Review Reports
- Carlo Buonerba 1,*,
- Paolo Montuori 1 and
- Pellegrino Cerino 1
- et al.
Reviewer 1: Onur Kenan Ulutaş Reviewer 2: Anonymous Reviewer 3: Anonymous Reviewer 4: Anonymous
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis manuscript presents a methodologically transparent post-hoc secondary analysis of the SPES biomonitoring cohort, examining associations between 20 serum trace-element channels and self-reported physician-diagnosed tumor history. I want to commend the authors at the outset for the restraint shown in the Discussion and Conclusions: despite Cr-52 emerging as the sole analyte-specific nominal signal, the authors clearly state that this result did not survive FDR correction (q = 0.226) and consistently avoid causal or protective claims throughout. This kind of unembellished, appropriately cautious reporting of a null/exploratory finding is commendable and increasingly rare, and it substantially strengthens the credibility of the work.
The statistical framework is rigorous and well suited to the study design: censoring-aware multiple imputation for values below the limit of quantification, Benjamini–Hochberg FDR correction applied consistently across the primary panel and each sensitivity family, and three complementary sensitivity analyses (extended-covariate adjustment, quantification-state analysis, quantified-only analysis) that triangulate the robustness of the Cr-52 signal. Convergence and singularity diagnostics are transparently reported across all 30 completed datasets.
A few points merit attention before this work is fully ready for publication:
Statistical power: With only 86 reported tumor histories against 4,051 non-cases, the study is markedly underpowered to detect modest associations, particularly in the quantified-only sensitivity analysis (41 cases). The authors should explicitly discuss the implied minimum detectable effect size or power for the primary analysis, so readers can better calibrate the weight given to the Cr-52 finding.
Outcome misclassification: The reliance on a single, broad self-administered "tumor" item (not linked to pathology or registry data) is a substantial limitation that is well-acknowledged in the Discussion. I recommend the authors add one clarifying sentence to the Abstract's Conclusions stating that the estimand is a prevalence odds ratio for a heterogeneous, non-validated self-reported outcome, so that readers who consult only the abstract are not left with an inflated impression of the outcome's specificity.
Reverse causation / temporality: Because serum was collected after or around the time of diagnosis, post-diagnostic treatment, dietary change, or altered metabolism could plausibly explain the observed inverse Cr-52 association. This is discussed, but a brief sensitivity consideration (stratifying by time since diagnosis, using the available diagnosis-age/interval data) would strengthen the causal-inference discussion, if feasible with the available sample size.
Reproducibility: Given that individual-level data cannot be shared for privacy reasons, I encourage the authors to deposit the analytic code (already stated as available upon request) in a public repository (Zenodo, OSF) with a persistent identifier, which would enhance transparency and reproducibility standards expected by Medicina.
Discussion length/focus: The comparison with the Dongfeng–Tongji cohort is well-executed and adds valuable context; however, the paragraph on chromium biology (Cr(III) and insulin signaling) could be tightened, as it partially duplicates points made earlier regarding EFSA's conclusions on chromium's uncertain essentiality.
Overall, this is a well-conducted, honestly reported null/exploratory study that adds a useful, transparently reported data point to the sparse literature on serum trace elements and cancer history. I recommend minor revision to address the points above, particularly clarifying statistical power limitations and strengthening the abstract's framing of the outcome's limitations.
Author Response
Response to Reviewer 1
Manuscript ID: medicina-4510894
Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study
- Statistical power and minimum detectable effect
With only 86 reported tumor histories against 4,051 non-cases, the study is markedly underpowered to detect modest associations, particularly in the quantified-only sensitivity analysis (41 cases). The authors should explicitly discuss the implied minimum detectable effect size or power for the primary analysis, so readers can better calibrate the weight given to the Cr-52 finding.
Response: We thank the Reviewer. The parent SPES sample size was fixed for the biomonitoring program rather than for this post hoc outcome, and the 86 reported histories limit precision, particularly in the measured-data subsets.
We therefore avoided retrospective observed power and calculated an approximate detectable effect from the pooled standard error of the exactly reproduced primary Cr-52 model. Under a two-sided Wald test, 80% power at alpha = 0.05 corresponded to odds ratios of approximately 0.73 or 1.37 per one-standard-deviation increase in log Cr-52. Using alpha = 0.0025 only as a conservative 20-comparison benchmark, the corresponding odds ratios were approximately 0.65 or 1.55; this benchmark is not presented as a fixed Benjamini–Hochberg threshold.
These calculations are conditional on the observed model precision and are not an a priori power analysis. They show that modest associations could have been missed and do not alter the false-discovery-rate conclusion for Cr-52 (q = 0.226). We described the calculation and its assumptions in the Supplementary Statistical Methods and added the resulting limitation to the Discussion.
Changes made: We added the precision-calibrated detectable-effect calculation to the Supplementary Methods and Results and the corresponding limitation to the Discussion. Location in revised manuscript: Supplementary Methods S1.3, p. 2; Supplementary Results S2.3, p. 8; Discussion, p. 14, lines 378–387.
- Outcome misclassification and Abstract framing
The reliance on a single, broad self-administered "tumor" item (not linked to pathology or registry data) is a substantial limitation that is well-acknowledged in the Discussion. I recommend the authors add one clarifying sentence to the Abstract's Conclusions stating that the estimand is a prevalence odds ratio for a heterogeneous, non-validated self-reported outcome, so that readers who consult only the abstract are not left with an inflated impression of the outcome's specificity.
Response: We agree. Although the Abstract already identifies the outcome as self-reported physician-diagnosed tumor history, the conclusion should state the endpoint limitation and estimand directly rather than relying on the full text.
We revised the Abstract to identify the analysis as secondary and cross-sectional, state that the outcome was heterogeneous, self-reported, and not validated through pathology or registry linkage, and avoid any implication that the result concerns confirmed or incident cancer. The Abstract Results describes the complementary analyses as non-independent, and the Conclusions describes the Cr-52 pattern as nominal and exploratory.
Changes made: The Abstract conclusion now reads: “No channel met the FDR criterion. The Cr-52 pattern was nominal and exploratory. These prevalence associations with a heterogeneous, unvalidated self-reported outcome do not establish temporality, causality, chromium speciation, or a protective effect.” Location in revised manuscript: Abstract, p. 1, lines 24–27.
- Reverse causation and time since diagnosis
Because serum was collected after or around the time of diagnosis, post-diagnostic treatment, dietary change, or altered metabolism could plausibly explain the observed inverse Cr-52 association. This is discussed, but a brief sensitivity consideration (stratifying by time since diagnosis, using the available diagnosis-age/interval data) would strengthen the causal-inference discussion, if feasible with the available sample size.
Response: We agree. The four available timing categories contained only 23, 24, 25, and 14 reported histories, so separate stratum-specific models would have been too sparse for stable interpretation.
We therefore evaluated exclusions based on time since diagnosis. After excluding the 23 participants who reported a diagnosis within two years before enrollment, the analysis included 4,114 participants and 63 reported histories. The estimate remained inverse (OR 0.684, 95% CI 0.525–0.892; nominal p = 0.0049); all 30 models converged, none was singular, and no warnings occurred.
The two-year exclusion did not attenuate the inverse estimate, but it does not establish temporality. Longer-term treatment effects, sustained dietary or metabolic change, survivor selection, residual confounding, outcome heterogeneity, and chance remain possible. We report the stable analysis in the Supplementary Materials and added a brief, cautious statement to the Results and Discussion. This targeted analysis is not treated as an independent replication or as a new discovery family.
Changes made: We added the stable two-year diagnosis-exclusion analysis to the Results and Supplementary Results and strengthened the reverse-causation discussion. Location in revised manuscript: Results, p. 11, lines 281–285; Discussion, p. 13, lines 329–337; Supplementary Results S2.3, p. 8.
- Public deposition of the analytical code
Given that individual-level data cannot be shared for privacy reasons, I encourage the authors to deposit the analytic code (already stated as available upon request) in a public repository (Zenodo, OSF) with a persistent identifier, which would enhance transparency and reproducibility standards expected by Medicina.
Response: We agree. The final analysis scripts and public reproducibility materials have now been deposited in a Zenodo software record under the persistent DOI https://doi.org/10.5281/zenodo.22071985. The public record contains the final analysis scripts corresponding to the manuscript results, a public input schema and data dictionary, software and package versions, aggregate expected outputs, numerical acceptance checks, a synthetic smoke-test dataset, and SHA-256 checksums.
Individual-level SPES data and the archived multiple-imputation object are not included in the public deposit because they contain governed health and biomonitoring information. A separate restricted master archive retains the complete data-linked execution record, including input hashes, console logs, fit-level diagnostics, reviewer-analysis outputs, the BMI-corrected extended family, and the alternative-optimizer audit.
The manuscript Data Availability Statement has been updated to cite the Zenodo record and to distinguish the public reproducibility materials from the restricted participant-level data and archived multiple-imputation object.
Changes made: The active Zenodo DOI was added to this response and to the manuscript Data Availability Statement. Location in revised manuscript: Data Availability Statement, p. 16, lines 448–457.
- Discussion length and focus
The comparison with the Dongfeng-Tongji cohort is well-executed and adds valuable context; however, the paragraph on chromium biology (Cr(III) and insulin signaling) could be tightened, as it partially duplicates points made earlier regarding EFSA's conclusions on chromium's uncertain essentiality.
Response: We agree. The mechanistic discussion is longer than warranted by a cross-sectional total-Cr-52 result that did not meet the FDR criterion and for which chromium speciation information was unavailable.
