Impact of Dermatologic Screening and Methods on Breslow Thickness in Melanoma: A Retrospective Cohort Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

[General comments to Authors] 

The authors conducted a retrospective study of 475 histopathology‑confirmed cutaneous melanomas in 397 patients to challenge current screening strategies and to assess the potential implications for personalized screening recommendations in clinical practice. 

They show that screening is associated with thinner melanomas and a higher proportion of in situ diagnoses (relative to invasive cases) compared with first‑visit diagnoses. Their findings support risk‑adapted screening strategies, in which resource‑intensive digital techniques such as mole mapping and total body photography are best suited for high‑risk patients. 

Strategies for melanoma screening are widely debated and remain to be optimized both in terms of patient benefit and cost to healthcare systems. In this regard, the authors’ efforts to identify better approaches should be supported. This retrospective study is well written, scientifically sound, and places the findings within the broader context of cutaneous melanoma diagnosis, including the problem of overdiagnosis. The authors interpret the results carefully and discuss the study’s limitations transparently. 

[Detailed comments to Authors] 

The introduction is well written and clearly focused on the role of screening in preventing melanoma‑associated deaths. The authors effectively introduce the reader to the subject and rationale of their study. 

The patient population and statistical approach are described transparently and with sufficient detail to allow reproducibility and proper assessment of the methods. 

Table 1 is missing the header row. 

The main results of this retrospective study: - that patients who underwent regular or irregular screening had significantly lower mean Breslow thickness of invasive melanomas compared with patients without prior screening, and  - that there were no statistically significant differences in Breslow thickness between the groups with irregular, regular, or close screening -  are consistent with findings from recent studies. However, the detailed analysis of almost 500 melanomas adds important value to our understanding of the impact of different screening strategies in distinct patient groups. 

The discussion provides adequate context for the presented results. Relevant literature is properly cited, with appropriate commentary on limitations and potential biases. Finally, the authors propose efficient screening strategies based on their findings and discuss risk prediction models, which aim to further enhance efficiency by reducing the number needed to screen through risk‑adapted approaches. 

The conclusions are consistent with the evidence and arguments presented in the paper. The main objective—to contribute to the discussion on personalized screening recommendations tailored to the specific needs of low‑ and high‑risk melanoma patients—is convincingly addressed.

Author Response

The header row has been added to Table 1.

We thank the reviewer for the thorough and positive evaluation of our manuscript and for the constructive comments. We are pleased that the reviewer found the study to be well written, scientifically sound, and appropriately contextualized within the ongoing debate on melanoma screening and overdiagnosis. 

Reviewer 2 Report

Comments and Suggestions for Authors

In this manuscript, Wunderlich, Potiez, and colleagues address a relevant clinical question: whether the frequency of dermatologic screening and the diagnostic modality influence melanoma stage at diagnosis, with a particular focus on Breslow thickness and melanoma in situ. The study benefits from a well-defined cohort, access to a cancer registry comparator, and a clinically meaningful stratification by melanoma risk. The conclusions supporting risk-adapted screening strategies and targeted use of digital mole mapping are broadly consistent with the presented data and current guidelines. However, the statistical methodology and reporting require clarification and strengthening to support the causal interpretation of the findings and to avoid overstatement. Several analyses are primarily univariable, potentially confounded, and inconsistently reported. Addressing the points below would significantly improve the robustness and interpretability of the study.

• This retrospective single-centre study addresses an important and clinically relevant question. Nevertheless, the design inherently limits causal inference, and selection and referral biases merit more explicit consideration, particularly in the Methods and Results sections.
• Patients diagnosed at a first visit differ substantially from those under longitudinal follow-up (e.g., referral patterns and symptom-driven presentations). Although the authors appropriately perform sensitivity analyses excluding first-visit cases, this distinction would benefit from being pre-specified and more clearly integrated into the analytic framework, rather than presented as a post hoc adjustment. Overall, the analyses should be explicitly framed as exploratory and associational, with clear acknowledgement of potential confounding by indication when comparing first-visit and follow-up patients.
• The dataset includes multiple melanomas per patient (475 melanomas in 397 patients). While the authors state that random-effects models were used to account for repeated measures, it is not always clear which analyses rely on mixed-effects models versus univariable non-parametric tests, and no model estimates or confidence intervals are presented. Clarifying the modelling strategy and reporting effect estimates with 95% confidence intervals would strengthen the statistical interpretation.
• Breslow thickness is right-skewed, making the use of non-parametric tests appropriate. However, the manuscript inconsistently reports means ± SD and medians [IQR] across tables. Using median and IQR consistently, or justifying the use of means (e.g., via transformation in regression models), would improve clarity and consistency.
• Given the number of subgroup and stratified analyses, the approach to multiple testing should be clarified. If no adjustment was applied, this should be acknowledged, and marginal p-values interpreted cautiously.
• Finally, while risk-stratified analyses are a strength, it would be helpful to clarify whether formal interaction testing between melanoma risk and screening frequency or modality was performed. Similarly, comparisons with Belgian Cancer Registry data should be interpreted as descriptive rather than inferential, given likely differences in case mix and follow-up intensity.

