Paperonium

Prostate cancer · PLoS Med · 2019

The summary’s 70% accuracy for a prostate scan is an earlier trial’s figure. This trial: 63%.

This trial, 338 men63%Its own count213 of 33870%The summaryThe earlier trial, 161 menthe summary used this row113 of 16160%657075
This trial, 338 men63%Its own count: 213 of 33870%The summaryThe earlier trial, 161 menthe summary used this row113 of 16160%657075

Share of men the scan got right: alarm for serious cancer, all-clear otherwise

The paper says

This trial repeated an earlier one-hospital trial of a prostate MRI scan (161 men) at several hospitals (338 men). Its summary says 338 men completed it, then gives the scan’s accuracy, the share of men it got right, as “70% (113/161)”.

Its own data show

The paper’s table comparing the two trials prints 70% (113 of 161) for the earlier trial and 63% for this one. This trial’s counts agree: the scan raised the alarm for 142 men with serious cancer and gave the all-clear to 71 without it, 213 of 338. Of 125 wrong calls, 121 were false alarms.

“Serious” means a tissue sample showed cancer above the lowest grade of aggressiveness. A repeat trial exists to check whether a result holds at other hospitals, so its own number is the news: accuracy fell 7 points. The summary’s 70% makes it look unchanged. The paper’s key point stands: only 4 of 75 men the scan cleared had serious cancer.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

Validation of IMPROD biparametric MRI in men with clinically suspected prostate cancer: A prospective multi-institutional trial · PLoS Med (2019) · doi:10.1371/journal.pmed.1002813