Figure–text discrepancy and methodological considerations in evaluating KL-6, SP-D, and CCL-18 as biomarkers of SSc-ILD in Indian patients

Dear Editor,

We read with great interest the article by Dubey et al. on serum biomarkers KL-6, SP-D, and CCL-18 in systemic sclerosis–associated interstitial lung disease (SSc-ILD) in Indian patients, recently published in Lung India.[1] The clinical question addressed is both timely and relevant, particularly given the relative paucity of Indian data in this domain. We would, however, like to highlight certain observations that may influence the interpretation of the findings.

First, the manuscript reports the correlation between KL-6 and FVC as r = −0.58 (P = 0.01) in both the text and Table 3, whereas Figure 1 depicts r = −0.38 (P = 0.01) for what appears to be the same analysis. These values are not concordant and cannot both be correct. The discrepancy is not trivial: an r value of − 0.58 corresponds to approximately 34% of the variance in FVC being explained by KL-6, indicating potential clinical utility, whereas an r of − 0.38 accounts for only about 14%, which is of considerably more limited relevance. The conclusion that KL-6 “may serve as a valuable biomarker for disease monitoring and treatment escalation” is therefore contingent on the accurate estimate, and clarification from the authors, or correction, is warranted.

Second, the analysis includes correlation testing of three biomarkers against five clinical or radiological parameters, along with five inter-group comparisons (dcSSc vs lcSSc, ILD vs non-ILD, Scl-70 status, dyspnoea grade, and disease progression) for each biomarker. This results in more than thirty statistical tests, performed without adjustment for multiple comparisons. Several headline associations, including SP-D versus HRCT (P = 0.04), SP-D versus mRSS (P = 0.02), and SP-D versus DLCO (P = 0.017), would not survive a Bonferroni or false-discovery-rate adjustment.[2] At α =0.05, one or two false positives are statistically expected from such an analysis even under the null.

Third, the comparison most directly relevant to the stated objective—biomarker performance in distinguishing ILD from non-ILD—yields a CCL-18 result of P = 1.00 (65 vs 11 patients). A P value at the maximum, even allowing for the small non-ILD arm, indicates effectively zero discrimination. This null is difficult to reconcile with the abstract’s conclusion that all three biomarkers “are promising tools for monitoring disease progression”; for CCL-18, the present data do not support that claim.

Fourth, immunosuppressive exposure at the time of sampling is not reported. Mycophenolate, cyclophosphamide, rituximab, and antifibrotics modulate KL-6 and CCL-18 levels independent of underlying disease activity,[3,4] and in a cross-sectional design, the resulting confounding is non-trivial. The cutoffs employed (KL-6 ≥500 U/mL, SP-D ≥110 ng/mL, CCL-18 ≥53 ng/mL) are imported from Japanese cohorts using different ELISA platforms.[5] In the index study, KL-6 was measured with a MyBiosource kit and SP-D and CCL-18 with BT-Assay kits—neither matches the original derivation assays. A clinician applying the recommended thresholds to a patient measured on a third platform may misclassify lung involvement in either direction. Locally derived ROC-based cutoffs, indexed to the assay actually used, would address this directly.

These points are not intended to diminish the underlying contribution; an Indian dataset on SSc-ILD biomarkers is genuinely needed. We would, however, encourage the authors to clarify the KL-6/FVC correlation, present a treatment-adjusted analysis, and report assay-specific cutoffs before the markers are recommended for routine monitoring.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.