Algorithm-based smartphone ‘apps’ for assessment of the risk of skin cancer in adults: a systematic review of diagnostic accuracy studies

Research output: Contribution to journalArticle

Authors

  • K. Freeman
  • Susan Bayliss
  • RN Matin
  • A Jain
  • Fiona M Walter
  • HC Williams

Colleges, School and Institutes

External organisations

  • Institute for Applied Health Research

Abstract

Background: Skin cancer has one of the highest global incidences of any cancer. Early detection and treatment, particularly of melanoma, can improve survival. Algorithm-based smartphone applications (apps) potentially offer a means of ensuring that the right people seek specialist medical attention by providing an instant risk assessment for a new or changing mole. However these apps may cause harm from failure to identify potentially fatal skin cancers or from over-investigation of false positive results.

Objective: To assess the validity and findings of studies of the accuracy of algorithm-based smartphone apps for the risk assessment of suspicious skin lesions.

Design: A systematic review of studies of diagnostic test accuracy.

Data sources: Cochrane Central Register of Controlled Trials; MEDLINE; Embase; CINAHL; CPCI; Zetoc; Science Citation Index; and online trial registers (database inception to 10 April 2019).

Eligibility criteria for selecting studies: Studies of any design evaluating algorithm-based smartphone apps to assess images of skin lesions suspicious for skin cancer were eligible. Reference standards included i) histological diagnosis or follow-up, and ii) expert recommendation for further investigation or intervention. Two authors independently extracted data and assessed validity using the QUADAS-2 tool. Estimates of sensitivity and specificity were reported for each app.

Results: Nine studies evaluating six different identifiable smartphone apps were included; six verified results using histology/follow-up (n=725 lesions) and three verified results using expert recommendations (n=407 lesions). Studies were small and of poor methodological quality with selective recruitment, high rates of unevaluable images, and differential verification. Lesion selection and image acquisition was by clinicians rather than smartphone users. Two CE-marked apps are available for download. SkinScan was evaluated in a single study (n=15, 5 cases) with 0% sensitivity and 100% specificity for the detection of melanoma. SkinVision, evaluated in two studies (n= 252, 61 cases) has achieved sensitivity of 80% (95% CI 63% to 92%) and specificity 78% (95% CI 67% to 87%) for the detection of malignant or pre-malignant lesions. Accuracy of the SkinVision app verified against expert recommendations was poor (3 studies).

Conclusions: Current algorithm-based smartphone apps cannot be relied on to detect all cases of melanoma or other skin cancers. Test performance is likely to be poorer than reported here when used in clinically relevant populations and by the intended users of the apps. The current regulatory process for award of CE-markings for algorithm-based apps does not provide adequate protection to the public.

PROSPERO registration: CRD42016033595.

Details

Original languageEnglish
Article numberm127
JournalBMJ
Volume2020
Issue number368
Publication statusPublished - 10 Feb 2020