Interesting study here from Italy:
file:///C:/Users/User/Downloads/excli2025-8400.pdf
The Covid Vax apparently improves (decreases) all cause mortality for the time of the study while increasing several types of cancer.
Ovarian was really high, but not significant. I have trouble with that too.
Whatâs the source? (your upload didnât work)
Try this:
Thanks. It seems to be a fairly shitty journal, which makes me question the validity of the findings. If very good research I would expect a better journal to publish it. Still, a massive reduction in all-cause death is what matters to me over 30 month. The finding for cancer hospitalization âvaried by infection status, cancer site, and the minimum lag-time after vaccinationâ so it seems less reliable.
Scientists have previously found the vaccine appears to cut the risk of diabetes. This may be due to the hepatitis B virus â which infects the liver and spreads through blood, semen and vaginal fluids â disrupting the organâs ability to store sugar from the blood. This could raise the risk of diabetes, where blood sugar levels are persistently too high.
But prior studies have not looked at whether the vaccine might reduce diabetes risk among a group of both immunised and non-immunised people who havenât contracted hepatitis B, which would suggest the effect acts independently of just preventing the infection.
To explore this, Nhu-Quynh Phan at Taipei Medical University in Taiwan and her colleagues analysed the health records of more than 580,000 people living across the US, Europe, Africa, Latin America, the Middle East and the Asia-Pacific. On average, these records spanned nearly four years for each individual between 2005 and 2023.
The team found the vaccinated participants had an overall 15 per cent lower rate of diabetes â defined as them either receiving a diagnosis, having a persistently high blood sugar level or being prescribed diabetes drugs â than their unvaccinated counterparts. The vast majority of cases were type 2 diabetes, the most common form of the condition. The results will be presented at the European Association for the Study of Diabetes Annual Meeting in Vienna, Austria.
Although this is observational research, the scientists also found a dose-response effect, where the vaccinated participants with higher levels of hepatitis B-specific antibodies were less likely to develop diabetes than those with lower levels. Differences in antibody levels may be a reflection of how many vaccine doses the individual participants received, how recently they were immunised or general variation in immune responses.
Read the full story:
@Bicep , an ChatGPT analysis of that study, notice the weaknesses:
Hereâs a concise, methods-focused critique of the EXCLI Journal cohort study you linked.
What the study did (in brief)
- Population-wide cohort of residents â„11 y in Pescara, Italy, followed June 27 2021âDec 31 2023. Exposure = COVID-19 vaccination (â„1 dose; â„3 doses). Outcomes = all-cause mortality and first hospital admission with a cancer diagnosis (skin cancers excluded), identified via ICD-9 codes. Cox models adjusted for age, sex, selected comorbidities and recorded prior SARS-CoV-2 infection; sensitivity analyses imposed 90/180/365-day lags from vaccination to outcome. (excli.de)
- Key results: vaccinated groups had lower all-cause mortality; cancer hospitalization risk was modestly higher for ââ„1 doseâ at 180 days (HRâ1.23) but the association disappeared or reversed with a 365-day lag (â„3 doses HRâ0.90). Authors emphasize findings are preliminary due to healthy-vaccinee bias and unmeasured confounding. (excli.de)
Strengths
- Very large, population-level dataset with deterministic linkage across health registries. (excli.de)
- Prespecified adjustments for major clinical covariates; proportional-hazards assumptions checked; multiple lag-period sensitivity analyses. (excli.de)
- Authors openly discuss healthy-vaccinee bias and testing misclassification, appropriately tempering causal claims. (excli.de)
Main weaknesses / sources of bias
-
Outcome â incidence (ascertainment bias).
âCancerâ is proxied by first hospital admission with a cancer code, not incident diagnosis. Admission practices, care-seeking, and screening intensity differ by age, sex, SES, and health behaviorsâfactors unevenly distributed between vaccinated and unvaccinated groups. This can inflate apparent cancer âriskâ without reflecting true incidence (e.g., more screening â more admissions for surgery). The paper acknowledges higher vaccinated age/comorbidity profiles but lacks data on screening participation, primary-care use, or socioeconomic variables. (excli.de) -
Exposure handling may induce time-related bias.
Follow-up for vaccinated starts 180 days after the 1st/3rd dose; unvaccinated follow-up begins on fixed calendar dates. Vaccination status appears treated as group membership rather than a time-varying exposure, raising risk of immortal-time and misclassification bias (person-time prior to vaccination may be misattributed). The paper does not report time-updated vaccination in the Cox model. (excli.de) -
Competing risks not modeled.
Vaccination is strongly associated with lower all-cause mortality (HRâ0.42 for â„1 dose). Lower death rates leave more vaccinated individuals alive to be hospitalized later, potentially increasing observed hospitalization counts independent of cancer biology. No Fine-Gray or other competing-risk analyses were presented. (excli.de) -
Residual confounding is substantial and mostly unaddressed.
Adjustments cover a few hospital-coded comorbidities and recorded infections. Missing are smoking, alcohol, obesity/BMI, occupational exposures, family history, medications, and socioeconomic statusâkey determinants for breast, colorectal, bladder, and hematologic cancers. The authors concede unmeasured confounding and healthy-vaccinee bias but cannot quantify them. (excli.de) -
Fragility to analytic choices / multiple testing.
Signals fluctuate across lags, sexes, cancer sites, and vaccine products (e.g., 180-day HR>1 for several sites; 365-day lag attenuates or reverses). No correction for multiple comparisons is reported despite many subgroup and sensitivity analyses, increasing false-positive risk. (excli.de) -
Infection status misclassification.
âPrior SARS-CoV-2 infectionâ relies on recorded tests only, in a context where testing policies changed and many infections went unrecorded. Stratified results differ sharply by recorded infection, which the authors warn should be interpreted cautiously. (excli.de) -
Generalizability and construct validity.
Single-province setting; outcome is hospital admission, not registry-confirmed incidence or stage. Skin cancers excluded; stage, treatment intent, and pathology data unavailable. Manufacturer-stratified results are presented, but product groups differ in age/indication mixes; confounding by indication likely. (excli.de) -
Mechanistic citations lean on speculative literature.
The discussion cites hypothesis papers and preprints proposing oncogenic mechanisms of vaccination (e.g., mRNA/LNP effects, DNA contamination) and explicitly says they âneed validation.â Including these does not strengthen causal inference from the current observational signal. (excli.de)
How the analysis could be strengthened
- Use time-varying vaccination and infection status (target-trial emulation), with calendar-time and time-since-vaccination terms to address dynamic pandemic conditions.
- Model competing risks (death) and present absolute risks and risk differences alongside HRs.
- Link to cancer registry data (date of diagnosis, histology, stage) and screening records(mammography/colonoscopy) to reduce misclassification and screening bias.
- Include negative-control outcomes/exposures to detect residual bias; consider self-controlled designs for short-term diagnostic surges.
- Adjust or report for multiple testing (e.g., FDR) across numerous site-/sex-/product-specific analyses.
Bottom line
The studyâs own sensitivity analyses and caveats point to instability of associations for cancer hospitalization and major susceptibility to residual and time-related biases. As designed, it does not establish a causal link between vaccination and cancer risk (in either direction). The clearest signalâlower all-cause mortality among the vaccinatedâis itself likely influenced by healthy-vaccinee and other unmeasured factors the authors acknowledge. Overall, the paper is hypothesis-generating and underlines the need for registry-based incidence analyses with rigorous time-varying and competing-risk methods before drawing conclusions about cancer risk. (excli.de)