Author Correction: Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials

Correction to: npj Digital Medicine 10.1038/s41746-022-00613-w, published online 08 June 2022

The original version of the published Article contained an error in the description of the prostate tissue samples in the Methods and Results sections, which stated that “pretreatment biopsy samples” and “pretreatment prostate biopsies” were used in this study. Both instances have been updated to “pretreatment and posttreatment prostate tissue”, and these changes are reflected in the HTML and PDF versions of the Article. Furthermore, we have added descriptions of the use of posttreatment prostate tissue during model development throughout the Article.

In the interest of transparency and clinical relevance, we repeated the evaluation of model performance in the validation set after removing any cases with post-treatment prostate tissue. This results in a slightly smaller cohort (n = 931) for validation, compared to the 20% validation set used in the original paper, and still demonstrates similar performance to the MMAI algorithms using only pretreatment tissue for each clinical endpoint. Table 2 has been added to the HTML and PDF versions of the Article.

Table 2.

Validation results for the subset of patients from the 20% validation set reported in the Article that includes patients with pretreatment slides only (n = 931).

Clinical outcome	NCCN AUC estimates (95% CI)	MMAI AUC (95% CI)	Differential AUC estimate (MMAI - NCCN)	Comparative test p value
Distant metastasis (5-year)	0.72 (0.67–0.78)	0.83 (0.78–0.88)	0.11	<0.001
Distant metastasis (10-year)	0.69 (0.64–0.74)	0.78 (0.73–0.84)	0.09	<0.001
Biochemical failure (5-year)	0.61 (0.57–0.64)	0.69 (0.65–0.73)	0.08	<0.001
Biochemical failure (10-year)	0.62 (0.58–0.66)	0.68 (0.63–0.72)	0.06	0.004
Prostate cancer-specific survival (10-year)	0.67 (0.61–0.73)	0.77 (0.70–0.83)	0.10	<0.001
Overall survival (10-year)	0.57 (0.54–0.61)	0.65 (0.61–0.69)	0.08	<0.001

The subcohort represented in this table is smaller than reported in the Article (n = 1109).

AUC area under the curve, CI confidence interval, MMAI multimodal artificial intelligence, NCCN National Comprehensive Cancer Network.

Data availability

The data availability statement has also been updated in the PDF and HTML versions of the Article to read, “The data published in this article will be publicly available six months from publication, through requests made to NRG Oncology at APC@nrgoncology.org”.