We retained only a focused statement that recent intake, protein binding, renal handling, insulin-related physiology, inflammation, treatment, and other metabolic features may correlate with circulating chromium. Detailed cellular, animal, supplement, and inflammatory-marker examples were removed or compressed. The revised paragraph foregrounds EFSA’s conclusion, randomized-trial evidence, unavailable SPES speciation, and the absence of any basis for a protective or supplementation interpretation.
Changes made: The chromium-biology discussion was substantially shortened and overlapping cautions were consolidated into focused paragraphs on reverse causation, chemical non-specificity, and prospective research needs. Location in revised manuscript: Discussion, p. 13, lines 338–356, and p. 15, lines 408–416.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper is covering an important aspect of the field and the authors indicated the limitations but there are several concerns which should be addressed before publication.
Major issues
1. Due to fact that the main target of the paper is Cr-52, the precision and validity should be clearly introduced. Take into consideration that serum is a carbon rich matrix, so clearly justify that 1:1. dilution was sufficient in order to avoid from high variability of carbon load.
2. The most important limitation and concern of the paper is that LOQ and quality control data are not introduced.
3. When the outcome is not part of imputation model, the imputed values are conditionally independent of the outcome.
4. The main comparison study estimated that Cr association in a mutually adjusted 12 metal model. So the comparison between 2 studies is not symmetric.
5. The cited trial of Masharani et al. indicates a decline of insulin sensitivity after chromium supplementation, which is different statement.
Minor issues
- The title should be changed. I think that biomonitoring study or sample should be included in the title.
- The abstract should be changed, it states that the inverse direction "was observed across complementary analyses", but the Results section explains that these are not independent replications. The abstract should state that Cr-52 was below LOQ in 50% of the participant with secondary history.
- Introduction style should be changed. Hypothesis should be added.
Minor editing is needed.
Author Response
Response to Reviewer 2
Manuscript ID: medicina-4510894
Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study
Major issue 1. Cr-52 analytical precision and the serum matrix
Due to fact that the main target of the paper is Cr-52, the precision and validity should be clearly introduced. Take into consideration that serum is a carbon rich matrix, so clearly justify that 1:1 dilution was sufficient in order to avoid from high variability of carbon load.
Response: We thank the Reviewer. Cr-52 was not preselected as the sole or principal endpoint; the prespecified primary family comprised all 20 analytical channels. Because Cr-52 emerged as the only nominal analyte-specific signal, we agree that its analytical context warrants particularly clear reporting. We also appreciate the opportunity to clarify that the documented SPES procedure used a 1:10 (v/v) dilution, not a 1:1 dilution: a 500 µL serum aliquot was diluted tenfold with a nitric-acid diluent before ICP-MS analysis.
The recently published companion SPES analysis provides a peer-reviewed description of the same serum trace-element workflow: Coppola A, Montuori P, Triassi M, Federico A, Dallio M, Di Duca F, De Rosa E, Buonerba C, Cerino P. Food-Source Patterns and Serum Trace-Metal Biomarkers in the SPES Population Biomonitoring Study, Campania, Italy. J. Xenobiotics 2026, 16, 153. doi:10.3390/jox16040153. The documented procedure used a NexION 350 ICP-MS platform with a concentric nebulizer, cyclonic spray chamber, quartz torch, and 2-mm injector; high-purity argon and helium were recorded as instrument gases; yttrium, rhodium, and lutetium were introduced online as internal standards; element-specific five-point calibration curves were prepared daily and achieved R² > 0.99; each sample was analyzed in duplicate and the mean was retained; chemical blanks were monitored; LOQ estimation was based on the variability of 20 reagent blanks; and accuracy and precision were assessed using internal quality controls and certified reference materials.
Tenfold dilution reduced the serum-matrix burden, but dilution alone cannot establish complete removal of potential carbon-derived interference at m/z 52. The frozen secondary-analysis archive did not include the exact collision/reaction-cell settings or Cr-specific numerical recovery and precision summaries required for a new retrospective interference assessment. We therefore reported the documented analytical safeguards without claiming that dilution alone eliminated all mass-52 interference.
Separately, chromium speciation information was unavailable in the archived analytical dataset used for this secondary analysis. The biomarker is therefore reported as the total serum Cr-52 analytical channel and cannot be assigned to Cr(III), Cr(VI), or another chromium species. We did not make a species-specific interpretation of the observed association.
Changes made: We retained and made more prominent the documented 1:10 dilution in the main Methods, expanded the Supplementary Methods to report the analytical safeguards, cited the newly published companion SPES paper, and stated clearly that the available secondary-analysis archive does not permit a new Cr-specific matrix-interference validation. Chromium speciation is described as unavailable, and the Cr-52 result remains explicitly exploratory and hypothesis-generating. Location in revised manuscript: Methods Section 2.2, p. 5, lines 124–147; Supplementary Methods S1.2, pp. 1–2; Discussion, p. 14, lines 388–399.
Major issue 2. Limits of quantification and quality-control information
The most important limitation and concern of the paper is that LOQ and quality control data are not introduced.
Response: We agree. The submitted Supplement reported assay completion and below-LOQ frequencies, but the actual archived LOQs and the complete assay-support structure required clearer consolidation.
We added Supplementary Table S2B for all 20 analytical channels.
For Cr-52, 4,134 assays were performed, 1,691 results were below the LOQ (40.9% of performed assays), 2,443 were quantified, no result carried an HI20 flag, and three unperformed assays were eligible for completion under the parent low-nonperformance rule. The primary Cr-52 model therefore included 4,137 participants.
Table S2B is intended to provide the complete archived LOQ and assay-support structure; it is not presented as a substitute for analyte-specific recovery or precision data. The published SPES methods report the use of internal quality controls and certified reference materials, but numerical Cr-specific recovery and within-run or between-run precision summaries were unavailable in the secondary-analysis archive and therefore cannot be supplied retrospectively.
We also replaced wording that could be read as an instrument- or batch-QC failure. HI20 was a prespecified post-analytical high-value plausibility flag, not a failed calibration, failed duplicate, or failed analytical batch. HI20 source values were set to missing and were not retained as observed concentrations or used for model fitting or donor selection; eligible cells were completed under the archived missing-data workflow. Archived LOQ values were constant within channel and were the thresholds used by the frozen workflow; they were not newly derived for this revision. Separate instrument-level LODs were unavailable and were not inferred from the archived LOQs.
Changes made: We revised Supplementary Methods S1.2, designated the assay-completion table as Table S2A, added Table S2B without changing the numbering of subsequent tables, and added a concise main-text cross-reference. The newly published Journal of Xenobiotics companion paper and its Supplementary Methods are cited as the current peer-reviewed methodological source. The response and table notes distinguish archived assay-support metadata from numerical analyte-specific recovery and precision information. Location in revised manuscript: Methods Section 2.2, p. 5, lines 132–139; Results Section 3.1, p. 8, lines 233–234; Supplementary Methods S1.2, pp. 1–2; Supplementary Table S2B, p. 4.
Major issue 3. Outcome information in the imputation models
When the outcome is not part of imputation model, the imputed values are conditionally independent of the outcome.
Response: We thank the Reviewer for raising this important multiple-imputation principle. We agree that omission of an analysis outcome from an exposure-imputation model can create an uncongenial imputation-analysis relationship. We therefore re-audited the exact parent imputation script, the archived 30-imputation object used for the submitted analysis, and the target-specific predictor-selection records.
Although the present secondary analysis was specified after the parent imputation run, the original standardized questionnaire item used to define tumor history was already present in the parent dataset as a non-imputed auxiliary variable. The questionnaire item remained unchanged and was not itself imputed.
For Cr-52, the archived target-specific audit confirmed that the tumor-history indicator was retained after predictor selection in both relevant components: the censored Tobit model for below-LOQ observations and the predictive-mean-matching component for eligible non-censored missing values. The completed Cr-52 concentrations were therefore not generated independently of tumor-history information.
Retention was target-specific rather than universal. For channel-specific models in which the item was not retained, the imputed cells—not the observed quantified concentrations—were generated without direct conditioning on tumor-history status. This creates potential imputation–analysis uncongeniality and could attenuate or otherwise bias an association under some missing-data structures. The revised Supplementary Methods state this limitation explicitly. The parent geospatial paper and its Web Appendix document the responder-only, censoring-aware imputation framework, and the published Journal of Xenobiotics companion analysis independently reports reuse of the same archived 30-imputation object. The 11 channels with at least 10% left-censoring were also examined in quantification-state and quantified-only analyses based on observed assay information; these analyses reduce, but do not eliminate, the concern and do not constitute independent replication. The remaining channels had less than 10% left-censoring and no more than 5% eligible assay nonperformance by design. Unperformed Hg-202 and Ni-60 assays were not imputed.
Because tumor history was already included in both Cr-52 imputation components, we did not regenerate the entire parent imputation resource solely for this revision. The Cr-52 quantification-state analysis (OR 0.616, 95% CI 0.401–0.948) and quantified-only analysis (OR 0.375, 95% CI 0.154–0.912) additionally remained inverse without using completed sub-LOQ concentrations; neither met its own false-discovery-rate threshold, and neither constitutes an independent replication.
Changes made: The Methods now distinguish the later specification of this secondary analysis from the prior availability of the questionnaire item during imputation. The Supplementary Methods document the Cr-52 predictor audit, clarify that the questionnaire outcome was not imputed, describe the target-specific nature of predictor retention, and state the potential imputation–analysis uncongeniality for channels in which the item was not retained. Location in revised manuscript: Methods Section 2.3, p. 6, lines 164–181; Supplementary Methods S1.2–S1.3, pp. 1–2; Discussion, p. 14, lines 388–399.
