All statistical analyses were carried out in accordance with a statistical analysis plan agreed before the data were inspected. Stata®version 13.0 (StataCorp LP, College Station, TX, USA) was used to carry out the

analyses. The primary analysis was based on all evaluable data, excluding men without all three test results and any data that were rejected as part of the external mpMRI quality control/quality assurance process. For each comparison, 2 × 2 contingency tables were used to present the results and calculate the diagnostic accuracy estimates with 95% CIs. Given the paired nature of the test results, McNemar’s tests were used for the head-to-head comparisons of sensitivity and specificity between mpMRI and TRUS-guided biopsy. Because the PPV and NPV are dependent on disease prevalence, a generalised estimating equation (GEE) logistic regression model was used to compare the PPV and NPV for mpMRI and TRUS-guided biopsy against those for TPM-biopsy.89,90

The sensitivities, specificities and predictive values were calculated for mpMRI based on the overall

radiological score for mpMRI [assigned to definition 2 on the mpMRI case report form (CRF)] and definition 1 for CS cancer on TPM-guided biopsy.

The format of the 2 × 2 table is shown inTable 5. Specificity=d/(c+d), where d=the number of men

testing negative on mpMRI and negative for CS cancer on TPM-biopsy and c=the number of men testing

positive on mpMRI and negative for CS cancer on TPM-biopsy. The NPV=d/(b+d), where d=the number

of men testing negative on mpMRI and negative for CS cancer on TPM-biopsy and b=the number of

men testing negative on mpMRI who test positive for CS cancer on TPM-biopsy. Sensitivity=a/(a+b),

TABLE 4 Required sample size for McNemar’s test for different levels of agreement between mpMRI and TRUS-guided biopsy

mpMRI results

TRUS-guided biopsy result (for true cases)a

Required number of casesb

Required sample size Negative Positive –ve +ve ve +ve Prevalence of 15% (definition 1) Prevalence of 25% (definition 2) Sensitivity=70% 0.29 0.23 0.01 0.47 48 321 192 0.25 0.27 0.05 0.43 66 441 264 Independence assumptionc 0.156 0.364 0.144 0.336 107 714 428 0.05 0.47 0.25 0.23 153 1021 612 0.01 0.51 0.29 0.19 170 1134 680

a Sensitivity of TRUS-guided biopsy=48% in all cases. b For 90% power and two-sided 5% significance.

c Assumes that results of mpMRI and TRUS-guided biopsy are independent for each individual. Note

The shaded rows reflect the scenario in which virtually all cancers are detected by either mpMRI or TRUS-guided biopsy, and so there is extremely low agreement between mpMRI and TRUS-guided biopsy. This does not make clinical sense and is very unlikely but is included for completeness.

METHODS

NIHR Journals Library www.journalslibrary.nihr.ac.uk

where a=the number of men testing positive on mpMRI and positive for CS on TPM-biopsy and b=the

number of men testing negative on mpMRI who have CS cancer on TPM-biopsy. The PPV=a/(a+c),

where a=the number of men testing positive on mpMRI and positive for CS on TPM-biopsy and c=the

number of men testing positive on mpMRI who do not have CS cancer on TPM-biopsy.

For the comparison of TRUS-guided biopsy and mpMRI, McNemar’s test was used to compare the agreement between mpMRI (radiological score of≥3 assigned to definition 2 on the mpMRI CRF) and TRUS-guided

biopsy (definition 1) in the subset of men found to have CS prostate cancer on TPM-biopsy according to definition 1. For the secondary analysis, all analyses that were carried out for definition 1 were repeated for definition 2 and, at the request ofThe Lancetreviewers, for any Gleason score of≥7.

Exploratory analyses will be performed in the future to evaluate the impact on the sensitivity, specificity and predictive values of mpMRI when each of the 12 regions of interest are correlated between mpMRI and TPM-biopsy. Agreement between mpMRI and TPM-biopsy in identifying CS cancer in the same region will be based on a nearest neighbourhood approach.91Sensitivity to this approach will be tested by also

presenting results according to complete match (the most stringent rule) and using a left/right rule (the less stringent rule). The sensitivity, specificity and predictive values of each of the individual MRI reporting sequence combinations, namely T2, T2+DW and T2+DW+DCE, will inform additional important tertiary

analyses in subsequent publications.

Odds ratios (ORs) represent the odds of each test correctly detecting the presence or absence of disease. For specificity and NPV, the coding logic was reversed as the correct test result is a negative test result. Ratios are presented for TRUS-guided biopsy relative to mpMRI, so ratios of>1 favour TRUS-guided biopsy

and ratios of<1 favour mpMRI.

TABLE 5 2 × 2 tables to demonstrate the accuracy of mpMRI with respect to TPM-biopsy

TPM-biopsy mpMRI +ve ve Total +ve a b a+b –ve c d c+d Total a+c b+d

Chapter 3

Results

T

he main results of PROMIS have been published by Ahmedet al.92Sections of this chapter have been

reproduced from Ahmedet al.92© 2015 The Authors. Published by Elsevier Inc. This is an Open Access

article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

In document Multiparametric MRI to improve detection of prostate cancer compared with transrectal ultrasound-guided prostate biopsy alone : the PROMIS study (Page 53-56)