The association between marital status and the risk of cardiovascular, cancer, and all-cause mortality: An updated systematic review and meta-analysis

The study pool mixes incompatible populations — five disease-specific survival studies alongside seven general-population cohorts — producing pooled numbers that describe neither quantity. The included studies adjusted for vastly different sets of other factors: one of the largest (Burgoa 1998, over 327,000 participants) controlled only for age and sex, while others controlled for fifteen or more variables including income, education, smoking, and diet. Pooling these does not estimate any single coherent quantity.

The paper's own measure of between-study consistency exceeded 94% for nearly every analysis, reaching 99.9% for widowed overall mortality. The studies share essentially no common signal — the pooled number is arithmetic, not science.

Marital status was measured once at enrollment, then treated as fixed over follow-up periods of two to eighteen years. Many participants would have changed status during that time. In countries where cohabitation is common, administrative registries likely classified partnered people as single.

The internal contradictions — at least six text-versus-figure discrepancies and one confirmed duplication — mean the reported numbers cannot be taken at face value even setting aside the design problems.

The random-effects model is the right tool in principle, but the degree of heterogeneity makes pooling meaningless. When virtually all variation reflects real differences between studies rather than chance, a single summary number is not informative.

The subgroup analyses split an already small pool (seven to nine studies) into groups of two to four. One outlier can determine the entire result. No correction was applied for multiple comparisons across roughly thirty to forty tests, so some results reaching conventional thresholds are expected by chance.

The pervasive numerical contradictions make the effect sizes uninterpretable. For widowed heart disease mortality alone, the text reports 67% higher risk while the figure reports 37%. The publication bias assessment contradicts itself: the visual check suggested missing studies, but a formal test that lacked the sensitivity to detect bias with this few studies was treated as reassurance.

The data extraction section mentions "definition of Preeclampsia" — a condition unrelated to the topic — suggesting the methods template was copied from a different review.

The study pool does not measure one thing. The reported numbers contradict themselves. The design cannot distinguish between marriage causing better health and healthier people being more likely to marry.

On a claim-by-claim basis: the direction of higher mortality among unmarried people is consistent with prior literature, but the specific magnitudes are untrustworthy. The null finding for divorced people and cancer is the most reliable result. The sex-difference claim for divorced men versus women rests on duplicated data and was never actually reported. The regional comparisons rest on two to three studies per group, with the apparent US effect for widowed cancer mortality driven by disease-specific survival studies rather than general-population data. The policy recommendations require causal evidence that observational data cannot provide.

The entire analysis assumes that marriage causes lower mortality. The most parsimonious alternative — that healthier people are more likely to marry — is acknowledged in one sentence and set aside. The conclusion recommends public health interventions, which only makes sense under the causal interpretation.

The abstract presents precise, confident numbers. The body reveals those numbers sit atop heterogeneity so extreme that the included studies share no common signal. A reader of the abstract alone would form a confident, specific, wrong impression.

The sex-difference finding for divorced men — featured in the abstract and discussion as evidence that men benefit more from marriage — is a copy-paste error. The discussion treats it as confirmed and builds interpretation around it.

Health selection, misclassification from one-time measurement of marital status, and confounding by socioeconomic factors are each mentioned briefly and set aside rather than incorporated into the interpretation.

The protocol was pre-registered, the search strategy is published, and the authors disclose no conflicts or funding. But the extracted data and analysis code are unavailable, which matters enormously when the published numbers are self-contradictory. A reader who wanted to determine the actual results has no way to do so.

The limitations section acknowledges heterogeneity and language bias but omits the most consequential problems: the incoherent study pool, the contradictory numbers, the single-time-point measurement of marital status, and the inconsistent confounder adjustment across studies.

The paper claims "previous meta-analyses have generally found that married individuals have better survival outcomes than their unmarried counterparts, but none have specifically addressed sex differences." It then repeatedly cites Wang 2020 — a meta-analysis of 21 studies and 7.9 million people that specifically addressed sex differences in marital status and mortality. The paper's claimed novelty is contradicted by its own reference list.

The biological mechanism citations in the introduction are decorative. Six papers are cited to build an implied chain from unmarried status through cortisol to disease. None of those six study marital status — they connect cortisol to cardiovascular biomarkers, cancer survival, or metabolic syndrome without any link to relationship status.

Key cited papers contain inconvenient findings that go unmentioned. Rendall 2011, cited for marriage's protective effect, found little evidence of mortality differences between unmarried categories — directly challenging the paper's structure. Shor 2012, a much larger meta-analysis of widowhood and mortality, found smaller effects than this paper reports but is cited only in passing about Singapore's eldercare system. Goldman 1993, cited for the marriage selection hypothesis, argues that aggregate mortality inferences of the kind this paper draws "are simply not justified."

The paper also ignores well-established literature on cohabitation (which blurs the unmarried category), marriage quality (which can reverse the apparent benefit), and temporal trends in the marriage-mortality association.

There are no conflicts of interest. The authors are medical students at Iranian universities with no external funding and no industry ties. The paper's problems are methodological — consistent with inexperience — not motivational.