Major issue 4. Asymmetric comparison with the Dongfeng–Tongji cohort
The main comparison study estimated that Cr association in a mutually adjusted 12 metal model. So the comparison between 2 studies is not symmetric.
Response: We agree completely. The two studies do not estimate the same quantity and should not be presented as direct replications.
The Dongfeng–Tongji analysis followed 4,573 older adults with type 2 diabetes who were cancer-free at baseline, identified 541 medically confirmed incident cancers during a median 10.2 years, measured fasting plasma metals before diagnosis, and estimated chromium in a Cox model mutually adjusted for the other 11 metals (HR per 1-SD higher log chromium 0.88, 95% CI 0.82–0.94). SPES analyzed current total serum Cr-52 in a younger population-based biomonitoring sample, used a heterogeneous self-reported prevalent tumor-history outcome, collected serum after or around the reported diagnosis, and fitted separate channel-specific logistic models rather than a mutually adjusted metal-mixture model.
We therefore use the term directional concordance only. The revised text states that the shared direction provides context for hypothesis generation, not replication, and that the broader metal profiles were not concordant. It also avoids implying that the SPES single-channel coefficient has the same conditional interpretation as the Dongfeng–Tongji mutually adjusted chromium coefficient.
Changes made: The comparison paragraph was shortened and revised so that the differences in population, matrix, timing, outcome validation, model estimand, and multimetal adjustment are explicit before the shared chromium direction is discussed. Location in revised manuscript: Discussion, p. 12, lines 319–328, and p. 13, lines 357–364.
Major issue 5. Interpretation of the Masharani trial
The cited trial of Masharani et al. indicates a decline of insulin sensitivity after chromium supplementation, which is different statement.
Response: We agree and thank the Reviewer for identifying this imprecision. In the randomized trial, chromium picolinate did not improve insulin sensitivity compared with placebo (the primary between-group comparison was not significant), and higher achieved serum chromium among supplemented participants was associated with worsening insulin sensitivity. Describing the trial only as showing "no improvement" omitted the latter finding.
This correction strengthens rather than weakens the need for restraint: the trial does not support a beneficial metabolic interpretation of higher circulating chromium and provides no basis for a cancer-prevention or supplementation claim.
Changes made: The sentence now states: “In randomized trials, chromium picolinate did not improve insulin sensitivity relative to placebo in healthy adults, and chromium treatment produced no glycemic benefit in poorly controlled, insulin-treated type 2 diabetes; within the chromium-treated group in Masharani et al., higher achieved serum chromium was associated with worsening insulin sensitivity.” Location in revised manuscript: Discussion, p. 13, lines 344–348.
Minor issue 1. Title
The title should be changed. I think that biomonitoring study or sample should be included in the title.
Response: We agree that both the parent-study context and the secondary cross-sectional design should be immediately visible. The submitted title already includes "population-based biomonitoring cohort," but "cohort" may be read as implying longitudinal outcome ascertainment.
We therefore propose the more explicit title: "Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study".
Changes made: The title was changed consistently in the manuscript and Supplementary Materials to: “Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study”. Location in revised manuscript: Title, p. 1, lines 2–4; Supplementary Materials title, p. 1.
Minor issue 2. Abstract wording and below-LOQ frequency
The abstract should be changed, it states that the inverse direction "was observed across complementary analyses", but the Results section explains that these are not independent replications. The abstract should state that Cr-52 was below LOQ in 50% of the participant with secondary history.
Response: We agree. The revised Abstract explicitly describes the analyses as non-independent and reports the observed Cr-52 quantification frequencies.
Among participants with an assay performed, Cr-52 was below the LOQ in 45 of 86 participants reporting tumor history (52.3%) and in 1,646 of 4,048 participants reporting no history (40.7%). These counts describe assay state; they do not themselves establish a concentration gradient or a causal effect.
Changes made: The Abstract Results now reports 45/86 (52.3%) versus 1,646/4,048 (40.7%) below the LOQ and uses the phrase “non-independent complementary analyses.” The Abstract conclusion uses the same cautious wording provided to Reviewers 1 and 4. Location in revised manuscript: Abstract, p. 1, lines 19–27.
Minor issue 3. Introduction style and research hypothesis
Introduction style should be changed. Hypothesis should be added.
Response: We agree. The Introduction was made more direct by separating established carcinogenicity of defined chemical forms and exposure settings from the narrower question addressed by a current total-serum biomarker. The knowledge gap and the hypothesis are now explicit.
The final Introduction states: “We hypothesized, without specifying a direction or causal effect, that current serum trace-element distributions might differ according to self-reported physician-diagnosed tumor history.” The objective is aligned with the actual estimand: prevalence associations across all 20 serum analytical channels with FDR control, rather than incident cancer risk or tumor development.
Changes made: The final Introduction paragraph, Abstract, Discussion, and Conclusion were revised so that the design, non-directional hypothesis, outcome, and estimand remain consistent. Location in revised manuscript: Introduction, pp. 2–3, lines 31–72; Abstract, p. 1, lines 12–27; Discussion, pp. 12–15, lines 309–416; Conclusions, p. 15, lines 417–425.
Quality of English language
Minor editing is needed.
Response: We agree. The complete manuscript, Supplementary Materials, figure captions, and reviewer responses underwent a final language and terminology edit. Particular attention was given to article use, sentence length, the distinction between tumor history and confirmed cancer, and consistent use of exploratory, nominal, and FDR-controlled terminology.
Changes made: A final author-level language and terminology review was completed against the definitive scientific text. Location in revised manuscript: Entire main manuscript and Supplementary Materials; final page and line locators are provided throughout these responses.
Reviewer 3 Report
Comments and Suggestions for AuthorsI understand what this paper aims to convey. The authors transparently reused existing SPES cohort data to explore associations between 20 serum trace-element analytical channels and self-reported tumor history, reporting that only Cr-52 emerged as an analyte-specific signal under an FDR-controlled multiple-comparison framework.
The authors explicitly and repeatedly stated that this signal does not imply a causal or protective effect, and candidly acknowledged the fundamental limitations of interpretation (inability to distinguish chemical species, post-diagnosis serum collection, heterogeneity of the self-reported outcome variable, failure to meet the FDR threshold, etc.). This self-restraint and transparency should be recognized as scientific integrity.
However, I believe the authors' responses to the fundamental questions below will be decisive in determining whether this paper is suitable for publication.
Q1. What is the added value of this paper?
The core finding of this paper is that none of the 20 channels passed the FDR threshold. The sole signal, Cr-52, yielded q = 0.226, falling 4.5-fold short of the FDR criterion (q < 0.05). The consistent inverse direction observed across four complementary analyses demonstrates internal consistency, but all four analyses derive from the same 86 participants and therefore do not constitute independent validation.
In this context, the most important question is as follows:
Why should this paper be published? If no result passed the FDR threshold, interpretation is impossible with total chromium alone, and the authors have no plans or capacity for further experiments, what contribution does this paper make to the scientific literature?
In particular, the Dongfeng–Tongji cohort has already reported an inverse chromium association with stronger evidence from a prospective design, observing 541 medically confirmed incident cancers among 4,573 adults with type 2 diabetes.
The inverse direction for chromium persisted in the mutually adjusted 12-metal model, and consistent dose–response trends were observed across quartiles. Given that stronger evidence already exists, what added value does the Cr-52 nominal signal in this paper provide to the scientific literature? I acknowledge the concordance in direction, but please clarify whether that alone constitutes sufficient justification for publication.
Q2. What can be said with total chromium alone?
The biological effects of chromium are determined entirely by its oxidation state. Cr(VI) is an IARC Group 1 carcinogen with established associations to lung cancer via inhalation, whereas Cr(III) has uncertain essentiality and no evidence of carcinogenicity. Attempting to interpret total chromium measurements when the two species have opposite biological effects represents a fundamental limitation.
It is a reasonable assumption that chromium in serum exists predominantly as Cr(III)-protein complexes (particularly transferrin-bound chromium). When Cr(VI) enters the bloodstream, it is rapidly reduced to Cr(III) by erythrocytes and trapped intracellularly, so chromium remaining in serum is expected to be primarily Cr(III). However, this paper does not discuss this expected species composition.
Accordingly, I ask the following:
-
Why was the expected species composition of serum chromium not discussed? Please explicitly address the fact that serum chromium is predominantly Cr(III) and explain the implications this holds for interpreting the inverse association.
-
Given that only total chromium was measured, can the inverse association be attributed to a (hypothetical) protective effect of Cr(III), a metabolic environment with low Cr(VI) exposure, or an entirely different metabolic correlate? If these cannot be distinguished, what are the authors' views on the chemical interpretability of this result?
-
Even assuming serum chromium is predominantly Cr(III), evidence that Cr(III) is an effective factor in cancer prevention is currently very limited. EFSA does not recognize an essential metabolic function for Cr(III), and randomized trials have shown no effect in healthy non-diabetic adults. There are even concerns that Cr(III) supplements may generate carcinogenic Cr(VI) through reactions with biological oxidants. Given this literature context, do you consider a Cr(III)-based protective-effect interpretation to be plausible?
Author Response
Response to Reviewer 3
Manuscript ID: medicina-4510894
Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study
Q1. Added value of the study
Why should this paper be published? If no result passed the FDR threshold, interpretation is impossible with total chromium alone, and the authors have no plans or capacity for further experiments, what contribution does this paper make to the scientific literature? In particular, the Dongfeng-Tongji cohort has already reported stronger prospective evidence. Please clarify whether directional concordance alone constitutes sufficient justification for publication.
