What Does “Success” Mean in the Cost-Effectiveness of Fetal Fibronectin Testing? [Letter] Authors: Güney Sarıoğlu Published: 2026 Dear editor I was interested in the study by Abushanab et al evaluating fetal fibronectin (fFN) testing in women presenting with preterm labor symptoms in Qatar.1 The authors should be commended for addressing an important local health-economic question using real-world data and for attempting to link diagnostic testing with downstream resource use. Their finding that fFN testing was associated with lower costs is potentially relevant for institutions seeking to reduce unnecessary admission and treatment. However, three internally testable issues appear central to interpretation of the reported dominance of fFN and merit clarification. First, the direction and meaning of the effectiveness endpoint are difficult to reconcile across the manuscript. The Methods define the primary measure as cost per “successful delivery within 48 hours” and also describe this as “successful prolongation of delivery within 48 hours”. Table 3 then classifies delivery within 48 hours as the primary outcome and reports a higher rate in the fFN group (94% vs 80%). Yet the Discussion states that the primary outcome was successful prolongation of pregnancy beyond 48 hours, which is presented as clinically important for corticosteroid administration and short-term tocolysis.1 These are not equivalent endpoints; in fact, they point in opposite directions. The ambiguity therefore concerns the estimand itself, rather than terminology alone. This distinction matters because fFN is primarily a risk-stratification test for imminent preterm birth, with greatest clinical value often derived from its ability to identify women at low short-term risk and thereby avoid unnecessary admission, transfer, corticosteroids, or tocolysis.2,3 If the intended effectiveness endpoint is delivery within 48 hours, this should be interpreted as an observed short-term delivery event relevant to risk classification, rather than as a favorable therapeutic outcome attributable to testing. Conversely, if the intended clinical benefit is prolongation of pregnancy for at least 48 hours to permit a therapeutic window, women remaining undelivered beyond 48 hours would represent success. Because these definitions imply opposite directions of effectiveness, the prespecified estimand and its coding should be stated explicitly before interpreting the reported dominance of fFN. Recalculating the analysis using the clinically intended direction of benefit, or reporting both formulations as sensitivity analyses, would show whether the conclusion of dominance depends on endpoint coding. Second, the modeled management pathways are not readily reconciled with the observed cohort. The decision tree states that women with positive fFN received one cycle of atosiban plus antenatal corticosteroids, whereas women with negative fFN received neither.1 However, the observed fFN cohort shows atosiban use in only 12.1% and corticosteroid use in 45.7%. Given that approximately 31% of the fFN pathways were positive, these aggregate treatment frequencies do not by themselves prove model misspecification because treatment use is not reported separately by fFN result. Nevertheless, they do not transparently correspond to a model in which every positive result triggers both interventions. The no-fFN branch is similarly simplified relative to observed care. This raises a clinically important question: is the economic model evaluating the real-world strategy that actually occurred, or a hypothetical protocol in which fFN results deterministically govern treatment? That distinction matters because a diagnostic test has economic value through changes in management. If the decision tree is intended to reproduce the retrospective management actually observed in the cohort, treatment, admission, and other major cost-driving probabilities should ideally be parameterized using empirical frequencies conditional on positive and negative fFN results. If, instead, the model represents a protocol-adherent strategy in which fFN results determine management, identify this explicitly as a hypothetical test-guided scenario. Otherwise, the estimated cost difference may reflect the assumed treatment protocol as much as the test itself. Prior economic evaluations of threatened preterm labor have explicitly compared diagnostic-management strategies rather than the test in isolation.3,4 Reporting both empirical and protocol-adherent scenarios would help distinguish the economic effect of observed practice from that of an assumed management pathway. Third, Table 2 contains an apparent arithmetic inconsistency that is directly relevant to reproducibility. For the “negative fibronectin, no delivery within 48 hours” branch, the reported probability is 0.0233 and the reported cost is QAR75,980. Applying the same apparent probability-weighting used in the other rows gives 0.0233 × QAR75,980 = QAR1,770.33, rather than the reported proportional cost of QAR172.68.1 Conversely, QAR172.68 divided by 0.0233 implies an underlying cost of approximately QAR7,411. We therefore cannot determine from the table whether the probability-weighted cost is incorrect, whether the listed branch cost contains a typographical error, or whether a different denominator or costing convention was applied. More broadly, specifying whether the “Cost, QAR” column represents mean per-patient branch cost or aggregate expenditure for all patients in that branch would make the derivation of the reported QAR6,470 per-person saving transparent. CHEERS 2022 emphasizes clear reporting of model inputs, valuation, and analytic assumptions precisely so that decision makers can reproduce economic results.5 These observations do not diminish the value of generating Qatar-specific evidence. Rather, they focus on whether the economic conclusion answers the clinical question the authors intended to study. Probabilistic sensitivity analysis can quantify uncertainty around chosen parameters, but it cannot resolve ambiguity in the direction of the effectiveness endpoint, determine whether modeled treatment pathways correspond to observed practice, or correct an internally inconsistent model input.5,6 Clarifying the endpoint, specifying whether the decision tree is empirical or protocol-based, and resolving the Table 2 calculation would allow readers to judge whether fFN remains dominant under a clinically coherent and reproducible model. The study has the potential to make a useful contribution because it connects a familiar diagnostic test with real-world resource allocation in a setting with limited local economic evidence. Addressing these specific points would, in my view, substantially strengthen that contribution and provide answers that clinicians and health-system decision makers can act on. Funding Statement I received no funding for this work. Declaration of Generative AI and AI-Assisted Technologies in the Manuscript Preparation Process During preparation of this work, the author used ChatGPT (version: GPT-5.6 Sol, OpenAI) to identify relevant methodological literature and improve language. Ethical Statement This Letter to the Editor is based exclusively on the critical appraisal of a previously published article and does not involve human participants, animals, identifiable personal data, or new data collection. Therefore, ethical approval and informed consent were not required. Author Contributions The author made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; has agreed on the journal to which the article has been submitted; and agrees to be accountable for all aspects of the work. Disclosure I declare no conflicts of interest in this communication. References