Endocrinology · Medical Literature
[Correspondence] How definitions shape obesity-attributable death estimates – Authors' reply
Abstract
Ayoub Boulares and Nicola Luigi Bragazzi raise four important methodological considerations regarding the association between adult obesity and severe infection risk reported in our study.1 Given the potential public health implications of our findings, we welcome the opportunity to clarify and provide additional context.
Clinical relevance
Is adult obesity an independent risk factor for severe infectious disease outcomes (hospitalization and infection-related mortality) across a broad spectrum of bacterial, viral, parasitic, and fungal pathogens, and what proportion of global infection-related deaths is attributable to obesity? __SNAPSHOT__ {"studyDesign":"Multicohort prospective observational study with population attributable fraction (PAF) impact modelling using Global Burden of Disease (GBD) data","setting":"Finland (two occupational cohorts: Finnish Public Sector study and Finnish Health and Social Support study) and the United Kingdom (UK Biobank); national hospital and death registries used for outcome ascertainment","sampleSize":"n ≈ 547,264 total (Finnish cohorts: n = 67,766; UK Biobank: n = 479,498)","population":"Adults with no history of severe infection-related hospitalization at baseline; Finnish cohorts mean age approximately 42 years, 73% women; UK Biobank mean age approximately 57 years, 54% women; participants categorized by BMI at baseline into healthy weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and obesity class I (30.0–34.9), class II (35.0–39.9), and class III (≥40.0 kg/m²)","intervention":"Exposure: baseline BMI category, with obesity (BMI ≥30 kg/m²) as the primary exposure of interest; additional adiposity measures (waist circumference, waist-to-height ratio) examined as sensitivity analyses","comparator":"Healthy weight (BMI 18.5–24.9 kg/m²)","primaryOutcome":"Severe infectious disease, defined as infection-related hospitalization or infection-related death, across 925 bacterial, viral, parasitic, and fungal infectious disease diagnoses; secondary: population attributable fraction (PAF) of global infection-related deaths attributable to obesity","followUp":"Mean approximately 13–14 years (Finnish cohorts followed through approximately 2022; UK Biobank through approximately 2022)","fundingSource":"Government/non-profit: Wellcome Trust, Medical Research Council (UK), and Research Council of Finland; no industry funding declared for the primary study. One co-author (Naveed Sattar) disclosed consulting fees and research grants from multiple pharmaceutical companies unrelated to this work."}
Key findings
Across both Finnish cohorts and UK Biobank, individuals with obesity (BMI ≥30 kg/m²) had approximately 70% higher risk of severe infectious disease (hospitalization or death) compared with healthy-weight individuals (relative effect size reported as approximately HR ~1.7; exact pooled HR with 95% CI not fully reproducible from the abstract alone — specific figures derived from press release and secondary reporting). Risk scaled with obesity severity: class III obesity (BMI ≥40 kg/m²) was associated with approximately three times the risk of severe infection relative to healthy weight (HR ~3.0; exact CI not reported in the abstract). Overweight (BMI 25.0–29.9 kg/m²) was associated with only modestly elevated risk. The association was consistent across obesity measures (BMI, waist circumference, waist-to-height ratio), demographic and clinical subgroups, and a wide range of infection types — including non-fatal and fatal, acute and chronic, bacterial and viral (including subtypes), and parasitic and fungal infections. Notable exception: no significant association was found between obesity and HIV or tuberculosis. Cox proportional hazards models were used; specific HRs with 95% CIs for the primary composite endpoint are not fully reported in the available abstract. Applying pooled HRs to GBD mortality data using Monte Carlo simulation (1,000 iterations), PAFs for obesity-attributable infection-related deaths were: 8.6% (95% CI 6.6–11.1%) in 2018 (pre-COVID-19), 15.0% (95% CI 12.8–17.4%) in 2021 (during COVID-19 pandemic), and 10.8% (95% CI 8.6–13.6%) in 2023 (post-pandemic). This corresponds to approximately 0.6 million of 5.4 million global infectious disease deaths in 2023 being linked to obesity. Country-level variation was substantial: PAF was approximately 25% in the United States and approximately 17% in the United Kingdom, versus as low as 1.2% in Vietnam. NNT/NNH: not calculable from available data. No major drug-related adverse events applicable (observational exposure study). NOTE: The document submitted for summarization is a Correspondence reply (August 2026), not the original study; all quantitative findings above are drawn from the original Nyberg et al. Lancet 2026 article (DOI: 10.1016/S0140-6736(25)02474-2), as only the correspondence reply — which itself contains no data — was submitted.