Response: We appreciate this central question. We agree that directional concordance with Dongfeng–Tongji is not, by itself, sufficient justification for publication, and the revised manuscript does not present it as such.
The first contribution is complete, FDR-controlled reporting of all 20 serum analytical channels in a population-based biomonitoring sample from the general population, including the predominantly null results. Reporting the full panel rather than only the nominal association reduces selective-reporting bias and provides an interpretable account of what this dataset does and does not support.
Second, SPES addresses a different epidemiological question from Dongfeng–Tongji. Dongfeng–Tongji studied prediagnostic fasting plasma metals and medically confirmed incident cancer in older adults with type 2 diabetes using a mutually adjusted 12-metal Cox model. SPES studied current total serum channels and a broad self-reported prevalent tumor-history outcome in a younger population-based sample using separate channel-specific prevalence models. These differences make the SPES evidence weaker for causality but distinct rather than redundant.
Prospective findings outside Dongfeng–Tongji are not uniform. EPIC-Florence reported no chromium association in 150 prediagnostic breast-cancer cases and 150 matched controls, and a small 2025 exploratory EPIC-Florence analysis of 50 cases each of lung, bladder, and prostate cancer with matched controls found no significant case–control differences in the measured serum-metal panel. These studies do not resolve the question, but they show that prospective findings are not uniformly concordant.
Third, the Cr-52 pattern was examined through the full censoring-aware distribution, observed quantification state, and quantified values. These non-independent analyses do not validate one another, but they show that the direction is not solely a consequence of assigning completed sub-LOQ values. The study also quantifies what cannot be concluded from a total-serum, post-diagnosis, heterogeneous outcome and specifies the design needed for a decisive follow-up: prediagnostic specimens, validated registry-linked outcomes, species-appropriate matrices, and prespecified multimetal inference.
The published SPES geospatial analysis established pronounced small-area heterogeneity across the 20 serum channels, and the published food-source analysis identified individual dietary correlates in the same responder population. The present paper adds a distinct participant-level health-history layer to that exposure and source context, while explicitly acknowledging that the questionnaire endpoint is not confirmed cancer.
Finally, public-health relevance does not substitute for statistical evidence. It does, however, increase the value of transparent reporting in a region marked by long-standing community concern about environmental contamination and in which SPES was embedded in direct engagement with participating communities. In that setting, a complete, predominantly null participant-level analysis helps distinguish observed biomonitoring patterns from individual-level causal interpretations that this cross-sectional design cannot support. The revised manuscript therefore presents the complete, FDR-controlled characterization of all 20 analytical channels—and the limits of that evidence—rather than the nominal Cr-52 p value, as the principal justification for publication.
Changes made: The knowledge-gap paragraph and the opening and closing Discussion paragraphs were strengthened. The revised text states explicitly that directional concordance is contextual, that no association met FDR control, and that the study contributes a transparent population-based, predominantly null, hypothesis-generating result rather than evidence of a protective chromium effect. Location in revised manuscript: Introduction, pp. 2–3, lines 52–72; Discussion, p. 12, lines 309–328, and pp. 14–15, lines 400–416; Conclusions, p. 15, lines 417–425.
Q2a. Expected species composition of serum chromium
Why was the expected species composition of serum chromium not discussed? Please explicitly address the fact that serum chromium is predominantly Cr(III) and explain the implications this holds for interpreting the inverse association.
Response: We agree that oxidation state and protein binding are central to chromium toxicology. Chromium speciation information was unavailable in the archived analytical dataset used for this secondary analysis; the measured variable was the total serum Cr-52 analytical channel.
Published human-serum speciation work identifies chromium–transferrin, a trivalent complex, as the main detectable chromium species in human serum at physiological concentration levels. This context was added. However, the archived SPES dataset contained no speciation information, so the published finding cannot quantify species proportions in SPES, exclude Cr(VI)-related exposure history, or assign the total Cr-52 association to Cr(III).
The revised manuscript therefore acknowledges the protein-bound Cr(III) context while treating unavailable speciation as a fundamental limitation. The implication is not that the association is a Cr(III) effect, but that species-specific interpretation is impossible with the archived total-serum channel.
Changes made: One concise sentence on chromium–transferrin was added, followed immediately by the statement that speciation information was unavailable and that the total Cr-52 channel cannot identify the contributing species or their proportions. Location in revised manuscript: Methods Section 2.2, p. 5, lines 145–147; Discussion, p. 13, lines 349–356.
Q2b. Chemical interpretability of the inverse association
Given that only total chromium was measured, can the inverse association be attributed to a hypothetical protective effect of Cr(III), a metabolic environment with low Cr(VI) exposure, or an entirely different metabolic correlate? If these cannot be distinguished, what are the authors' views on the chemical interpretability of this result?
Response:
We agree that the inverse Cr-52 association has limited chemical interpretability because the assay quantified total serum chromium and did not distinguish chromium species. Accordingly, the observed association cannot discriminate among Cr(III)-related metabolic processes, lower Cr(VI) exposure, or other biological or exposure-related determinants, and it cannot be assigned to any specific chromium mechanism. We therefore interpret Cr-52 only as an assay-level measure of total serum chromium and the inverse estimate as a nominal, hypothesis-generating association. It should not be interpreted as evidence of a protective effect of chromium or as an indicator of a particular chromium species, source, or exposure pathway.
Changes made: Language that could be read as attributing the association to Cr(III) biology was replaced by an explicit statement that total serum chromium cannot resolve chromium species, source, exposure pathway, or mechanism. The Cr-52 estimate is now described as nominal and hypothesis-generating. Location in revised manuscript: Discussion, p. 13, lines 338–356, and p. 14, lines 367–399.
Q2c. Plausibility of a Cr(III)-based protective interpretation
Even assuming serum chromium is predominantly Cr(III), evidence that Cr(III) is an effective factor in cancer prevention is currently very limited. EFSA does not recognize an essential metabolic function for Cr(III), randomized trials have shown no effect in healthy non-diabetic adults, and supplement chemistry raises additional concerns. Do you consider a Cr(III)-based protective-effect interpretation to be plausible?
Response: We do not consider a Cr(III)-based protective interpretation to be supported or defensible from the present data. A total-serum, cross-sectional, post-diagnosis association that did not meet the FDR criterion cannot establish a protective nutrient effect.
EFSA concluded that an essential metabolic function of Cr(III) in humans had not been substantiated and found no evidence of beneficial effects of chromium intake in healthy subjects. In the Masharani randomized trial, chromium picolinate did not improve insulin sensitivity relative to placebo; within the chromium-treated group, higher achieved serum chromium was associated with worsening insulin sensitivity. These observations argue against using the SPES result to endorse nutritional deficiency, supplementation, or cancer prevention.
The literature cited by the Reviewer concerning possible oxidative conversion of some Cr(III) supplement complexes provides an additional reason not to advance a simple protective interpretation; SPES did not assess supplement use or chromium species and cannot test that pathway. The mechanistic discussion was shortened and now states directly that the data do not support chromium supplementation, a protective Cr(III) effect, or a low-Cr(VI) interpretation.
Changes made: The revised Discussion and Conclusions reject protective, deficiency, supplementation, and species-specific interpretations and retain only a brief hypothesis-generating discussion of metabolic correlates. Location in revised manuscript: Discussion, p. 13, lines 338–356; Conclusions, p. 15, lines 417–425.
Reviewer 4 Report
Comments and Suggestions for AuthorsThe manuscript entitled “Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Cross-Sectional Analysis of a Population-Based Biomonitoring Cohort” addresses a relevant topic in environmental health and human biomonitoring and presents scientific merit, particularly due to the use of a population-based dataset and a detailed statistical approach. However, some adjustments are necessary to improve the methodological clarity, statistical interpretation, and discussion of the findings before the manuscript can be considered for publication. Below are the reviewer's comments regarding the adjustments that need to be made to the manuscript. I kindly ask that all changes made to the manuscript be highlighted in a different color in the text or highlighted in yellow.
Abstract:
1- Please make the study design clearer in the abstract.
2- Please clarify that the outcome was self-reported physician-diagnosed tumor history, rather than confirmed cancer or incident cancer.
3- In the results, please report the total number of participants and the number reporting tumor history (4,137 and 86, respectively), together with the main effect estimates.
4- Please make the conclusion more cautious and consistent with the cross-sectional design. The results should be described as exploratory/hypothesis-generating rather than as evidence that chromium influences tumor development.
Introduction:
5- Please strengthen the epidemiological background regarding tumors/cancer and environmental exposure to trace elements, providing quantitative information and a clearer justification for the relevance of this topic.
6- Please make the knowledge gap more explicit. What is still unknown regarding circulating trace elements and tumor history, and what specifically does the present study add to the existing literature?
7- At the end of the introduction, please state the research hypothesis and align the objective precisely with the outcome evaluated. Avoid wording that implies a causal relationship between exposure and tumor development.
Materials and Methods:
8- Since this is a secondary cross-sectional analysis, please provide a clearer description of the original SPES study and explain how the current analytical population was obtained. I also recommend following the STROBE recommendations and providing the corresponding checklist as supplementary material.
9- Please describe in greater detail how tumor history was assessed. Clarify whether the information was self-reported, whether it was confirmed through medical records or cancer registries, and whether benign and malignant tumors were combined into the same outcome.
10- The outcome is highly heterogeneous, including tumors from different anatomical sites and potentially different biological behaviors. Please provide a stronger justification for combining these conditions into a single outcome and discuss the implications of this decision.
