Methodological Limitations in “Self-Reported Vision-Related Quality of Life and Functional Impact in Portuguese Patients with Uveitis: A Cross-Sectional Study” [Letter]

Dear editor

We read with interest the cross-sectional study examining self-reported quality of life among Portuguese patients with Uveitis.¹ While this study is clinically important, and there is scarce data among this population, we have identified several methodological limitations that are of concern.

First, the study’s online questionnaire format means the uveitis diagnosis relies on self-reporting without any clinical verification, relying instead on consistency between reported symptoms and duration.¹ Because the questionnaire was deployed online rather than through an in-person or clinician‑checked method, there was no opportunity to cross-check responses against medical records at the point of data collection. The discussion then suggests recruitment through patient associations “may have increased the likelihood of prior ophthalmological confirmation” but this is speculation, not verification, and should not be read as mitigating a limitation that is inherent to the online survey design. This is a source of information bias and distorts associations due to under- or over-reporting.² The self-administered web questionnaires are known to bring social desirability bias, since participants respond without any interviewer present to probe or clarify answers,³ and they also invite recall bias, since disease burden was characterized retrospectively rather than captured in the moment or verified against clinical documentation. We would recommend the authors explicitly acknowledge that the online, self-report nature of the survey prevents diagnostic verification, and exclude the recruitment-based mitigation argument in favor of stating plainly that diagnostic misclassification could not be ruled out.

Second, the control group was 89.3% female (n=50) vs only 10.7% male (n=6), while the uveitis group skewed the opposite way at 55.6% male (p<0.0001). Since prior literature cited in the introduction links female sex to worse psychosocial outcomes in uveitis generally,^4,5 this imbalance likely works against finding a group difference rather than inflating one. Yet in the adjusted models, male sex was independently linked to markedly lower odds of daily-activity difficulty and driving difficulty, which are large effects drawn from a control cell of just six men. We would ask the authors to re-run these models using exact or Firth logistic regression, which are better suited to sparse strata.

This imbalance is also a product of how participants were selected. Eligibility required self-reported diagnosis, Portuguese language, and online accessibility. Participants were recruited through patient associations and professional networks, creating a very narrow eligible pool and a sampling bias that Table 1 documents but does not statistically correct for.⁶ A recruitment flow diagram (number of individuals eligible, approached, and enrolled) would let readers assess the extent of participation bias.

Finally, despite five separate multivariable logistic regression models, none is presented in tabular form. The adjusted odds ratios, CIs, and p-values appear only in running text, and for the emotional/social domain only three of nine measured outcomes are reported at all (“among others”), with the rest omitted. Table 1 itself includes only three baseline variables, leaving pain, driving status, and activity-limitation data absent from any consolidated comparison, a gap notable given the authors’ stated adherence to STROBE guidelines.⁷ We would ask the authors to publish full regression tables for all models, to report all outcomes (or move the remainder to supplementary table), to expand Table 1 to include the omitted baseline variables, and to reconcile their reporting against a completed STROBE checklist.

Separately, while it is asserted that the NEI VFQ-25 “has been culturally and linguistically validated and adapted”, the only citation given is the instrument’s original English-language development paper, and no Portuguese validation study is referenced, so this claim of local adaptation is unverified.⁸ We would request the authors to cite a Portuguese‑language validation study if one exists, or clarify the adaptation methodology (eg, pilot testing) used.

Taken together, these issues, ie, unverified diagnosis, unstable sex‑adjusted estimates, and the incomplete reporting of regression models, bear directly on whether the paper’s central associations can be interpreted with confidence. We believe addressing them would substantially strengthen this valuable contribution to a data‑scarce area.

Disclosure

The authors report no conflicts of interest in this communication.