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

Dear editor

We thank the correspondents for their interest in our study and for the opportunity to clarify the methodological and reporting issues raised in their Letter. We agree that the findings should be interpreted as exploratory patient-reported outcomes.

The correspondents correctly note that the uveitis diagnosis was not independently verified through clinical examination or medical records. We clarify, however, that eligibility for the uveitis group required participants to self-report a previous ophthalmological diagnosis of uveitis; consistency between the reported diagnosis, symptoms, and disease duration was used only as an additional data-quality check and did not replace that eligibility requirement. The published Methods explicitly state that no external clinical verification of diagnosis was performed and that no clinical examination or medical records were obtained to determine uveitis subtype, disease activity, or treatment status. We therefore agree that diagnostic misclassification, recall bias, social desirability bias, and other response-related biases cannot be excluded. The statement concerning recruitment through patient associations was intended only to describe the recruitment context and should not be interpreted as clinical verification or as mitigation of this limitation.

We agree that the marked sex imbalance, particularly the small number of men in the group without uveitis, warrants additional examination. The published Table 1 reported 20 men and 16 women in the uveitis group and 6 men and 50 women in the group without uveitis. In direct response to the correspondents’ request, we repeated the two specifically highlighted models using Firth penalised logistic regression. For any difficulty in daily activities, the Firth estimates were OR=0.072 (95% CI 0.009–0.395; p=0.002) for male sex and OR=6.03 (95% CI 0.69–66.23; p=0.104) for uveitis. For daytime driving difficulty, the corresponding estimates were OR=0.193 (95% CI 0.027–0.887; p=0.033) and OR=19.65 (95% CI 2.77–235.04; p=0.002), respectively. These sensitivity analyses yielded conclusions consistent with the originally reported models, although the wide confidence intervals continue to indicate substantial imprecision.

We also agree that the recruitment strategy may have introduced selection bias. Recruitment was conducted through an openly distributed online questionnaire using patient associations, optometry clinics, social networks, and professional groups, and there was no defined sampling frame from which the number of individuals exposed to the invitation could be determined. Consequently, the numbers approached and eligible but not enrolled, and therefore a conventional participation rate, cannot be reconstructed retrospectively without introducing unsupported estimates. For the same reason, a recruitment flow diagram containing approached and eligible denominators cannot be produced reliably. This limitation should be considered when assessing representativeness and generalisability.

We agree with the correspondents that complete reporting of the multivariable analyses improves transparency. Each model included age, sex, and uveitis status, with female sex and absence of uveitis as the reference categories. For constant vision concern, the estimates for age, male sex, and uveitis were OR=1.20 (95% CI 1.08–1.35; p=0.001), OR=1.520 (95% CI 0.439–5.199; p=0.502), and OR=0.754 (95% CI 0.168–3.189; p=0.704), respectively. For severe ocular pain, the corresponding estimates were OR=1.309 (95% CI 1.142–1.580; p<0.001), OR=7.377 (95% CI 1.007–76.252; p=0.061), and OR=0.063 (95% CI 0.004–0.649; p=0.027), respectively. For any difficulty in daily activities, they were OR=1.088 (95% CI 0.991–1.215; p=0.096), OR=0.052 (95% CI 0.005–0.331; p=0.004), and OR=7.896 (95% CI 0.773–113.953; p=0.095), respectively. For daytime driving difficulty, they were OR=0.953 (95% CI 0.876–1.029; p=0.235), OR=0.145 (95% CI 0.017–0.761; p=0.042), and OR=28.975 (95% CI 3.493–422.498; p=0.005), respectively. For nighttime driving difficulty, they were OR=0.928 (95% CI 0.840–1.015; p=0.117), OR=0.272 (95% CI 0.049–1.396; p=0.116), and OR=14.721 (95% CI 1.540–209.788; p=0.029), respectively.