11- Please clarify the temporal relationship between tumor diagnosis and blood collection. Were serum samples collected before diagnosis, after diagnosis, or at variable times relative to diagnosis? This information is essential because it directly affects the possibility of reverse causation.
12- Please provide a clearer rationale for the selection of covariates included in the primary and extended models. Explain whether these variables were selected a priori based on previous literature or an epidemiological framework.
13- Please provide more details regarding the treatment of values below the LOQ. For Cr-52, a substantial proportion of participants were below the LOQ, and this may have an important influence on the estimated associations.
14- Please clarify the multiple-imputation procedure. Since the 30 imputed datasets were originally generated for previous analyses, please state whether the tumor-history outcome was incorporated into the imputation models and discuss the implications if it was not.
15- Please clarify how the log-transformed concentrations were standardized and whether the same standardization procedure was applied across the 30 imputed datasets.
16- Please provide additional information regarding model convergence and the treatment of singular fits. For the analytes in which the geographic random-effect variance was essentially zero, consider presenting a sensitivity analysis using an appropriate model without the random effect.
17- Please explain more clearly the multiple-testing strategy. The primary analysis included 20 analytical channels, whereas additional sensitivity analyses were also performed. Please clearly distinguish primary analyses from exploratory analyses.
Results:
18- Please begin the results by describing the participant selection process, including the initial sample, exclusions, and final analytical population. A flowchart would make this information more didactic.
19- In Table 1, please provide a complete description of the participants, including the main demographic and clinical characteristics, and clearly indicate the number of participants with and without tumor history.
20- Please review the presentation of continuous variables and ensure that the descriptive statistics correspond to their distributions. Non-normally distributed variables should preferably be presented as median and interquartile range.
21- Please ensure that the sample size for each analytical channel is clearly presented, particularly because the number of available participants varies between elements and is substantially smaller for some sensitivity analyses.
22- Please present the primary regression results in a more didactic manner. A forest plot showing ORs and 95% CIs for the 20 analytical channels could considerably improve the presentation.
23- Please revise the interpretation of the Cr-52 finding. The OR of 0.751 (95% CI 0.601–0.937; p=0.011) is nominally significant, but the FDR-adjusted q-value is 0.226. Therefore, this finding should be clearly described as exploratory rather than statistically significant after correction.
24- The sensitivity analyses for Cr-52 and Ni-60 should be interpreted cautiously. In particular, the quantified-only analyses considerably reduce the number of participants with tumor history. Please provide the exact sample size and number of cases for each analysis and clarify that these analyses do not represent independent replication.
Discussion:
25- The discussion should emphasize that the study cannot establish temporality or causality. Because serum measurements were obtained after or around the tumor diagnosis, reverse causation is a plausible explanation that should be discussed more prominently.
26- Please moderate the biological interpretation of the Cr-52 finding. The proposed mechanisms are interesting, but they should be presented as possible explanations or hypotheses rather than mechanisms demonstrated by the present study. The discussion should also emphasize that total serum chromium does not identify the specific chromium species responsible for the measured concentration.
27- Please expand the limitations section to include the cross-sectional design, self-reported and heterogeneous tumor outcome, relatively small number of tumor cases, multiple comparisons, substantial values below the LOQ for some elements, potential selection in quantified-only analyses, and residual confounding.
28- Please revise the final interpretation of the study so that it is consistent with the statistical results. The main contribution of the study is the identification of an exploratory pattern that may justify further investigation, preferably through prospective studies using prediagnostic samples and validated cancer outcomes. The conclusion should not suggest that chromium protects against or causes tumors.
Author Response
Response to Reviewer 4
Manuscript ID: medicina-4510894
Serum Trace Elements and Self-Reported Physician-Diagnosed Tumor History: A Secondary Cross-Sectional Analysis of the Population-Based SPES Biomonitoring Study
As requested, all revisions are highlighted in yellow in the accompanying marked manuscript and Supplementary Materials; clean versions are also provided.
Comment 1. Study design in the Abstract
Please make the study design clearer in the abstract.
Response: We agree. The Abstract identifies the work at its first substantive sentence as a secondary cross-sectional analysis of archived SPES biomonitoring data. This wording distinguishes the present analysis from the prospectively implemented parent program and from a longitudinal cancer-incidence study.
Changes made: The Abstract Background and Methods now identify the work as a secondary cross-sectional analysis and retain the fixed parent-study sample derivation. Location in revised manuscript: Abstract, p. 1, lines 12–18.
Comment 2. Meaning of the outcome in the Abstract
Please clarify that the outcome was self-reported physician-diagnosed tumor history, rather than confirmed cancer or incident cancer.
Response: We agree. The revised Abstract states that the endpoint was a heterogeneous self-reported history of a physician-diagnosed tumor, not registry-confirmed cancer, pathology-confirmed malignant neoplasia, or incident cancer. The outcome was not externally validated, and the analysis estimates prevalence odds ratios.
Changes made: This clarification appears in both the Abstract Methods and Conclusions. Location in revised manuscript: Abstract, p. 1, lines 12–16 and 24–27.
Comment 3. Participant counts and main estimate in the Abstract
In the results, please report the total number of participants and the number reporting tumor history (4,137 and 86, respectively), together with the main effect estimates.
Response: We agree. The revised Abstract presents the analytical population, outcome-positive count, nominal p value, and FDR-adjusted q value together so that they cannot be separated from the null panel-level conclusion.
The Abstract also reports the observed Cr-52 below-LOQ frequencies: 45/86 (52.3%) among participants reporting a history and 1,646/4,048 (40.7%) among those reporting no history with an assay performed.
Changes made: The Abstract Results reports N = 4,137, 86 reported histories, OR 0.751 (95% CI 0.601–0.937), p = 0.011, q = 0.226, and the below-LOQ counts. Location in revised manuscript: Abstract, p. 1, lines 15–23.
Comment 4. Cautious Abstract conclusion
Please make the conclusion more cautious and consistent with the cross-sectional design. The results should be described as exploratory/hypothesis-generating rather than as evidence that chromium influences tumor development.
Response: We agree. The Abstract does not state or imply that chromium influences tumor development. It identifies the pattern as nominal and exploratory, states that the complementary analyses are non-independent, and makes the outcome and design limitations explicit.
Changes made: The Abstract conclusion now reads: “No channel met the FDR criterion. The Cr-52 pattern was nominal and exploratory. These prevalence associations with a heterogeneous, unvalidated self-reported outcome do not establish temporality, causality, chromium speciation, or a protective effect.” Location in revised manuscript: Abstract, p. 1, lines 24–27.
Comment 5. Epidemiological background and quantitative context
Please strengthen the epidemiological background regarding tumors/cancer and environmental exposure to trace elements, providing quantitative information and a clearer justification for the relevance of this topic.
Response: We agree. The Introduction was strengthened with concise quantitative context rather than expanded into a general review of pollution and cancer.
The revised text distinguishes established carcinogenicity of defined chemical forms and exposure settings from circulating total-element biomarkers. It notes that the principal prospective comparison included 4,573 adults with type 2 diabetes and 541 incident cancers over a median 10.2 years, whereas EPIC-Florence reported no association between prediagnostic serum chromium and breast-cancer incidence in 150 cases and 150 matched controls, with more than half of chromium measurements below the quantification limit. A small 2025 exploratory EPIC-Florence analysis of 50 cases each of lung, bladder, and prostate cancer with matched controls likewise found no significant case–control differences in the measured serum-metal panel. It also states that the published SPES serum-metals analysis included 4,188 participants and 20 analytical channels and documented pronounced small-area heterogeneity in current concentrations.
These numbers justify the question while also showing why the available evidence is sparse, population-specific, matrix-specific, and not sufficient for causal inference.
Changes made: The first two Introduction paragraphs now retain quantitative information directly relevant to circulating trace elements, analytical censoring, and cancer outcomes. Location in revised manuscript: Introduction, p. 2, lines 31–51.
Comment 6. Knowledge gap and added value
Please make the knowledge gap more explicit. What is still unknown regarding circulating trace elements and tumor history, and what specifically does the present study add to the existing literature?
Response: We agree. The gap is not whether defined occupational exposures to carcinogenic metal species can cause cancer; that is established for specific agents and routes. The unresolved question is whether current circulating total-element biomarkers differ according to tumor history in a general-population biomonitoring cohort, particularly when the outcome is historical, heterogeneous, and collected after diagnosis and when several channels are left-censored.
The present study adds a complete 20-channel, FDR-controlled participant-level analysis in SPES; transparent reporting of the predominantly null panel; explicit handling of below-LOQ measurements; observed-assay-state and quantified-only sensitivity views; and a precise account of what total serum Cr-52 cannot establish. It does not add evidence of causality or a protective chromium mechanism.
The revised manuscript states that directional agreement with Dongfeng–Tongji is contextual rather than sufficient justification by itself. The primary contribution is a transparent population-based, predominantly null, hypothesis-generating result that defines the requirements for a prospective registry-linked study.
Changes made: An explicit knowledge-gap paragraph was added before the final Introduction paragraph and aligned with the added-value response to Reviewer 3. Location in revised manuscript: Introduction, p. 2, lines 52–65; Discussion, p. 12, lines 309–318, and pp. 14–15, lines 400–416.
Comment 7. Research hypothesis and objective
At the end of the introduction, please state the research hypothesis and align the objective precisely with the outcome evaluated. Avoid wording that implies a causal relationship between exposure and tumor development.
Response: We agree. The hypothesis is non-directional and refers to current serum distributions and self-reported history, not tumor development.
