Letter to the Editor about “Mortality from Alzheimer’s disease and other dementias and its heterogeneity across states in India: findings from GBD 2021 study”
Year: 2026
Type: letter
Abstract: Dear Editor, I read with great interest this study, which uses estimates from the Global Burden of Disease 2021 (GBD 2021) and supplementary analytical tools to examine long-term trends in dementia mortality and related risk factors in India[1]. While the topic is of substantial public health relevance, several methodological and interpretive issues warrant careful consideration before strong conclusions can be drawn. We also note that the article complies with the TITAN guidelines[2]. The analysis relies entirely on the GBD cause-of-death model, which itself is based on sparse and highly heterogeneous primary data from India. Many states have long faced incomplete vital registration, misclassification of causes of death, and irregularities in the issuance of death certificates. Yet the authors treat model-derived estimates of unexpected increases in mortality as if they were directly observed data, without adequately considering that the apparent rise could stem from model revisions, improvements in data inputs, or algorithmic reclassification rather than genuine epidemiological change. Without sensitivity analyses using alternative modeling assumptions or triangulation with local mortality data, the robustness of these reported trends remains uncertain. The bidirectional Mendelian randomization (MR) analysis raises substantial concerns about external validity. The authors exclusively employ GWAS-derived genetic instruments from populations of European ancestry to infer causal pathways for dementia in India, despite marked differences in allele frequencies, linkage disequilibrium patterns, and environmental interactions[3]. These discrepancies may lead to biased or non-transferable causal estimates. The study neither engages with these issues nor provides justification for why European instruments would be biologically or contextually appropriate proxies for Indian populations. At a minimum, sensitivity analyses or simulations assessing instrument portability are needed to support such causal claims. Patterns in the MR findings themselves should have prompted caution. For instance, in the forward MR, smoking appears to have a protective effect for certain dementia subtypes, an observation inconsistent with established clinical and biological evidence[4]. Rather than investigating potential pleiotropy, selection bias, or collider structures, the manuscript treats these results at face value. This undermines confidence in the causal interpretations. MR-PRESSO and Cochran’s Q alone cannot rule out population stratification or genetic confounding, particularly when instruments are applied across ancestral groups. The Bayesian APC model’s projections to 2050 assume that past trends will continue without structural breaks. Given the rapid changes occurring in India, rising dementia awareness, evolving diagnostic standards, shifts in healthcare access, and demographic transitions, such extrapolations may be highly unstable. The manuscript does not disclose the uncertainty associated with model structure, creating an impression of precision that is unlikely to be justified. The attribution of deaths to high fasting glucose, elevated BMI, and tobacco use relies on the GBD comparative risk assessment framework, which assumes globally derived relative risks. These assumptions may not hold in India, where metabolic phenotypes (e.g., the “thin-fat” Indian phenotype), dietary patterns, and life-course exposures differ substantially[5]. Presenting attributable fractions as definitive estimates without discussing potential mismatches between global and local risk profiles may lead to misleading interpretations. This paper addresses an important topic, yet several of its strongest claims rest on heavily modeled, indirect, and partly non-transferable evidence. To avoid overstating the findings, it is essential to clearly acknowledge data limitations, interpret MR results with greater caution, and rigorously address issues in the spatial and APC analyses. A revised version that incorporates these methodological considerations would offer a more balanced and scientifically reliable contribution.
Source: International Journal of Surgery
Authors: Hao Lu, Dianmei Yang, Yan Shi, Junwen Wang
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $3,450
OpenAlex ID: W7124141163