Last month, a preprint, partially authored by the Czech group SMIS, linked COVID-19 vaccination to decreased fertility in Czech women. Six journal editors read the submission and rejected it before peer review even began. SMIS and its followers interpreted this as censorship and an attempt to suppress inconvenient truths, but in reality, it reflected a rapid and justified response to fundamental scientific flaws. That speed isn't arrogance; it’s professional triage. Editors handle hundreds of manuscripts annually and are trained to spot basic methodological landmines instantly. Here are the ones they saw, landmines so elementary that every first-year research trainee learns to avoid them in year one of their education.
First, no inferential statistics. The authors compare frequencies using barplots, vaccinated versus unvaccinated, but never test whether the difference could appear by chance. A picture is not a p-value. Without it, the "effect" may be pure noise. Imagine flipping a coin ten times and getting seven eagles. Does that mean the coin is rigged? You might expect five eagles and five lions, but small deviations are normal. To decide whether the result is meaningful or just random variation, we need to apply basic statistics. You can't just eyeball a graph and declare a discovery.
Second, the ecological fallacy. Births were counted for thousands of women grouped into two giant buckets. Drawing conclusions about individual biology from bucket averages is a classic error that can even reverse the true direction of an association. A common example is the claim that countries with higher chocolate consumption win more Nobel prizes. While true at the population level, it says nothing about whether eating chocolate makes any individual more likely to win an award. Similarly, linking average birth rates to vaccination status without analyzing individuals leads to misleading inferences.
Third, exposure is guessed, not measured. The team "estimated" who was vaccinated before conception by subtracting doses given during pregnancy. Any misclassification here ripples straight into the result.
Now, these three problems do not automatically mean that the result of the study is wrong. Maybe there really is a statistically significant difference in birth rates between vaccinated and unvaccinated women. Maybe the misclassifications are minimal and don’t distort the picture. Perhaps the population is so homogenous that ecological fallacies and confounding don’t change the conclusion much. But these are all maybes. And in scientific publishing, these kinds of flaws are enough to trigger rejection before any of that can be tested.
And we haven’t even reached the most significant problem yet: causality.
The analysis does not adjust for several plausible confounders. Maternal age, socio-economic status (income and education), contraceptive use and whether the pregnancy was planned are all associated with both COVID-19 vaccination uptake and fertility outcomes. Such variables are classic confounders, third factors that can distort the apparent exposure-outcome relationship. Unless they are measured and controlled, any association we observe cannot be interpreted causally. Omitting them does not merely affect statistical precision; it introduces systematic bias that can make correlation masquerade as causation. The authors acknowledge the limitation of not having individual-level data, but acknowledgement alone cannot neutralise the bias. Without additional data, sensitivity analyses, or design features that break the confounding link, causal language exceeds what the evidence can support.
Finally, strong causal language without a causal bridge. Speculation is fine; stating it as fact is not. Yet the language of the paper repeatedly leans toward one-sided conclusions, implying causal links that the data cannot support. For example, the authors mention in vitro studies where spike protein exposure may have affected ovarian cells. These findings, while interesting, are from highly artificial lab settings and have no direct bearing on birth rates in national populations. Suggesting otherwise is a leap that no trained scientist should make without substantial bridging evidence.
With those five problems lined up, editors did not need referees to conclude that the study’s conclusions outran its data.
What Robust Studies Find Instead
Below are four example investigations that follow the epidemiology rulebook. These studies either use individual-level data or rely on rigorous meta-analytic synthesis, adjust for confounders, and report statistical uncertainty. This is important because the authors behind the Czech preprint, especially on their social media platforms, have repeatedly claimed that no one is seriously examining the apparent drop in birth rates. That is simply not true. These studies do look into the issue, and they do so using methods that avoid the major pitfalls described above. Ignoring such robust evidence while insisting the topic is being neglected is misleading at best. Sound science means engaging with all the data, not only the fragments that fit a chosen narrative.
-
North-American couples cohort (Wesselink et al., 2022): Researchers followed 2,126 couples actively trying to conceive, logging vaccination status prospectively and analyzing time-to-pregnancy cycle by cycle. Multivariable models showed no difference in fecundability for vaccinated women or men; if anything, recent SARS-CoV-2 infection, not vaccination, briefly reduced male fertility.
-
Norwegian miscarriage registry (Magnus et al., 2021): Using national linked health records, investigators compared more than 18,000 first-trimester miscarriages with ongoing pregnancies. After adjusting for age and calendar time, vaccinated women were not at higher risk of miscarriage (odds ratio ~0.9). Large data, rigorous linkage, clear result.
-
Global meta-analysis of 40 studies (Fernández-García et al., 2024): This systematic review pooled >150,000 pregnancies. Vaccination reduced severe maternal COVID-19, had no adverse signal for conception, miscarriage or stillbirth, and slightly improved some neonatal outcomes. When dozens of datasets point the same way, the weight of evidence is hard to ignore.
-
Assisted-reproduction meta-analysis (Chamani et al., 2024): For people undergoing IVF, an ideal setting to scrutinize eggs, embryos, and implantation, researchers combined data from eleven studies. Ovarian response, embryo quality, and clinical pregnancy rates were identical in vaccinated and unvaccinated patients. That is about as close to a controlled fertility stress-test as one can get.
Take-Home Message
Flashy graphs on social media are not proof of vaccine-related reproductive harm, such as reduced fertility, miscarriage, or disrupted menstrual cycles, especially when the analysis skips the first chapters of every epidemiology handbook. Yet this is exactly the kind of material that can mislead the general public. When presented with confident charts and scientific-sounding language, even educated laypeople can be fooled by those who are either reckless or intentionally deceptive.
In reality, when scientists collect individual-level data, measure exposure accurately, and adjust for obvious confounders, the alarming fertility signal vanishes. COVID-19 vaccines remain a safe, effective way to protect adults, including those planning a family, from the real risks of the virus itself.
No comments:
Post a Comment