, in accordance with Cochrane guidance, to minimize selection bias. There were divergent cutoff points for categorizing BMI groups, which potentially posed a challenge to our analytical efforts. Similarly, there was considerable heterogeneity in the methods used to measure ADL. Different assessment tools and criteria were utilized, making it difficult to synthesize results and draw comprehensive conclusions. Furthermore, in the included studies, the definitions of obese, overweight, and underweight groups varied slightly. Due to the heterogeneity in BMI categories and ADL measurement methods, the majority of included articles were not considered for the meta-analysis. Some of the included studies exhibited a high risk of bias in participant selection, confounding variables, and exposure measurements. However, none of them were included in the meta-analysis. Some of the meta-analyses included a limited number of studies, and therefore the corresponding results should be interpreted with caution. A further limitation is that we did not GRADE the evidence.

Conclusion

Individuals with a higher BMI, as well as those who are underweight, exhibit an elevated risk of limitations in ADL. Our findings underline the importance of initiatives addressing both obesity and underweight status to mitigate the risk of ADL impairment.