Original educational illustration. The groups are conceptual, not a numerical chart.
A large survey can tell us something important about its respondents without telling us how common infertility is across an entire country. The Polish National Fertility Test illustrates why every striking percentage needs three things: a defined population, a clear question and the correct denominator.
This is not a reason to dismiss participants’ experiences. Reproductive difficulties matter, including when a person is still seeking a diagnosis. It is a reason to distinguish listening to those experiences from estimating the prevalence of a clinically defined condition in Poland. “National” describes the project; it does not certify representative sampling.
Who took part?
The researchers analysed the first 8,000 adult respondents to an anonymous online survey conducted in 2024. Invitations were distributed through Medonet, national media and social media. Participation was voluntary and unpaid. People opted in; they were not randomly selected from Poland’s adult population. The authors explicitly describe a self-selected convenience sample rather than a probabilistically representative sample. Methods, accepted manuscript, page 5.
That recruitment pathway matters. To respond, someone must encounter the invitation, be interested enough to open it and choose to complete the questions. People looking for fertility information may have different experiences or motivations from people who ignore the invitation. The survey cannot tell us what those nonrespondents would have reported.
Nationwide distribution does not remove this limitation. Responses can come from many locations without reproducing the population’s composition. Adding participants may describe the selected group more precisely, but it does not turn self-selection into random sampling. Sample size and representativeness answer different questions.
An important correction: the main analysis excluded people who had never tried
A fair critique must describe the actual methods. The main analysis excluded 1,243 respondents who had not yet tried to conceive and 1,225 who answered that the question did not apply. That left 5,532 people: 3,316 women and 2,216 men. They had tried to conceive at some point, not necessarily only at the time of the survey. Methods and results, pages 7 and 12.
It would therefore be incorrect to accuse the main analysis of simply combining people who had never tried with everyone else. The legitimate distinction is between having ever tried and currently trying. Different histories, durations and timings of attempts may affect comparisons, but they are not the same as never having attempted pregnancy.
What does 28.9% mean?
In the main analytical group, 1,601 of 5,532 respondents reported reproductive-health problems during attempts to conceive: 28.9%. Among women, the figure was 1,215 of 3,316, or 36.6%; among men it was 386 of 2,216, or 17.4%. These are proportions within this survey sample, not representative estimates of clinically diagnosed infertility in Poland. Table 3, page 36.
The question asked about ever experiencing reproductive-health problems during attempts, giving examples such as trying for more than a year or receiving an infertility diagnosis. A participant’s answer is not equivalent to independently checking identical diagnostic criteria in every respondent. A broad question captures a broad range of experiences.
The authors acknowledge that the groups may include diagnosed infertility, ongoing investigation and distress associated with trying to conceive. All deserve attention. However, combining them does not produce a single clinically verified diagnosis. The study’s limitations explicitly distinguish reported difficulties from confirmed clinical infertility. Pages 23–24.
Different percentages are not interchangeable
The paper also reports 20.0% and 14.0%. These should not be selected according to which makes the strongest headline. The 20.0% figure counts the same 1,601 reports against all 8,000 respondents. The 14.0% figure refers to 1,122 participants classified using a different composite definition in the full sample. Different definitions and denominators produce different quantities. Definitions, pages 7–8; Table 2, page 35.
The additional definition combined reported specialist diagnosis or more than a year of trying together with fertility-related help-seeking in the preceding year. Descriptive analyses using that definition included the full sample regardless of conception history. This creates a legitimate question about comparing people with different opportunities to recognise a problem. It must not be confused with the restricted main analysis.
Why the sample’s composition matters
Across all 8,000 respondents, 56.7% held a master’s degree and 29.2% lived in cities with more than 500,000 residents. These describe the full sample, not separately the subgroup of 5,532. The authors discuss demographic skew and limited generalisability. Participant characteristics and limitations, pages 9–10 and 23.