The final sentence states: “We hypothesized, without specifying a direction or causal effect, that current serum trace-element distributions might differ according to self-reported physician-diagnosed tumor history.” The objective states that we estimated prevalence associations across the 20 analytical channels with FDR control and prespecified sensitivity analyses. The manuscript states explicitly that the analysis did not test incident cancer risk, historical etiologic exposure, or causal effects.
Changes made: The final Introduction paragraph was replaced with aligned hypothesis and objective language. Location in revised manuscript: Introduction, pp. 2–3, lines 66–72.
Comment 8. Original SPES study and derivation of the analytical population
Since this is a secondary cross-sectional analysis, please provide a clearer description of the original SPES study and explain how the current analytical population was obtained. I also recommend following the STROBE recommendations and providing the corresponding checklist as supplementary material.
Response: We agree. We made the relationship between the parent SPES programme and the present secondary analysis more explicit. SPES was a population-based human-biomonitoring programme conducted in Campania in 2016–2017, with participants sampled from municipal registries across predefined geographic clusters. The parent programme enrolled 4,205 participants; 17 participants residing outside the study inclusion area were excluded, leaving the frozen 4,188-record analytical dataset. Of these, 4,137 had a usable response to the standardized tumor-history item and entered the present outcome models, while 51 with a blank response were excluded because the outcome was not imputed.
We cited the parent SPES design publication, the peer-reviewed serum-metals analysis, and the recently published companion analysis by Coppola et al. in Journal of Xenobiotics, which used the same 4,137 questionnaire responders and provides additional methodological detail. Figure 1 already shows the 4,205-to-4,188-to-4,137 participant flow; the corresponding text was reinforced in the Methods and at the beginning of the Results.
Changes made: We expanded Section 2.1 and the first Results paragraph. Completed STROBE and STROBE-ME checklists are supplied as separate reporting documents. Location in revised manuscript: Methods Section 2.1, p. 3, lines 74–93; Figure 1, p. 4, lines 95–97; Results Section 3.1, p. 8, lines 227–233; completed STROBE and STROBE-ME checklists supplied separately.
Comment 9. Ascertainment and meaning of tumor history
Please describe in greater detail how tumor history was assessed. Clarify whether the information was self-reported, whether it was confirmed through medical records or cancer registries, and whether benign and malignant tumors were combined into the same outcome.
Response: We agree. The endpoint came from a standardized self-administered lifestyle-questionnaire item asking whether a physician had previously diagnosed a tumor. When a history was reported, age at diagnosis and body site were recorded for descriptive purposes. The response was not verified through medical records, pathology reports, or cancer-registry linkage, and the outcome itself was not imputed.
The source item used the broad Italian term for tumor and did not reliably distinguish malignant, benign, or pre-invasive diagnoses. We therefore defined the endpoint as self-reported history of a physician-diagnosed tumor and interpreted the coefficient as a prevalence odds ratio for that questionnaire-defined history, not as confirmed cancer, malignant-neoplasm history, or incident cancer.
Changes made: We sharpened this wording in Section 2.1, Supplementary Methods S1.1, the Table 1 note, and the figure caption, and used the same terminology consistently throughout the manuscript. Location in revised manuscript: Methods Section 2.1, pp. 4–5, lines 98–115; Supplementary Methods S1.1, p. 1.
Comment 10. Justification and implications of the heterogeneous outcome
The outcome is highly heterogeneous, including tumors from different anatomical sites and potentially different biological behaviors. Please provide a stronger justification for combining these conditions into a single outcome and discuss the implications of this decision.
Response: We agree that the endpoint is heterogeneous. We did not combine separately ascertained malignant and benign diagnoses during this secondary analysis; rather, the original SPES instrument itself collected one broad, uniformly administered tumor-history item and did not preserve the histological or biological-behavior information needed for a defensible malignant-versus-benign reclassification. Any retrospective recoding into confirmed cancer categories would therefore require unsupported assumptions.
In addition, only 86 participants reported a history. Site-specific association models would be sparse and unstable and could invite selective interpretation. The combined questionnaire endpoint was consequently the only reproducible participant-level analysis of the source item.
We stated more explicitly that this choice may dilute an association confined to a particular tumor type, may include benign or pre-invasive conditions, and does not support interpretation as a unified biological cancer phenotype. Site categories remain descriptive only.
Changes made: We strengthened the justification in Section 2.1 and expanded the corresponding limitation in the Discussion. Location in revised manuscript: Methods Section 2.1, pp. 4–5, lines 103–115; Discussion, p. 14, lines 378–387.
Comment 11. Timing of diagnosis relative to serum collection
Please clarify the temporal relationship between tumor diagnosis and blood collection. Were serum samples collected before diagnosis, after diagnosis, or at variable times relative to diagnosis? This information is essential because it directly affects the possibility of reverse causation.
Response: We agree. Serum was collected at the SPES enrollment visit in 2016–2017. For participants reporting a tumor history, collection occurred after the reported diagnosis or, when the derived interval was 0 years, possibly within the same age-year; exact diagnosis and blood-collection dates were unavailable, so exact ordering within that age-year cannot be established. Years since diagnosis was calculated as age at enrollment minus reported age at diagnosis, both recorded in completed years, and all derived intervals were nonnegative. Among the 86 reported histories, the median interval was 5.0 years (IQR 2.0–8.0): 23 diagnoses were reported within 0–2 years, 24 at >2–5 years, 25 at >5–10 years, and 14 at >10 years before enrollment.
We added a stable Cr-52 sensitivity analysis excluding the 23 diagnoses reported within two years of enrollment. The inverse estimate was not attenuated (N = 4,114; 63 reported histories; OR 0.684, 95% CI 0.525–0.892; nominal p = 0.0049), but this does not establish temporality and cannot exclude longer-term post-diagnostic change, survivor selection, or residual confounding.
Changes made: We added the exact derivation of years since diagnosis to Supplementary Methods S1.1 and retained strengthened reverse-causation language throughout the revised manuscript. Location in revised manuscript: Results Section 3.1, p. 9, lines 245–251; Results Section 3.2, p. 11, lines 281–285; Discussion, p. 13, lines 329–337; Supplementary Methods S1.1, p. 1; Supplementary Table S6, p. 7.
Comment 12. Rationale for the covariate sets
Please provide a clearer rationale for the selection of covariates included in the primary and extended models. Explain whether these variables were selected a priori based on previous literature or an epidemiological framework.
Response: We agree. The core adjustment set was prespecified for this secondary analysis before fitting the channel-specific association models, using an epidemiological confounding framework rather than data-driven significance testing. Age, sex, smoking, and geographic cluster were selected because each can influence serum trace-element distributions and the prevalence or reporting of tumor history. Geographic cluster also reflects the SPES sampling structure and the marked small-area variation reported in the published SPES geospatial analysis.
The extended model added body mass index, self-reported occupational exposure, and socioeconomic deprivation as plausible additional confounders. During the requested recheck of continuous covariates, two additional outcome-negative BMI values derived from implausible adult height entries were set to missing; one BMI was already missing in the submitted complete-case extended analysis. The corrected 20-channel extended family therefore used 4,134 participants with valid BMI (86 reporting tumor history and 4,048 reporting no history). The corrected estimates were materially unchanged, and the primary models were unaffected. The valid-BMI population and descriptive summary are consistent with the published Journal of Xenobiotics companion analysis. Potentially post-diagnostic characteristics, such as treatment, current diet, inflammation, and organ function, were not included in the core adjustment set; where available, they could represent consequences of the reported diagnosis rather than baseline confounders, and adjustment could introduce overcontrol or collider bias.
Changes made: We added the a priori covariate rationale to Section 2.2 and Supplementary Methods S1.3. During the requested recheck, two additional outcome-negative BMI values derived from implausible adult height entries were set to missing, and the corrected extended-covariate family replaced the original sensitivity family; the primary analysis was unaffected. Location in revised manuscript: Methods Section 2.2, p. 6, lines 148–163; Supplementary Methods S1.3, p. 2; Table 1, pp. 8–9; Supplementary Table S9, p. 8.
Comment 13. Treatment of values below the LOQ
Please provide more details regarding the treatment of values below the LOQ. For Cr-52, a substantial proportion of participants were below the LOQ, and this may have an important influence on the estimated associations.
Response: We agree. Values below an analyte-specific LOQ were treated as left-censored observations; they were not replaced by zero, LOQ/2, or another fixed fraction. The archived 30-dataset workflow used prespecified channel-specific censoring-aware models: log-scale Tobit, fixed Box–Cox Tobit, or interval-censored Gamma models, with constrained positive draws below the archived LOQ and predictive mean matching as a numerical fallback where required.
For Cr-52, the archived LOQ was 0.0015 µg/L. Among 4,137 questionnaire responders, 4,134 assays were performed, 1,691 were below the LOQ (40.9%), 2,443 were quantified, and no Cr-52 result carried an HI20 plausibility flag. Three unperformed assays were eligible for completion under the parent low-nonperformance rule.
The primary model therefore estimates a full-distribution contrast under the censoring-aware workflow. The quantification-state analysis separately asks whether the probability of being above the LOQ differed by reported history; it does not estimate a concentration gradient. The quantified-only analysis assesses the direction within the quantifiable range but is selected on being above the LOQ. These analyses are complementary, non-independent, and neither resolves the separate question of analytical matrix interference.
Changes made: We expanded S1.2, designated the assay-completion table as Table S2A, added Table S2B containing the archived LOQ and assay-support information for all 20 channels, and added a concise cross-reference in the main Results. We also distinguished left-censoring, assay non-performance, and the post-analytical HI20 plausibility flag. Location in revised manuscript: Methods Sections 2.2–2.3, pp. 5–7, lines 132–139 and 164–211; Supplementary Methods S1.2–S1.3, pp. 1–2; Supplementary Table S2B, p. 4.
