Commentary: plasma-derived circALG8 and circCAMTA1 as a panel for early diagnosis of non-small cell lung cancer

1. To the editor

I read with great interest the article by Sheng et al., “Plasma-derived circALG8 and circCAMTA1 as a panel for early diagnosis of non-small cell lung cancer” [1]. The authors identify two plasma circRNAs and report that their combination – especially when integrated with serum markers – achieves AUCs up to 0.90 for early NSCLC. The demonstration of circRNA stability under RNase R, actinomycin D, freeze – thaw, and temperature stress is important. However, several methodological and statistical issues raise concerns about the robustness of these findings.

2. First, the study has limitations

The discovery phase relied on sequencing only three lung adenocarcinoma (LUAD) patients, three lung squamous cell carcinoma (LUSC) patients, and three healthy controls. These small cohorts are vulnerable to false discoveries and lack reproducibility [2–4]. Larger and ideally multicenter cohorts with multiple-testing corrections are needed.

3. Second, control group selection

Validation compared patients with NSCLC with only healthy volunteers. Without benign pulmonary disease, COPD, or infectious disease controls, the specificity is likely overestimated [2,4].

4. Third, the use of NSE and CYFR21-1, despite their nonsignificance

Table 2 shows that NSE (AUC 0.5310, p = 0.6462) and CYFR21-1 (AUC 0.5966, p = 0.1382) have no diagnostic value¹. Nevertheless, they are incorporated into combined models that report striking AUC increases (e.g., circCAMTA1 + NSE AUC = 0.9010; circCAMTA1 + NSE + CEA + CYFR21-1 AUC = 0.9255). Such jumps are implausible when individual markers are nearly random, suggesting possible model overfitting.

5. Finally, the absence of 95% confidence intervals

All AUCs and diagnostic metrics are presented without 95% CIs. Confidence intervals are critical to judge precision and overlap; without them, it is impossible to know if differences are statistically meaningful [5,6]^3,5. DeLong’s method should be used for ROC comparisons [6]⁵.

6. Conclusion

While circALG8 and circCAMTA1 appear stable and promising for NSCLC detection, key concerns – including the inclusion of statistically null markers in combined models, the absence of 95% CIs, small discovery cohorts, limited control groups, and incomplete statistical reporting – significantly weaken confidence in the reported performance. Future work should expand discovery cohorts, incorporate disease controls, apply rigorous normalization, and report full model details with validated performance before clinical application can be considered.

Disclosure statement

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.