Minor Comments
The term “random-effects linear models” should be replaced with “linear mixed-effects models” for clarity.
Tables 2 and 3 would benefit from explicitly stating whether p-values are from univariable or multivariable analyses.

Author Response

We thank the reviewer for the careful and constructive assessment of our manuscript. We fully agree that the retrospective, single-centre design limits causal inference and that the findings should be interpreted as associational. Our intention was not to imply causality, but rather to explore clinically relevant associations between screening patterns and melanoma stage at diagnosis.

Causal interpretation, selection bias, and analytic framing : 

We acknowledge the inherent limitations related to selection and referral bias, particularly when comparing first-visit diagnoses with melanomas detected during longitudinal follow-up. These limitations are already discussed in the manuscript. The sensitivity analyses excluding first-visit cases were included to illustrate the robustness of the observed associations and were not intended as post hoc causal adjustments.

Confounding and exploratory nature of the analyses : 

We agree that confounding by indication cannot be excluded in this retrospective setting. The analyses are exploratory and hypothesis-generating. Given the observational design and the available data, no claims of causal inference are made.

Repeated measures and statistical approach : 

We confirm that random-effects models were used to account for multiple melanomas per patient where appropriate. Univariable non-parametric tests were used for descriptive comparisons. 

Comparison with Belgian Cancer Registry data : 

Comparisons with Belgian Cancer Registry data were intended to be descriptive and contextual, not inferential, and are interpreted as such.

 

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript entitled “Impact of Dermatologic Screening and Methods on Breslow Thickness in Melanoma: A Retrospective Cohort Study” evaluated the association between screening frequency, exam type, diagnostic approach andmelanoma in situ proportion and Breslow thickness using a single-center retrospective cohort. The study concluded that the risk-adapted screening strategies, especially with digitally supported longitudinal follow-up is beneficial to the high-risk patients. It is an interesting study with potentially significant implications; however, major concerns need to be addressed to improve the conclusion.

1. It has a potential risk of over-interpretation of association. Firstly, in this manuscript, first-visit at Erasme was implied as no prior screening. However, this may not be accurate/true. For example, some of these patients may have been initially evaluated in other hospitals/clinics before referral to Erasme. This could potentially increased the selection bias. Despite this point was acknowledged in Discussion, the Screening vs first-visit results still relied heavily on this comparison. It would be useful to include more information for those patients, such as being screened outside Erasme or not, referral source, and how it was handled. This can further improve the transparency of this study. Second, digital mole mapping is typically used in people with certain risk factors, such as with many nevi/atypical nevi, prior melanoma, strong family history, etc. This could, at least partially, explain why the low-risk mole mapping subgroup was so small. Therefore, how to rule out or take into account other factors in those with digital mole mapping? Thirdly, Breslow thickness usually is right-skewed and often has outliers, making mean comparisons potentially misleading (for example, some data like 1.8+/-3.48 mm, suggest heavy skew). Using median is recommended, as was in Table 4.
2. Some wording of the manuscript can be misleading. For example, the title reads like causation. Summary of each section and conclusions all framed results as if screening “produces/causes” thinner tumor.
3. Need to improve the method transparency. For example,  “patients with missing data were excluded from corresponding analyses” is not enough. Please specify missing-data amounts per key variable.

Author Response

Comment on first-visit classification and prior screening : 

We thank the reviewer for raising this important point. We fully agree that a first visit at Erasme does not necessarily imply the absence of prior dermatologic evaluation in another institution or private practice. Due to the retrospective nature of the study and the structure of the available medical records, information regarding prior screening performed outside Erasme was not systematically available.

Even if such information had been accessible, we acknowledge that the quality, frequency, and diagnostic modalities of screening performed in other institutions may differ substantially, making direct comparison with screening conducted at Erasme difficult and potentially misleading. For this reason, screening outside Erasme could not be reliably categorized or integrated into the analytic framework.

This limitation is acknowledged in the Discussion and further supports our decision to interpret the findings as associational rather than causal. The comparison between first-visit and follow-up diagnoses should therefore be understood as a pragmatic clinical distinction reflecting different care pathways within our institution, rather than as a definitive measure of prior screening exposure.

Digital mole mapping and potential confounding : 

We agree with the reviewer that digital mole mapping is preferentially used in patients with established melanoma risk factors. This clinical practice likely explains the small size of the low-risk subgroup undergoing digital mole mapping.

Due to the retrospective design and the lack of systematic documentation of all factors influencing the decision to use digital mole mapping, these variables could not be fully accounted for in the analyses. As a result, confounding by indication cannot be ruled out. The findings related to digital mole mapping should therefore be interpreted as associational and reflective of real-world clinical practice, rather than as evidence of a causal effect.

Breslow thickness : 

We agree that Breslow thickness is right-skewed and may include outliers, which can make mean values sensitive to extreme observations. For this reason, non-parametric statistical tests were used, and median values are reported in key analyses. Mean and standard deviation values were included for descriptive purposes and for comparability with the data of Belgian Cancer Registry.

Missing data : 

Exact counts of missing values per variable were not systematically recorded in the retrospective dataset. Missing data were limited and did not affect the primary outcome variables.