Comment 14. Multiple imputation and tumor history
Please clarify the multiple-imputation procedure. Since the 30 imputed datasets were originally generated for previous analyses, please state whether the tumor-history outcome was incorporated into the imputation models and discuss the implications if it was not.
Response: We thank the Reviewer. We re-audited the exact parent imputation script, the archived 30-imputation object used in this study, and the target-specific predictor-selection records.
Although the present secondary analysis was specified after the parent imputation run, the original standardized questionnaire item used to define tumor history was already present in the parent dataset as a non-imputed auxiliary variable. For Cr-52, the archived audit confirmed that the tumor-history indicator was retained after predictor selection in both the censored Tobit component for below-LOQ observations and the predictive-mean-matching component for eligible non-censored missing values. The outcome itself was not imputed and remained unchanged across completed datasets.
Retention was target-specific and was not universal across all channel-specific models. For channels in which the item was not retained, the imputed cells—not the observed quantified concentrations—were generated without direct conditioning on tumor-history status. This creates potential imputation–analysis uncongeniality and could attenuate or otherwise bias an association under some missing-data structures. The revised Supplementary Methods acknowledge this explicitly. The 11 channels with at least 10% left-censoring were also examined using quantification-state and quantified-only analyses based on observed assay information; these analyses reduce, but do not eliminate, the concern and do not constitute independent replication. The remaining channels had less than 10% left-censoring and no more than 5% eligible assay nonperformance by design. Unperformed Hg-202 and Ni-60 assays were not imputed.
Changes made: The imputation-timing statement was corrected in the main and Supplementary Methods, the Cr-52 predictor audit was documented, the outcome was identified as non-imputed, and universal outcome retention across all 20 channels was not implied. Because tumor history was already included in both Cr-52 components, the parent imputation resource was not regenerated solely for this revision. The potential implications for channels in which the item was not retained are now stated explicitly as a limitation. Location in revised manuscript: Methods Section 2.3, p. 6, lines 164–181; Supplementary Methods S1.2–S1.3, pp. 1–2; Discussion, p. 14, lines 388–399.
Comment 15. Standardization across completed datasets
Please clarify how the log-transformed concentrations were standardized and whether the same standardization procedure was applied across the 30 imputed datasets.
Response: We agree. For each analytical channel, the mean and standard deviation of the natural-log concentration were calculated separately in each completed dataset. The 30 means were averaged to obtain one analyte-specific log-mean constant, and the 30 standard deviations were averaged to obtain one analyte-specific log-SD constant.
The same pair of constants was then applied to every completed dataset. The identical constants were also used in the quantified-only analysis, so its odds ratio remained expressed per the same one-standard-deviation increment as the primary analysis rather than being rescaled within the quantified subset.
Changes made: This calculation is now described in the main and Supplementary Statistical Methods, and the exact constants remain in the reproducibility archive. Location in revised manuscript: Methods Section 2.3, p. 7, lines 184–195; Supplementary Methods S1.3, p. 2.
Comment 16. Singular fits and models without a random effect
Please provide additional information regarding model convergence and the treatment of singular fits. For the analytes in which the geographic random-effect variance was essentially zero, consider presenting a sensitivity analysis using an appropriate model without the random effect.
Response: We thank the Reviewer. The diagnostic table reports convergence, singularity, warnings, and random-intercept variance across all 30 completed datasets. For Hg-202 and Ni-60, all 30 primary mixed models were singular because the geographic random-intercept variance was estimated at zero or effectively zero; the exposure coefficients and variances remained finite.
We fitted logistic models without a random effect using the same exposure standardization, age spline, sex, and smoking terms. The model without any cluster term reproduced the boundary mixed-model estimates essentially exactly: Hg-202 OR 1.043 (95% CI 0.824–1.319; nominal p = 0.726) and Ni-60 OR 1.381 (95% CI 0.998–1.911; nominal p = 0.051).
To retain geographic adjustment without estimating a variance component, we also fitted cluster fixed-effects logistic models. Hg-202 remained null (OR 1.003, 95% CI 0.764–1.317; nominal p = 0.982), whereas the Ni-60 estimate was attenuated and more imprecise (OR 1.314, 95% CI 0.933–1.851; nominal p = 0.118). All 30 fits in both alternative specifications converged without warnings. Because the fixed-cluster model estimates many additional parameters with only 47 reported histories in the Ni-60 measured subset, we regard it as a robustness check rather than a replacement primary model. Separately, 37 of 600 BMI-corrected extended-model fits produced scaled-gradient warnings; targeted refitting with independent optimizers yielded materially concordant pooled estimates, no change in FDR classification, and no change in the rounded Cr-52 result.
Changes made: We retained the no-random-effect results and the corresponding convergence, singularity, and variance-component diagnostics in the Supplementary Materials and summarized the targeted alternative-optimizer audit for the BMI-corrected extended family. Location in revised manuscript: Methods Section 2.3, p. 7, lines 184–195; Results Section 3.2, p. 12, lines 296–308; Supplementary Results S2.2–S2.3, pp. 6–9.
Comment 17. Multiple-testing strategy
Please explain more clearly the multiple-testing strategy. The primary analysis included 20 analytical channels, whereas additional sensitivity analyses were also performed. Please clearly distinguish primary analyses from exploratory analyses.
Response: We agree. The only primary family comprised the 20 channel-specific models using the core adjustment set. Benjamini–Hochberg q values were calculated from the 20 unrounded primary p values, with q < 0.05 as the criterion for false-discovery-rate-controlled evidence.
The extended-covariate analysis formed a separate 20-channel sensitivity family. After applying the governed BMI-validity rule, the full extended family was rerun and Benjamini–Hochberg q values were recomputed across the 20 corrected models; no channel met q < 0.05. The quantification-state and quantified-only analyses evaluated different estimands and were each restricted to the 11 channels meeting the prespecified threshold of at least 10% left-censoring; Benjamini–Hochberg adjustment was applied separately within each 11-channel family.
The assay-completion models, no-random-effect models, and recent-diagnosis exclusion were targeted diagnostic or robustness analyses. Their nominal p values are descriptive and do not create additional discovery claims. The Cr-52 estimates are complementary and non-independent; their agreement does not constitute independent confirmation, and interpretation remains governed by the primary result (q = 0.226).
Changes made: This hierarchy remains explicit in the main and Supplementary Statistical Methods, and the corrected extended-family q values replaced the original sensitivity-family values. Location in revised manuscript: Methods Section 2.3, p. 7, lines 192–211; Results Section 3.2, pp. 9–12, lines 252–295; Supplementary Methods S1.3, p. 2; Supplementary Results S2.3, pp. 8–9.
Comment 18. Participant selection and flowchart
Please begin the results by describing the participant selection process, including the initial sample, exclusions, and final analytical population. A flowchart would make this information more didactic.
Response: We agree. Figure 1 already presents the participant flow from 4,205 enrolled participants to 4,188 eligible analytical records and 4,137 participants with an available tumor-history response, with 51 excluded from outcome models because the response was missing. The opening Results paragraph now mirrors that flow explicitly and cross-references the figure at the first mention of the analytical population.
Changes made: We strengthened the first paragraph of Section 3.1 and retained the existing flowchart. Location in revised manuscript: Methods Section 2.1, p. 3, lines 85–93; Figure 1, p. 4, lines 95–97; Results Section 3.1, p. 8, lines 227–233.
Comment 19. Completeness of Table 1
In Table 1, please provide a complete description of the participants, including the main demographic and clinical characteristics, and clearly indicate the number of participants with and without tumor history.
Response: We agree. Table 1 identifies the overall population (N = 4,137), participants reporting tumor history (n = 86), and participants reporting no history (n = 4,051), and includes age, sex, smoking, body mass index, obesity, occupational exposure, socioeconomic deprivation, and environmental-pressure stratum.
As described under Comment 12, the corrected valid-BMI denominator is 4,134: all 86 participants reporting tumor history and 4,048 participants reporting no history. Overall BMI was 25.4 (SD 4.5), with median 24.6 [IQR 22.3–27.7]; obesity was present in 577 participants (14.0%), including 12 of 86 reporting tumor history and 565 of 4,048 reporting no history. These values are consistent with the published Journal of Xenobiotics companion analysis. We did not add variables that were unavailable uniformly or that could be post-diagnostic consequences merely to enlarge the descriptive table.
Changes made: We corrected the BMI and obesity rows and the availability footnote in Table 1, while retaining the existing questionnaire-defined group totals and table structure. Location in revised manuscript: Results Section 3.1 and Table 1, pp. 8–9, lines 227–244.
Comment 20. Presentation of continuous variables
Please review the presentation of continuous variables and ensure that the descriptive statistics correspond to their distributions. Non-normally distributed variables should preferably be presented as median and interquartile range.
Response: We agree. We rechecked the continuous-variable distributions and the underlying validity rules. Age was approximately symmetric and remains summarized using mean and standard deviation. BMI was right-skewed and is reported using both mean (SD) and median [IQR] in the 4,134 participants with valid anthropometry. Socioeconomic deprivation is reported using median and interquartile range. Categorical variables remain n (%), and no between-group hypothesis tests were added.
Changes made: We report BMI as 25.4 (4.5); 24.6 [22.3–27.7] overall, 25.4 (4.3); 24.7 [22.4–28.1] among participants reporting tumor history, and 25.4 (4.5); 24.6 [22.3–27.7] among those reporting no history; the Table 1 footnote states the valid-BMI denominators explicitly. Location in revised manuscript: Table 1 and its note, pp. 8–9, lines 239–244.
