There is an emerging interest in the preoperative assessment of MRI pelvic floor measurements as possible predictors of post-prostatectomy incontinence. However, there is still no evidence about the reliability of these measurements or whether they are reliable when carried out by a urologist without previous experience in prostate MRI.
The objective of this research is to assess the interobserver agreement between two experts in pelvic floor imaging (radiologists), and two observers with less experience (urologists) for different measurements of the pelvic floor in preoperative prostate MRI.

Fifty-seven patients from our center who had undergone prostate MRI prior to robotic radical prostatectomy were selected by consecutive sampling.

Measurements were made by each observer blindly. The measurements were: membranous urethral length (MUL), intravesical prostatic protrusion (IPP), levator-ani muscle thickness (LAM), puborectal muscle thickness (PBR), internal-obturator muscle thickness (OMT), prostate volume (PV), ratio LAM/PV, urethral width (UW), angle between membranous urethra and prostatic axis (aLUMP) and prostatic urethra length (PUL). 

The interobserver agreement for the different MRI measurements was determined, using two methods: the intraclass correlation coefficient (ICC) and the Bland-Altman graphical method. Statistical significance was established for p-value <0.05.
To determine the strength of the agreement with ICC, the proposal by Landis Koch was used, with the degree of agreement being poor (ICC=0), slight (ICC 0.01-0.20 ), fair (ICC 0.21- 0.40), moderate (ICC 0.41-0.60), substantial (ICC 0.61-0.80), almost perfect (ICC 0.81-1.00).
Finally, the median of the agreement scores for each measurement and each combination of observers was calculated to obtain a summary measure.

Mean preoperative MRI anatomical measurements of each observer are summarized in TABLE 1.

Results of inter-observer agreement for pelvic floor measurements with Bland-Altman plots and ICC are summarized in TABLE 2.

A. INTER-OBSERVER CONCORDANCE FOR EACH MEASUREMENT WAS:

 1) MUL
o   Urologist 1-radiologist 1: Moderate; 
o   Urologist 2-radiologist 1: Fair; 
o   Urologist 1-urologist 2: Moderate; 
o   Urologist 1-radiologist 2: Moderate; 
o   Urologist 2-radiologist 2: Moderate; 
o   Radiologist 1-radiologist 2: Moderate; 

 Median concordance for MUL was MODERATE
 
2) IPP
o   Urologist1-radiologist1: Substantial; 
o   Urologist2-radiologist1: Moderate; 
o   Urologist1-urologist2: Substantial; 
o   Urologist1-radiologist2: Moderate; 
o   Urologist2-radiologist2: Substantial; 
o   Radiologist1-radiologist2: Substantial; 

Median concordance for IPP was SUBSTANTIAL.
 
3)  LAM
o   Urologist1-radiologist1: Fair; 
o   Urologist2-radiologist1: Slight; 
o   Urologist1-urologist2: Fair; 
o   Urologist1-radiologist2: Slight; 
o   Urologist2-radiologist2: Slight; 
o   Radiologist1-radiologist2: Fair; 

Median concordance for LAM was FAIR-SLIGHT.
 
4)  PBR
o	Urologist1-radiologist1: Slight; 
o	Urologist2-radiologist1: Fair; 
o	Urologist1-urologist2: Slight; 
o	Urologist1-radiologist2: Poor; 
o	Urologist2-radiologist2: Poor;
o	Radiologist1-radiologist2: Poor;

Median concordance for PBR was SLIGHT
 
5)       OMT
o   Urologist1-radiologist1: Substantial; 
o   Urologist2-radiologist1: Fair; 
o   Urologist1-urologist2: Slight; 
o   Urologist1-radiologist2: Moderate; 
o   Urologist2-radiologist2: Fair; 
o   Radiologist1-radiologist2: Moderate; 

Median concordance for OMT was FAIR-MODERATE
 
6)       PV
o   Urologist1-radiologist1: Almost perfect; 
o   Urologist2-radiologist1: Substantial; 
o   Urologist1-urologist2: Substantial; 
o   Urologist1-radiologist2: Substantial; 
o   Urologist2-radiologist2: Substantial; 
o   Radiologist1-radiologist2: Substantial; 

Median concordance for PV was SUBSTANTIAL
 
7)  LAM/VP ratio
Urologist1-radiologist1: Moderate; 
o   Urologist2-radiologist1: Fair; 
o   Urologist1-urologist2: Fair; 
o   Urologist1-radiologist2: Fair; 
o   Urologist2-radiologist2: Slight; 
o   Radiologist1-radiologist2: Moderate; 

Median concordance for LAM/VP ratio was FAIR

8)   UW
o   Urologist1-radiologist1: Fair; 
o   Urologist2-radiologist1: Fair; 
o   Urologist1-urologist2: Moderate; 
o   Urologist1-radiologist2: Fair; 
o   Urologist2-radiologist2: Fair; 
o   Radiologist1-radiologist2: Fair; 

Median concordance for UW was FAIR.
 
9)       aLUMP
o   Urologist1-radiologist1: Fair; 
o   Urologist2-radiologist1: Moderate; 
o   Urologist1-urologist2: Substantial; 
o   Urologist1-radiologist2: Moderate; 
o   Urologist2-radiologist2: Moderate; 
o   Radiologist1-radiologist2: Moderate; 

Median concordance for aLUMP was MODERATE.
 
10)       PUL:
o   Urologist1-radiologist1: Fair; 
o   Urologist2-radiologist1: Moderate; 
o   Urologist1-urologist2: Fair; 
o   Urologist1-radiologist2: Moderate; 
o   Urologist2-radiologist2: Fair; 
o   Radiologist1-radiologist2: Moderate; 

Median concordance for PUL was FAIR-MODERATE.
 

B.  MEDIAN INTER-OBSERVER CONCORDANCE FOR EACH COMBINATION OF OBSERVERS WAS:
o	Urologist 1-radiologist 1: FAIR-MODERATE
o	Urologist 2-radiologist 1:FAIR
o	Urologist 1-urologist 2: FAIR-MODERATE
o	Urologist 1-radiologist 2:MODERATE
o	Urologist 2-radiologist 2: FAIR-MODERATE
o	Radiologist 1-radiologist 2: MODERATE

The anatomical parameters with greater agreement have been the IPP and the PV. Most pelvic floor measurements have good or moderate interobserver agreement, except LAM thickness and puborectalis muscle thickness, the latter being one of the muscles that are part of the LAM complex. The LAM is a three-dimensional muscle made up of three muscle bundles and because of its complexity it is difficult to establish its measurements, so it does not seem to be a sufficiently reliable measurement to be used as a predictor variable for PPI. Therefore a standardized form of measurement of these muscles must be established.
The highest agreement was obtained for the two radiologists and for urologist-1 with the radiologist 1.The fact that the concordance was equally acceptable between urologist 1 and both radiologists seems to suggest that it is not necessary to have an extraordinary specialization to correctly carry out MRI measurements.

Most pelvic floor measurements have good or moderate interobserver agreement, except for LAM thickness and puborectalis muscle thickness.
For the latter, a standardized form of measurement must be established to improve their reliability.
It does not seem that an extraordinary specialization should be necessary to carry out the measurements correctly.