Practice implications
This large multicohort study establishes that adult obesity is a consistent, dose-dependent, and broadly applicable risk factor for severe infectious disease across diverse pathogen classes, populations, and clinical subgroups — extending well beyond the established obesity–COVID-19 and obesity–influenza associations. The finding that obesity class III carries roughly triple the severe infection risk of healthy weight is clinically important and adds to the rationale for considering obesity when stratifying infection risk and guiding vaccination prioritization. The PAF estimates (~10% of global infection deaths attributable to obesity in 2023) are derived from observational associations assumed to be causal — an assumption that cannot be verified in this study design. The authors themselves note that causality cannot be confirmed. PAF estimates will be sensitive to the BMI thresholds used to define obesity, which is precisely the methodological concern raised in the Boulares and Bragazzi correspondence that prompted the authors' reply (the content of which is not available for review). Findings apply most directly to predominantly White, working-age to older-adult populations in high-income Northern European settings; generalizability to lower-income countries, South/East Asian populations (where different BMI thresholds for obesity risk are used), or populations with very high infectious disease burden and low obesity prevalence (e.g., sub-Saharan Africa) is uncertain. The absence of an obesity–HIV and obesity–tuberculosis association is notable and biologically plausible, given the 'obesity paradox' documented in HIV and the inverse BMI–tuberculosis relationship. Practice implications include: (1) clinical and public health consideration of obesity as an infectious disease risk factor beyond cardiometabolic disease; (2) incorporation of obesity status into vaccination guideline discussions; (3) counseling patients with obesity about elevated infection risk and ensuring vaccine schedules are current. However, randomized evidence that weight reduction reduces infection-related hospitalization or mortality is limited; this study cannot substitute for such trials. CRITICAL LIMITATION NOTE: Only the abstract of a Correspondence reply — not the full original article — was submitted. All effect sizes reported here are sourced from publicly available secondary reporting and the original abstract on PubMed and The Lancet; specific within-subgroup HRs and full multivariable model outputs are not reproduced here and should be reviewed in the full primary publication. __LIMITATIONS__ Observational design precludes causal inference; residual confounding by socioeconomic status, comorbidities, immunosuppressive therapies, and health behaviors cannot be excluded.,BMI measured at a single baseline time point (1998–2002 for Finnish cohorts; 2006–10 for UK Biobank), introducing potential for misclassification due to weight change over long follow-up periods averaging 13–14 years.,PAF calculations assume that all observed associations are causal and that the GBD obesity prevalence data are accurately matched to the study populations — assumptions that are explicitly flagged as uncertain by the authors.,PAF estimates are highly sensitive to the BMI threshold chosen to define obesity (the core concern raised in the Boulares/Bragazzi correspondence), as varying cut-offs will produce materially different attributable fraction estimates, particularly for global projections applied to diverse ethnic populations.,UK Biobank is known to be a volunteer cohort with lower response rates and a healthier-than-average baseline profile compared to the general UK population, potentially underestimating true associations (healthy volunteer bias).,Generalizability is limited: study cohorts are predominantly White adults in Finland and the UK; WHO BMI thresholds used may underestimate obesity risk in South and East Asian populations, where lower BMI cut-offs are recommended.,No association between obesity and HIV or tuberculosis was found, but the study may have been underpowered for these outcomes in the specific cohorts studied, or the inverse association (weight loss with active disease) may have introduced reverse causality.,The Correspondence reply (the submitted document) addresses four unspecified methodological challenges that cannot be fully assessed without access to the full reply text, limiting the ability to determine whether the authors' responses adequately addressed concerns about definition-sensitive PAF estimates.,Surveillance definitions of 'severe infection' (hospitalization or death) may undercount milder but clinically significant infections, and registry-based ascertainment of 925 infection categories introduces heterogeneity in diagnostic coding quality across time periods and countries. __BOTTOM_LINE_CATEGORY__ potentially_practice_informing __BOTTOM_LINE__ This large multicohort study of over 540,000 adults provides robust evidence that adult obesity is a dose-dependent risk factor for severe infection across a broad spectrum of pathogens, with approximately 70% higher risk at BMI ≥30 and roughly three-fold risk at BMI ≥40 compared to healthy weight. Approximately one in ten infection-related deaths globally in 2023 was modelled as attributable to obesity, though this PAF estimate assumes causality and is sensitive to obesity definition thresholds — the methodological concern addressed in the companion Correspondence reply. Clinicians should consider obesity as a formal infectious disease risk factor when counseling patients, updating vaccination plans, and advocating for weight management, while acknowledging that direct trial evidence linking weight reduction to reduced infection mortality remains limited.