Scientists Reassess COVID-19 Genetic Risk Findings After Statistical Challenge

A scientific dispute over whether inherited blood-clotting and folate-metabolism gene variants influence COVID-19 outcomes has taken an unusually transparent turn. In a formal reply published in the open-access journal Immunity, Inflammation and Disease, a research team led by Hilal Çıralıoğlu has re-examined its own patient data after a reader, Dr. Panda, raised pointed methodological and statistical objections to the team’s earlier report linking Factor V Leiden and MTHFR polymorphisms to COVID-19 case status. The exchange offers a rare public window into how genetic association studies are audited, corrected, and reinterpreted, and it ends with a conclusion that is notably more cautious than the original paper.

The original study had compared 150 patients with COVID-19 against 300 healthy controls, genotyping two well-known variants: Factor V Leiden, the G1691A mutation in the F5 gene that causes activated protein C resistance and raises thrombosis risk, and the A1298C variant of the MTHFR gene, which encodes methylenetetrahydrofolate reductase, a key enzyme in folate and homocysteine metabolism. Both variants have long attracted attention in COVID-19 research because severe disease is strongly associated with abnormal coagulation, and the pandemic prompted a wave of candidate-gene association studies searching for inherited predictors of susceptibility and severity.

The first concern raised by the commenter involved Hardy–Weinberg equilibrium, a foundational quality check in population genetics. Under HWE, genotype frequencies in a randomly mating, stable population should follow predictable proportions determined by the allele frequencies themselves. When the control group of a case–control study deviates significantly from these expected proportions, it can signal population stratification, selection bias, sampling artifacts, or genotyping errors, any of which can distort the apparent association between a variant and disease status. Because the control group is meant to represent the baseline population, HWE testing in controls is considered an essential safeguard before any case–control comparison is interpreted.

When the authors re-ran the analysis, the results vindicated part of the critique. For Factor V Leiden, the control group showed 261 GG homozygotes, 39 GA heterozygotes, and no AA homozygotes, a distribution consistent with Hardy–Weinberg expectations: the Pearson chi-squared test yielded a statistic of 1.45 with a p value of 0.229, and an exact test was similarly non-significant at p = 0.623. For MTHFR A1298C, however, the picture was different. The controls comprised 184 AA and 116 AC individuals with no CC homozygotes, and this distribution departed sharply from equilibrium, with a chi-squared statistic of 17.23 (p = 3.31 × 10⁻⁵) and an exact test p value of 1.26 × 10⁻⁶. The authors acknowledged that this deviation should have been reported and discussed in the original publication and now flag it as an important limitation.

Importantly, the authors did not attempt to explain away the anomaly. They note that HWE deviations can arise from population structure, selection effects, sampling characteristics, or genotyping-related factors, and they concede that the available data do not allow them to determine which explanation applies. The original methods described PCR amplification followed by high-resolution agarose gel analysis, with genotype calls confirmed by DNA sequencing, but the published report did not document duplicate samples or negative controls, and the authors explicitly declined to retrospectively claim quality-control procedures that were never written down. This refusal to overstate what the record supports is itself a methodological lesson in scientific accountability.

The second set of concerns concerned effect-size reporting. P values alone, the commenter argued, convey little about the magnitude or precision of a genetic association; odds ratios with 95 percent confidence intervals and allele frequencies are far more informative. The authors agreed and recomputed the full set of estimates from the original individual-level data. For Factor V Leiden, the GA heterozygous genotype appeared in 34 of 150 patients (22.7 percent) versus 39 of 300 controls (13.0 percent). Relative to GG homozygotes, carrying one A allele was associated with higher odds of being in the COVID-19 group, with an odds ratio of 1.96 (95 percent CI 1.18–3.26; Fisher’s exact p = 0.010). At the allele level, the A-allele frequency was 11.3 percent in patients versus 6.5 percent in controls, giving an allele-based odds ratio of 1.84 (95 percent CI 1.13–2.98; p = 0.014).

