The study team evaluated five routine complete blood count–derived inflammatory indices—NLR, PLR, MLR, SIRI and SII—in patients with colorectal cancer (CRC) and in a very large healthy-screening population. The clinical dataset comprised 2,305 CRC patients and 78,281 health‑screening controls. An external replication cohort of 1,031 CRC patients was used to verify key signals. Investigators reported that all five indices were higher in CRC patients compared with controls and that the differences persisted after adjustment for measured confounders.
Authors described a head-to-head design that combined propensity score matching with multivariable survival modelling. The study team used Cox regression to evaluate associations with survival, restricted cubic splines to probe dose–response and potential non-linear relationships, and Harrell’s concordance index to quantify prognostic discrimination. The source did not report specific variable lists used for propensity matching in this summary; those methodological details were not provided in the article abstract.
Among the five indices, the monocyte–lymphocyte ratio (MLR) emerged as particularly informative in early disease. Data indicated that MLR retained both diagnostic and prognostic significance in Stage I CRC. The authors noted that this Stage I association was reproduced in the external replication cohort of 1,031 CRC patients, supporting the reproducibility of the early-stage MLR signal across independent samples.
Quartile-based analyses revealed dose–response trends across the five biomarkers. The investigators observed that higher quartiles generally corresponded with worse outcomes. For MLR and SIRI, however, spline analyses suggested non-linear threshold effects rather than strict linear increments in risk. The study abstract did not list exact inflection points or numeric thresholds for those non-linear effects; such numeric details were not reported in the summary provided.
The authors reported that while each index was independently associated with survival, the incremental prognostic gains beyond conventional tumour stage were modest when measured by concordance indices. In other words, adding CBC-derived indices to stage-based models produced only small improvements in discrimination. The article abstract did not provide specific concordance index values in this summary; those figures were not included in the available text.
To place biomarker values in clinical context, the study team derived population reference intervals for the five indices. The investigators found substantial overlap between the distributions observed in CRC patients and those in the healthy screening population. This overlap limits the value of fixed cutpoints for screening or for making treatment decisions. The authors explicitly advised that reference intervals should not be interpreted as CRC screening thresholds or as treatment decision boundaries.
The research group assessed cross‑population transferability by comparing their reference intervals with data from the National Health and Nutrition Examination Survey (NHANES). The comparison revealed non‑uniform transferability of reference intervals between populations. The abstract did not specify which indices showed the greatest discordance in transferability nor the numeric differences observed; those details were not reported in the provided summary.
The authors documented several practical limits of their findings. Reference intervals for CBC-derived indices substantially overlap between CRC and healthy populations, constraining single-measure screening use in isolation. Incremental prognostic improvement beyond tumour stage was small, which raises questions about clinical utility for risk stratification when staging information is already available. The source did not report specific analytic cutpoints, concordance values, or full variable lists for propensity matching in the abstract, and those granular data would be needed to evaluate thresholds and calibration. Remaining clinical questions include whether serial measurements or composite models combining CBC indices with molecular or imaging markers can improve discrimination, and how transferability issues should be addressed when applying reference intervals across diverse populations—questions the article indicates but does not resolve in the summary provided.