For the emotional/social outcomes, the estimates for age, male sex, and uveitis, respectively, were as follows: stopping activities, OR=1.112 (95% CI 1.031–1.215; p=0.010), OR=0.242 (95% CI 0.033–1.102; p=0.098), and OR=8.597 (95% CI 1.657–67.718; p=0.018); feeling limited, OR=1.063 (95% CI 0.985–1.154; p=0.125), OR=0.334 (95% CI 0.046–1.529; p=0.201), and OR=19.016 (95% CI 3.553–161.947; p=0.002); ocular discomfort leading to activity cessation, OR=1.211 (95% CI 1.095–1.372; p<0.001), OR=0.291 (95% CI 0.036–1.546; p=0.184), and OR=18.9 (95% CI 3.50–163; p<0.001); staying at home because of vision, OR=1.190 (95% CI 1.078–1.338; p=0.001), OR=0.844 (95% CI 0.167–3.862; p=0.830), and OR=6.451 (95% CI 1.224–39.742; p=0.033); frustration because of vision, OR=1.050 (95% CI 0.973–1.137; p=0.215), OR=0.322 (95% CI 0.044–1.470; p=0.186), and OR=22.098 (95% CI 4.050–194.260; p=0.001); loss of control because of vision, OR=1.116 (95% CI 1.025–1.230; p=0.017), OR=0.275 (95% CI 0.037–1.326; p=0.143), and OR=14.966 (95% CI 2.715–125.769; p=0.004); dependence on others, OR=1.102 (95% CI 1.020–1.200; p=0.018), OR=0.784 (95% CI 0.198–2.889; p=0.720), and OR=3.748 (95% CI 0.806–19.380; p=0.099); needing help from others, OR=1.189 (95% CI 1.089–1.324; p<0.001), OR=0.480 (95% CI 0.098–1.985; p=0.331), and OR=2.482 (95% CI 0.490–13.367; p=0.273); and fear of embarrassment, OR=1.190 (95% CI 1.078–1.338; p=0.001), OR=0.844 (95% CI 0.167–3.862; p=0.830), and OR=6.451 (95% CI 1.224–39.742; p=0.033). These calculations are provided as an additional source-data reanalysis for transparency and do not replace the numerical values reported in the published article.

We would, however, clarify that ocular pain, daily-activity limitations, driving difficulties, and emotional/social responses constitute study outcomes rather than baseline characteristics. STROBE distinguishes participant descriptive data (Item 14) from outcome data (Item 15), and the latter were reported throughout the Results section. We therefore do not consider it appropriate to retrospectively reclassify these outcomes as baseline variables in the published Table 1.

A completed STROBE checklist was submitted with the original article as Supplementary Material 1. With respect specifically to Item 13, we acknowledge that the original checklist could have documented the circumstances of recruitment more explicitly for Items 13(a) and 13(b). Because recruitment was based on an openly distributed online questionnaire without a defined sampling frame, the total number of individuals exposed to the invitation, potentially eligible, or eligible but not enrolled, as well as reasons for non-participation, cannot be reconstructed retrospectively. Item 13(c), concerning consideration of a flow diagram, was marked as not applicable. This clarification does not change the analysed sample or the reported results but more accurately describes the limitations inherent to the recruitment strategy.

Finally, we agree that the Portuguese-language basis of the NEI VFQ-25 should have been documented more precisely. A Portuguese-language version has undergone psychometric evaluation in Brazilian Portuguese,¹ and a subsequent clinical study conducted in Porto, Portugal, used what the authors described as the validated Portuguese version and cited that psychometric study as its supporting Portuguese-language reference.² The questionnaire used in our study was informed by the original NEI VFQ-25 and available Portuguese-language wording, with linguistic adaptation for European Portuguese. We do not claim that the specific European Portuguese wording used in our survey underwent a separate formal psychometric validation before administration. We therefore acknowledge this explicitly as a limitation and provide the relevant Portuguese-language references below.

We thank the correspondents for highlighting these issues. The additional Firth sensitivity analyses support the interpretation of the originally reported models while also confirming the substantial imprecision associated with the modest sample size and sparse strata. We hope that the additional analyses and clarifications provided here improve the transparency and interpretation of the study.

Disclosure

The authors report no conflicts of interest in this communication.