The MTHFR analysis pointed in the opposite direction. The AC genotype was present in 35 of 150 patients (23.3 percent) but 116 of 300 controls (38.7 percent), so relative to AA homozygotes the heterozygous genotype was associated with lower odds of belonging to the COVID-19 group (OR = 0.48, 95 percent CI 0.31–0.75; p = 0.0014). The C-allele frequency was 11.7 percent in patients versus 19.3 percent in controls, an allele-based odds ratio of 0.55 (95 percent CI 0.37–0.83; p = 0.0034). On purely numerical grounds, the pattern suggests that the C allele might be protective. Yet the authors immediately qualify this: because the MTHFR genotype distribution in controls violates Hardy–Weinberg equilibrium, both the genotype-based and allele-based estimates must be treated with caution, and the team now classifies the MTHFR case–control finding as exploratory, requiring confirmation in an independently recruited population before any protective association can be claimed.

A further structural limitation emerged from the data themselves: no Factor V Leiden AA homozygotes and no MTHFR CC homozygotes were observed in either group. This absence is not surprising, since both variants are relatively uncommon, but it means that homozygote-specific or recessive genetic models could not be meaningfully evaluated at all. In genetic epidemiology, the inability to test a full range of genotypes constrains the biological interpretations available, because dose-dependent effects of an allele can only be assessed when all genotype classes are represented. The reanalysis table therefore reports heterozygotes against reference homozygotes, with odds ratios calculated from 2 × 2 contingency tables using the log-odds (Wald) method and two-sided Fisher’s exact tests for significance.

The authors are careful to delineate the scope of the damage. The Hardy–Weinberg deviation was detected in the healthy control group, which primarily undermines the between-group MTHFR comparison, that is, the claim about who develops COVID-19. It does not, by itself, invalidate the within-cohort analyses that examined relationships between genotype and clinical outcomes among the 150 patients themselves, since those comparisons do not depend on the control group’s genotype distribution. This distinction matters for readers trying to judge which parts of the original study survive scrutiny and which do not: the Factor V Leiden association, supported by a control group in equilibrium and by confidence intervals that exclude the null value, retains its direction, while the MTHFR association is now explicitly downgraded to preliminary.

Beyond the specific variants, the exchange highlights what rigorous genetic association research demands. The authors list the lessons directly: prospectively defined genotyping quality-control procedures, explicit Hardy–Weinberg equilibrium assessment, complete reporting of genotype and allele frequencies, and effect estimates with confidence intervals, ideally in larger and independently recruited populations. The COVID-19 pandemic produced hundreds of candidate-gene studies conducted under intense time pressure, many with modest sample sizes and inconsistent reporting standards, and this reply demonstrates how post-publication critique can improve the scientific record without hostility. The authors close by thanking Dr. Panda for the opportunity to clarify their findings, a courteous ending to a dispute whose real value lies in modeling how statistical scrutiny, honest reanalysis, and transparent limitation-setting should work together to keep genetic medicine on solid ground.

Subject of Research: Hardy–Weinberg equilibrium reassessment of Factor V Leiden and MTHFR polymorphism associations with COVID-19 outcomes

Article Title: Reply to “Important Methodological and Statistical Concerns Regarding the Association of Factor V Leiden and MTHFR Polymorphisms With COVID‐19 Outcomes”

Article References: Çıralıoğlu, H., Adalı, Y., Özen, M., Oskay, A., Yılmaz, A., Seyit, M., Köseler, A., & Türkçüer, İ. (2026). Reply to “Important Methodological and Statistical Concerns Regarding the Association of Factor V Leiden and MTHFR Polymorphisms With COVID‐19 Outcomes”. Immunity, Inflammation and Disease, 14(10), Article e70556. https://doi.org/10.1002/iid3.70556

Image Credits: AI Generated

DOI: 10.1002/iid3.70556

Keywords: COVID-19, Factor V Leiden, MTHFR A1298C, Hardy-Weinberg equilibrium, genetic association study, odds ratio, case-control study, genotyping quality control, population genetics, thrombosis, scientific reproducibility, allele frequency