Education, information access, symptom recognition and contact with healthcare may affect both participation and reporting. Identifying those possibilities does not question participants’ honesty. It identifies a property of recruitment. A methodological criticism should target the inference being made, not the worth or credibility of people sharing their experiences.
A supplement association is not evidence of harm—or benefit
Associations between reported difficulties and supplement use can arise when people start buying products in response to a problem. They do not establish that supplements caused the difficulties. Equally, they do not demonstrate that those products treated them. The authors discuss help-seeking and reverse causation as possible explanations. Discussion, pages 21–23.
A cross-sectional survey often cannot establish what came first. Did behaviour precede the difficulty, follow it, or share another cause with it? Reverse causation is a plausible interpretation here, not a proven biography of each respondent. Neither a positive nor a negative association should become a treatment recommendation on that basis.
Adjustment does not mean every difference was removed
The researchers used separate regression models for individual predictors, adjusting for age. That is not the same as one comprehensively adjusted model accounting for all potentially relevant characteristics. Multiple-comparison correction was used for the main analyses, but not every detailed supplement ingredient or behaviour analysis. Statistical analysis, pages 8–9.
Age adjustment can improve a comparison without eliminating residual confounding. Unmeasured or insufficiently characterised differences may remain. Unexpected associations are therefore better treated as questions for further study than as instructions for prevention. Statistical modelling cannot, by itself, repair the recruitment pathway or validate every self-reported diagnosis.
What remains useful?
The survey documents how a particular group of internet users describes reproductive difficulties and related behaviours. It can highlight educational needs and help generate research questions. Its descriptive value does not depend on pretending that the sample represents everyone. Nor does its lack of representativeness make every observation meaningless.
The strongest defensible criticism is specific: this design has limited value for estimating population prevalence. That is more accurate than calling the entire study worthless. Importantly, the authors disclose the sampling limitation in their own work. Criticism should acknowledge that transparency rather than imply they concealed the problem.
How to read the next fertility headline
“Almost one person in three experienced a problem” is incomplete. One in three of whom: all adults, people planning a family, people trying for a defined period, or volunteers answering an online questionnaire? Each group could produce a different result. The denominator is part of the finding, not optional fine print.
The comparison between women and men also requires care. This survey compared respondents’ reports, not independently verified causes of infertility in matched couples. A higher reporting proportion does not establish that one sex accounts for a greater share of couples’ infertility. Turning those percentages into blame would be both methodologically wrong and harmful.
Similarly, not reporting a difficulty does not prove normal fertility. An answer describes a response to a particular question at a particular time. This was not a diagnostic examination of every participant. “Not applicable” is not a negative clinical test. Those categories should remain visible rather than disappear inside a single percentage.
Good communication need not remove the numbers. It should label them properly: all respondents, exclusions, the analytical group and the outcome definition. This lets readers see what was measured and what remains unknown. A less sensational headline can be far more useful to patients, clinicians and health educators.
Before sharing a result, ask: who could enter, who actually answered, what counted as a problem and which denominator was used? Then check whether the headline stays with those respondents or quietly generalises to society. A large number of responses cannot justify that last step.
The defensible conclusion is that 28.9% of a self-selected group who had ever tried to conceive reported difficulties. The study does not establish a representative prevalence of clinical infertility in Poland. For individual concerns, seek assessment based on personal circumstances—not an online survey percentage presented as personal prognosis.
Source and transparency
Pilarczyk R and colleagues. Self-reported fertility problems and their socio-demographic and health correlates: a cross-sectional online survey of 8000 Polish adults. Reproduction and Fertility, accepted manuscript, 2026. DOI: 10.1530/RAF-26-0056.
The authors report no conflicts affecting impartiality and no external research funding; publication charges were funded by Wrocław University of Science and Technology’s Faculty of Medicine. Affiliations include Medonet and Labplus R&D. Affiliations alone do not prove bias. This article is methodological education based on the full manuscript, not individual medical advice.