Gershenwald 2017 — AJCC 8th edition: melanoma staging and survival gradient
Citation
Gershenwald JE, Scolyer RA, Hess KR, Sondak VK, Long GV, Ross MI, et al. Melanoma staging: evidence-based changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA Cancer J Clin. 2017 Nov;67(6):472–492. DOI: 10.3322/caac.21409. PMID 29028110.
Study design and population
Pooled international retrospective and prospective cohort underpinning the AJCC 8th edition melanoma staging system. >46,000 stage I–III melanoma patients from 10 international centres in the International Melanoma Database and Discovery Platform.
Reported metrics
5-year melanoma-specific survival (MSS) by substage:
- IA 99 %; IB 97 %; IIA 94 %; IIB 87 %; IIC 82 %
- IIIA 93 %; IIIB 83 %; IIIC 69 %; IIID 32 %
Per-stratum 95 % CIs in primary tabular data. Tumour thickness (Breslow), ulceration and N-status confirmed as independent prognostic factors.
Surrogate-to-outcome linkage
Provides the load-bearing quantitative anchor for the entire Pillar-1 diagnostic-accuracy-to-outcome chain: stage-at-detection is the dominant determinant of melanoma-specific survival. Any intervention (including AI-assisted diagnosis) that moves detection to earlier T-stage translates, via this gradient, into measurable melanoma-specific survival gain. This is the regulator-recognised causal link that licenses diagnostic accuracy as a proxy in AI-dermatology devices targeting melanoma.
CRIT1–7 appraisal
| Criterion | Score | Justification |
|---|---|---|
| CRIT1 Relevance | 3 | Foundational evidence for the stage-to-survival link anchoring the surrogate argument. |
| CRIT2 Methodology | 3 | Pooled international multi-centre cohort; AJCC evidence base; very large sample. |
| CRIT3 Reporting | 3 | Substage MSS with CIs; Cox regression with adjusted hazard ratios. |
| CRIT4 Applicability | 3 | Contemporary staging system; regulator-recognised. |
| CRIT5 Evidence weight | 2 | Very large pooled cohort (not RCT, not meta-analysis — but highest-weight observational evidence available for this question). |
| CRIT6 Risk of bias | 2 | Retrospective pooling; heterogeneity across centres; pre-immunotherapy adjuvant era for earlier survival readouts. |
| CRIT7 Contribution | 3 | Central quantitative anchor — without this, the diagnostic-accuracy surrogate lacks an outcome linkage. |
Aggregate: very strong.
Limitations and notes
Retrospective pooling; heterogeneity; primarily Caucasian populations; pre-immunotherapy adjuvant-treatment era for earlier cohorts.
Strength as anchor
Essential. This is the quantitative anchor for the stage-at-detection → survival surrogate-to-outcome claim. Without Gershenwald 2017 (or the preceding Balch 2001 / AJCC 7th), the directional argument is qualitative rather than quantitative.