I read with great interest the recent publication by Rajtmajerova and colleagues, in which the authors performed whole-exome sequencing (WES) of 210 formalin-fixed paraffin-embedded (FFPE) tissue samples from 123 patients to characterize the genetic landscape of primary colorectal cancer (CRC) and its paired synchronous and metachronous liver metastases 1. CRC remains the second leading cause of cancer-related mortality globally, and metastatic disease accounts for the overwhelming majority of deaths attributable to this malignancy. By dividing samples into four analytical subgroups—primary tumors giving rise to synchronous metastases (PTsyn), their paired synchronous liver metastases (LMsyn), primary tumors associated with metachronous disease (PTmeta), and metachronous liver metastases (LMmeta)—the authors identified differential mutation frequencies in key driver genes such as APC and TP53, as well as chronicity-specific mutations in genes including VCAN, MPDZ, and FBN1. The study represents the largest paired WES cohort assembled to date for this research question and constitutes a meaningful contribution to understanding the molecular basis of CRC chronicity 1. However, some methodological and conceptual concerns warrant consideration before the findings can be accepted at face value, and I respectfully submit the following remarks in the hope that they may be addressed in a future work. The most pressing statistical concern relates to the reporting of nominal p-values without correction for multiple comparisons. The authors themselves candidly acknowledge this limitation, noting that their approach “is likely to lead to false positive results which need to be kept in mind.” In WES studies, where thousands of genes are simultaneously interrogated across multiple subgroup comparisons, the probability of encountering spurious statistically significant associations by chance alone is substantial. Well-established correction methods—such as the Benjamini–Hochberg false discovery rate procedure or Bonferroni adjustment—are standard practice in genomic analyses precisely because uncorrected p-values in multi-gene contexts are deeply unreliable guides to true biological signal 2. This concern is not merely theoretical in the present study: many of the differentially mutated genes highlighted in Table 2 as chronicity-specific, including SHROOM2, SPEG, GLI2, ZNF84, and PRAMEF15, are mutated in only five to seven samples within their respective groups. When Fisher's exact tests are applied across dozens of gene comparisons simultaneously without multiplicity correction, nominal p-values in the range of 0.01–0.04 carry very limited inferential weight. The authors' decision to present these results without correction should have been more prominently foregrounded in the abstract and results sections rather than confined to the limitations discussion, since readers and clinicians encountering the abstract alone may not be alerted to this critical qualification. We strongly encourage the authors to provide corrected p-values, even as supplementary material, so that readers can form an independent view of which associations are robust to stringent statistical scrutiny. A second concern pertains to the operational definition of metachronous metastasis and the potential for phenotypic misclassification to confound the central comparison of the study. The authors define metachronous metastases as those arising at least 6 months after removal of the primary tumor, while synchronous metastases are defined as those detected earlier. This 6-month threshold, although commonly employed in surgical oncology, is by no means universally standardized, and the clinical and biological literature reflects considerable variability in the cutoffs used across institutions and study populations 3. The temporal boundary between synchronous and metachronous disease may represent a continuous biological spectrum rather than a discrete dichotomy, and tumors classified as metachronous in the present cohort span a wide and unspecified range of intervals beyond 6 months. Patients with metachronous metastases appearing at 7 months post-resection may have fundamentally different molecular biology from those who develop liver recurrences at three or 5 years, yet both groups are pooled within the PTmeta and LMmeta subgroups. This heterogeneity in time-to-metastasis within the metachronous arm introduces uncontrolled biological variability that may dilute or distort the genetic differences being reported. Reboux and colleagues, in a large French population-based registry study, demonstrated that the incidence trajectories and survival outcomes of metachronous liver metastasis have evolved distinctly from those of synchronous disease over recent decades, suggesting these are biologically distinct entities that may further fractionate into temporal subgroups 4. The extent to which the interval to metachronous metastasis was distributed within the present cohort, and whether stratification by time interval would alter the genetic findings, remains unclear from the manuscript and deserves explicit discussion. A third issue concerns the estimation of tumor purity and its downstream impact on copy number variation (CNV) calling and sub-clonal mutation detection. The authors estimated tumor cellularity at a uniform 70% for all primary tumors and metastases, derived from hematoxylin–eosin staining of tissue slides. This approach, while pragmatic, does not capture the substantial inter-sample variability in tumor purity that characterizes real-world surgical specimens, particularly those derived from FFPE archival blocks, which are known to introduce nucleic acid fragmentation, cytosine deamination, and other artefactual changes that can confound variant allele frequency estimation 5. Applying a fixed purity estimate to set CNV thresholds means that samples with true tumor cellularity substantially above or below 70% will be systematically miscalled, potentially inflating or deflating the number of significant CNV events. The authors report a mean of 29 ± 17 CNVs per sample with a range of 3–143, suggesting marked inter-sample variability that is inconsistent with a uniform purity assumption. Furthermore, while the authors report a commendable overall duplicate read rate of 45% and an 83% on-target ratio, the mean coverage at 100× depth was only 19%, meaning that sub-clonal mutations—which are particularly relevant to understanding the evolutionary dynamics of metastatic progression—will have been missed in a substantial proportion of samples 6. The interplay between FFPE-derived DNA quality, variable tumor purity, and sub-clonal detection thresholds creates conditions under which differences in mutation frequencies between subgroups may partly reflect technical rather than biological variability. Computational tools such as FACETS, PURPLE, or similar algorithms that jointly estimate tumor purity and ploidy from WES data would have provided more rigorous and sample-specific foundations for CNV analysis, and the absence of such methods should be acknowledged as a substantive technical limitation rather than a secondary concern. Ayman M. Mustafa: conceptualization, writing – original draft, methodology, validation, visualization, supervision. The author declares no conflicts of interest.
Ayman M Mustafa (Wed,) studied this question.