Comment 21. Channel-specific sample sizes
Please ensure that the sample size for each analytical channel is clearly presented, particularly because the number of available participants varies between elements and is substantially smaller for some sensitivity analyses.
Response: We agree. The forest plot and Supplementary Tables S3 and S9-S11 report the model N and number of reported histories for every primary and sensitivity estimate. Hg-202 and Ni-60 are explicitly restricted to measured subsets, and the quantified-only analyses show the reduced denominators.
After applying the BMI-validity rule, the corrected extended-covariate family includes 4,134 participants and 86 reported histories for the 18 generally completed channels, 3,352 participants and 73 reported histories for Hg-202, and 2,277 participants and 47 reported histories for Ni-60. The primary, quantification-state, quantified-only, recent-diagnosis, and no-random-effect analyses are unaffected. Table S2B reports raw assay-performed and unperformed counts, below-LOQ counts, quantified counts, HI20 flags, and primary model N for all 20 channels.
Changes made: We replaced all 20 rows of Supplementary Table S9 with the BMI-corrected extended-family estimates and denominators and retained the main-text cross-references to Figure 2, Table S2B, and the complete association tables. Location in revised manuscript: Figure 2, p. 10, lines 260–264; Table 2, pp. 11–12, lines 289–295; Supplementary Tables S3 and S9–S11, pp. 4 and 8–9.
Comment 22. Forest plot for the primary associations
Please present the primary regression results in a more didactic manner. A forest plot showing ORs and 95% CIs for the 20 analytical channels could considerably improve the presentation.
Response: We agree with the requested presentation. The submitted manuscript already contains Figure 2, a forest plot of the adjusted prevalence odds ratios and 95% confidence intervals for all 20 analytical channels, with the channel-specific N and number of reported histories displayed.
Changes made: We retained Figure 2 and ensured that its caption states the one-standard-deviation exposure scale, adjustment set, channel-specific denominators, and the fact that no channel met q < 0.05. Location in revised manuscript: Figure 2 and caption, p. 10, lines 260–264.
Comment 23. Interpretation of the Cr-52 result
Please revise the interpretation of the Cr-52 finding. The OR of 0.751 (95% CI 0.601-0.937; p=0.011) is nominally significant, but the FDR-adjusted q-value is 0.226. Therefore, this finding should be clearly described as exploratory rather than statistically significant after correction.
Response: We agree. The revised text states consistently that no analytical channel met the prespecified false-discovery-rate criterion. Cr-52 is described as a nominal, exploratory, hypothesis-generating pattern rather than as statistically significant after multiplicity correction.
The nominal p value and 95% confidence interval are retained for transparency, but the panel-level conclusion remains based on q = 0.226. The sensitivity analyses do not supersede that conclusion and do not justify causal, protective, confirmatory, or independent-replication language.
Changes made: This wording was applied consistently to the statistical Methods, Results, and Discussion passages throughout the revised manuscript. Location in revised manuscript: Results Section 3.2, p. 9, lines 253–258; Discussion, p. 12, lines 309–318; Conclusions, p. 15, lines 417–425.
Comment 24. Denominators and non-independence of sensitivity analyses
The sensitivity analyses for Cr-52 and Ni-60 should be interpreted cautiously. In particular, the quantified-only analyses considerably reduce the number of participants with tumor history. Please provide the exact sample size and number of cases for each analysis and clarify that these analyses do not represent independent replication.
Response: We agree. For Cr-52, the primary analysis included 4,137 participants and 86 reported histories; the BMI-corrected extended-covariate analysis, 4,134 and 86 (OR 0.754, 95% CI 0.603–0.941; p = 0.013; q = 0.253); the quantification-state analysis, 4,134 and 86; and the quantified-only analysis, 2,443 and 41. The stable two-year recent-diagnosis exclusion included 4,114 participants and 63 reported histories.
For Ni-60, the primary and BMI-corrected extended-covariate analyses each included 2,277 participants and 47 reported histories; the quantification-state analysis included 2,271 and 47; and the quantified-only analysis included 1,584 and 36. The no-random-effect sensitivities used the same 2,277 participants and 47 reported histories as the primary Ni-60 measured subset.
Changes made: These corrected denominators and estimates are reported explicitly. The analyses continue to be described as complementary and non-independent; the quantified-only estimates apply only to participants with results above the LOQ and do not independently replicate the primary result. The BMI correction affects only the extended sensitivity family and does not alter the primary or measured-data analyses. Location in revised manuscript: Results Section 3.2, pp. 11–12, lines 265–308; Table 2, pp. 11–12, lines 289–295; Supplementary Results S2.3 and Tables S9–S11, pp. 8–9.
Comment 25. Temporality, causality, and reverse causation
The discussion should emphasize that the study cannot establish temporality or causality. Because serum measurements were obtained after or around the tumor diagnosis, reverse causation is a plausible explanation that should be discussed more prominently.
Response: We agree. The design is cross-sectional with respect to current serum Cr-52 and retrospective with respect to the questionnaire history. Exact diagnosis and collection dates were unavailable, and a zero-year interval does not establish within-year ordering. Therefore, the study cannot determine whether the measured chromium distribution preceded or followed the biological and behavioral consequences of diagnosis.
The stable exclusion of diagnoses reported within two years of enrollment reduces concern that the result is driven exclusively by the most recent diagnoses, but it does not solve reverse causation. Longer-term treatment, dietary change, metabolic alteration, survivor selection, outcome misclassification, residual confounding, and chance remain plausible explanations.
Changes made: Reverse causation was moved earlier in the Discussion, temporality is explicitly described as unestablished, and the two-year exclusion is presented as a limited robustness analysis rather than causal evidence. Location in revised manuscript: Discussion, p. 13, lines 329–337.
Comment 26. Biological interpretation and chromium species
Please moderate the biological interpretation of the Cr-52 finding. The proposed mechanisms are interesting, but they should be presented as possible explanations or hypotheses rather than mechanisms demonstrated by the present study. The discussion should also emphasize that total serum chromium does not identify the specific chromium species responsible for the measured concentration.
Response: We agree. Chromium speciation information was unavailable in the archived analytical dataset, and the total serum Cr-52 channel cannot identify Cr(III), Cr(VI), a protein-bound species, source, or historical dose.
The mechanistic section was shortened. Insulin signaling, nutrition, protein binding, renal handling, inflammation, treatment, and other metabolic features are presented only as non-exclusive hypotheses. The chromium–transferrin literature is acknowledged without assigning the SPES total-serum samples to Cr(III). The Discussion states directly that no protective, deficiency, supplementation, or low-Cr(VI) interpretation is supported.
Changes made: The extended mechanistic discussion was replaced by focused paragraphs centered on chemical non-specificity, unavailable speciation, randomized-trial evidence, and alternative explanations. Location in revised manuscript: Discussion, p. 13, lines 338–356.
Comment 27. Expanded limitations
Please expand the limitations section to include the cross-sectional design, self-reported and heterogeneous tumor outcome, relatively small number of tumor cases, multiple comparisons, substantial values below the LOQ for some elements, potential selection in quantified-only analyses, and residual confounding.
Response: We agree. The revised limitations section explicitly includes every item listed by the Reviewer.
The revised limitations section states that the design cannot establish temporality or causality; the outcome was heterogeneous, self-reported, and unvalidated; only 86 histories were available, limiting precision and precluding site-specific inference; 20 primary comparisons required FDR control; Cr-52 and several other channels had substantial left-censoring; quantified-only analyses condition on being above the LOQ and therefore apply to selected subsets; Hg-202 and Ni-60 used measured subsets; total serum measurements lack species and source information; some archived analytical performance details were unavailable; imputation predictor retention was target-specific; and residual confounding and post-diagnostic change remain possible.
The section also explains that the complete panel and sensitivity analyses improve transparency but do not overcome these design limitations.
Changes made: The limitations were consolidated into structured paragraphs ordered from outcome/design limitations to statistical, missing-data, and analytical limitations, with repetition reduced elsewhere. Location in revised manuscript: Discussion, p. 14, lines 378–399.
Comment 28. Final interpretation
Please revise the final interpretation of the study so that it is consistent with the statistical results. The main contribution of the study is the identification of an exploratory pattern that may justify further investigation, preferably through prospective studies using prediagnostic samples and validated cancer outcomes. The conclusion should not suggest that chromium protects against or causes tumors.
Response: We agree. The conclusion begins with the panel-level result: no channel met the prespecified FDR criterion. Cr-52 is described as the only nominal exploratory pattern, not as evidence that chromium protects against or causes tumors.
The final sentence specifies the appropriate next study: prospective collection or use of prediagnostic specimens, validated registry-linked cancer outcomes, species-appropriate biomarker matrices, adequate event numbers, and prespecified multimetal and multiplicity control.
Changes made: The Conclusion was rewritten to define the contribution as complete panel reporting and hypothesis generation, with no causal, protective, deficiency, supplementation, or species-specific implication. Location in revised manuscript: Conclusions, p. 15, lines 417–425.
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsCan be accepted in a current form
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors have provided a thorough and satisfactory response to resolve my concerns.
Reviewer 4 Report
Comments and Suggestions for AuthorsDear Authors,
Thank you for your thorough point-by-point response. I verified that all changes described in your letter are consistently reflected in the revised manuscript, and the cautious, well-aligned interpretation of the results is appropriate. I am happy to recommend acceptance in its present form.