Comprehensive analysis of the 1,000-lesion study were accepted for publication by Cancer Research, a peer-reviewd journal of the American Association of Cancer Research and the most widely cited cancer journal in the world. The manuscript, titled “Real-time Raman Spectroscopy for In Vivo Skin Cancer Diagnosis”, was first published in the online version of the journal in late March.
As a reminder, Verisante Technology Inc. (VRSEF) announced the preliminary statistical analysis earlier this year which were presented by Dr. Haishan Zeng at the SPIE (international society for optics and photonics) conference in San Francisco. Data was acquired on over 1,000 lesions from 848 patients. The study analysis focused on only those classes of lesions that are considered to be associated with concern of skin cancer – which resulted in a final dataset of 518 lesions from 453 patients. Specifically, the classes of skin lesions included in the analysis were those diagnosed as malignant or premalignant (i.e. – malignant melanoma, squamous cell carcinoma, basal cell carcinoma, and actinic keratosis – in aggregate, these represent essentially 100% of skin cancers) and benign conditions that look like skin cancer (i.e. – seborrheic keratosis, atypical nevi, malanocytic nevi, compound nevi, and intradermal and blue nevi).
The study focused on three distinct categories, specifically Aura’s ability to differentiate; 1) skin cancers and pre-cancers from benign skin lesions, 2) melanomas from non-melanoma pigmented lesions, and 3) melanomas from seborrheic keratoses. Statistical analysis was done using two methods; principal component with general discriminant analysis (GDA) and partial least squares. With sensitivities between 95% and 99%, Aura demonstrated specificities in these three tasks between 15% and 54% with GDA (in most instances using partial least squares resulted in higher specificities) – which resulted in a biopsy ratio (the number of non-melanoma lesions that undergo biopsy for each confirmed case of melanoma) of between 0.77:1 to 5.56:1. Specifically (see Table 3 from the study below) the GDA analysis showed, in differentiating 1) skin cancers and pre-cancers from benign skin lesions, specificity was 17% at 99% sensitivity and 41% at 95% sensitivity (partial least squares showed specificity of 24% at 99% and 52% at 95%) 2) melanomas from non-melanoma pigmented lesions was 15% at 99% sensitivity and 38% at 95% sensitivity, and 3) melanomas from seborrheic keratoses was 25% at 99% sensitivity and 54% at 95% sensitivity. The study authors concluded that, ” Raman spectroscopy can distinguish (1) malignant and premalignant lesions from benign disorders, (2) melanomas from benign pigmented skin lesions, and (3) melanomas from seborrheic keratoses.”
The biopsy ratio is a metric which helps gauge the accuracy of melanoma diagnosis – the lower the ratio, the more accurate the diagnosis. Misdiagnosis of skin cancer is rampant, often resulting in unnecessary biopsies. Biopsies can be costly, uncomfortable, and result in scarring. If Aura can improve the accuracy of diagnosis (which we believe it can), there should be substantial commercial demand for the device. Relative to the biopsy ratio, for reference the manuscript cites a retrospective study of ~5,000 pigmented skin lesions evaluated by ~500 general practitioners which resulted in a biopsy ratio between 58:1 and 21:1 (i.e. – 21 non-melanoma lesions biopsied for every confirmed case of melanoma). Other unrelated studies have indicated biopsy ratios in clinical practice may be even significantly higher than these figures. Although it’s not necessarily fair (as there could be significant differences in the make-up of the various studies) to make a direct comparison of the biopsy ratios in the Aura study (which appear comparably very strong) with these other studies, at least on the face of it, the Aura data looks potentially persuasive.
As we noted in our most recent update, the Aura data does look to be very compelling relative to clinical trial data from MELA Sciences Melafind device which showed a sensitivity of 98% and specificity of about 10% in the detection of melanoma. Aura’s biopsy ratio appears similarly relatively strong. Also noteworthy is that MELA opted to use a statistical method different than either GDA or partial least squares, both of which are considered conservative. As such, it’s possible that Aura’s specificities may have been relatively even stronger than Melafind’s had the statistical methods used for both devices’ studies been the same. Also as we’ve previously explained, Aura has other advantages to Melafind including a significant speed advantage (allowing for full body scans) and the ability to detect of all types of skin cancer whereas Melafind is used for only melanoma.
To read this article on Zacks.com click here.
Zacks Investment Research
