# Judge skin cancer AI by the thick melanomas it prevents, not the thin ones it finds

Source: https://onco.cc/ideas/idea-prev-melanoma-ai-thick-melanoma-metric/  
OnCo record `idea-prev-melanoma-ai-thick-melanoma-metric` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Melanoma diagnoses have soared while deaths barely changed, a sign of overdiagnosis. AI skin apps should be judged on whether dangerous thick melanomas fall, not how many spots they flag.

## Summary

Melanoma incidence has risen roughly six-fold in the US since 1975 with much smaller mortality change. AI dermatology tools risk amplifying detection of indolent in-situ lesions. Propose that regulatory and reimbursement evaluation of AI skin triage require thick (over 1 mm) melanoma incidence and biopsy-to-melanoma ratio as endpoints.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: AI triage tools judged by detection volume increase in-situ diagnoses and biopsies without reducing thick melanoma incidence; tools optimised against thick-melanoma endpoints adopt more conservative operating points.
- Rationale: The metric determines the algorithm's threshold.
- Proposed test: Cluster RCT of AI triage in primary care with three-year thick melanoma incidence and biopsy counts.
- Maturity: speculative
- Actor: regulator

## Sources

- Bottleneck evidence (Overdiagnosis and false alarms): Welch & Black, Overdiagnosis in cancer (JNCI 2010): https://doi.org/10.1093/jnci/djq099

## Connected records

- cancers: [Melanoma](https://onco.cc/cancers/melanoma/)
- fronts: [Early Detection & Screening](https://onco.cc/fronts/early-detection/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Overdiagnosis and false alarms](https://onco.cc/bottlenecks/b-overdiagnosis/)
- key papers: [Overdiagnosis in cancer](https://onco.cc/key-papers/paper-welch-j-natl-cancer-inst/)

---
JSON: https://onco.cc/api/v1/entities/idea-prev-melanoma-ai-thick-melanoma-metric.json