Prostate weight is the preferred measure of prostate size in radical prostatectomy cohorts

Transcription

1 Prostate weight is the preferred measure of prostate size in radical prostatectomy cohorts Matthew K.H. Hong, Henry H.I. Yao, Kathryn Rzetelski-West, Benjamin Namdarian, John Pedersen *, Justin S. Peters, Christopher M. Hovens and Niall M. Corcoran Division of Urology, Department of Surgery, University of Melbourne, Royal Melbourne Hospital, Parkville and Australian Prostate Cancer Research Centre, Victoria, Australia, * TissuPath Specialist Pathology, Mount Waverley, and Monash University Faculty of Medicine, Victoria, Australia Accepted for publication December 11 M.K.H.H., supported by scholarships from the Royal Australasian College of Surgeons and from the National Health and Medical Research Council, Australia OBJECTIVE To evaluate the accuracy of calculated prostate volume variables in a radical prostatectomy (RP) cohort, as many recent studies use these measures of prostate size instead of prostate weight. To determine whether this accuracy could be improved by modifying the mathematical model used in the volume estimation. PATIENTS AND METHODS Patients who underwent RP for prostate cancer at our associated institutions had calculated specimen volumes and weights from RP specimens determined at one pathology institution and transrectal ultrasonography (TRUS) volumes were recorded preoperatively ( n = 236). Correlation analysis was performed and errors were determined for calculated volume variables when compared with prostate weight. Bland-Altman plots were drawn and concordance coefficients calculated. Analysis was repeated with smaller prostates mathematically modelled as bullet-shaped rather than ellipsoid ( n = 165). RESULTS Although correlation was good for both TRUS and specimen volumes, they equally underestimated prostate weight with a large range of errors and poor concordance coefficients. Only 22% of TRUS volumes and 11% of calculated specimen volumes were within 1% of weight measurements. Application of a bullet-shaped mathematical model for prostates <55 g did not correct the large individual variation seen within these values. CONCLUSION Calculated prostate volume variables are prone to a large range of individual error regardless of the mathematical model used and should be avoided in statistical studies involving RP cohorts, and the more accurate prostate weight variable should instead be used as a size variable or correction factor. KEYWORDS prostatic neoplasms, radical prostatectomy, prostate weight, prostate volume, transrectal ultrasonography INTRODUCTION In urological studies involving the retrospective review of large radical prostatectomy (RP) cohorts, accurate measure of prostate gland size is important either as a standalone variable or as a correction factor for a number of prognostic variables in prostate cancer, e.g. percentage involvement of prostate cancer or PSA density (PSAD) [1 5 ]. Several measures of prostate size are available in these cohorts, and despite concerns about the accuracy of some measures [6,7 ], examples of each type can be found in publications [8 1 ]. RP specimens typically yield two different measures of prostate size: volume and weight. Whereas prostate weight is a direct measurement on scales, prostate volume is routinely derived by applying the prolate ellipsoid formula to three measured dimensions of the prostate [6 ]. A third measure of prostate size is the preoperatively measured TRUS volume, which similarly approximates prostate size based on in situ length measurements [11 ]. Other measures of prostate size, e.g. planimetric estimation using MRI, are relatively new and are not available in large RP cohorts, which often include many patients entered before the widespread use of this technology. Despite the accuracy and availability of prostate weight in RP cohorts [12 ], TRUS volumes or calculated specimen volumes are still used in many recent studies [4,9,1,13 15 ]. Whilst prostate weight and volume are theoretically approximately interchangeable as the specific gravity of prostate tissue is 1.5 [ 7 ], one is a single true measurement and the other is mathematically calculated. The inherent inaccuracy of calculated volume variables may thus potentially affect the validity of 12 THE AUTHORS 12 19, SUPPLEMENT 3,

2 HONG ET AL. TABLE 1 Patient demographics and baseline data Variable Value Number of patients 236 Mean ( SD ): age, years 61.1 (5.4) body mass index, kg/m (4.6) N (%): Pathological stage T2a 27 (11) T2b 23 (1) T2c 1 (51) T3a 56 (24) T3b 1 (4) Gleason Score 6 48 () (71) 8 9 (4) 9 11 (5) Median (range, IQR): TRUS volume, ml 37 (11 138, 19) specimen volume, ml 32 (1 13, 22) prostate weight, g 46 (18 126, ) the conclusions drawn in studies that rely on their accuracy [ 6,7 ]. For example, a study comparing TRUS volume calculated PSAD with prostate weight calculated PSAD showed the latter to be more strongly associated with adverse pathological features [3 ]. As previous studies suggest the ellipsoid formula to be the primary underlying factor contributing to the inaccuracy of prostate volume estimation [6 ], there have been attempts to improve accuracy by modifying the modelling formula used. In a group of brachytherapy patients, use of a bullet-shaped model has been shown to improve the accuracy of TRUS volume estimation [16 ]. In the present study, we first aimed to explore the general accuracy of TRUS and calculated specimen volumes as compared with prostate weight in a contemporary RP cohort and then attempted to improve their accuracy by applying a different mathematical model. PATIENTS AND METHODS Between 4 and 8, 856 men underwent RP for prostate cancer by participating surgeons at our affiliated TABLE 2 Correlation coefficients between calculated prostate volumes and prostate weights Group N TRUS volume vs weight, ρ Specimen volume vs weight, ρ TRUS volume vs specimen volume, ρ All TRUS volume, ml < Specimen volume. ml < ρ, Spearman s correlation coefficient; P <.1 for all correlation coefficients. institutions. Relevant clinicopathological data including age, preoperative PSA concentration and specimen pathology data were recorded prospectively in an electronic database. Within this group, 236 men had TRUS volume estimation recorded alongside pathological variables and were retrospectively recruited into this study. All TRUS volume estimations were performed by one experienced operator preoperatively whilst postoperative RP specimen dimensions and weight were reported by one pathology provider with pathologists adhering to a single protocol in specimen processing. In particular, specimen weight was determined on digital kitchen scales after formalin fixation and removal of the seminal vesicles. TRUS prostate volumes were calculated by using the prolate ellipsoid formula on dimensions obtained on US (width height length π /6) [11 ]. The same formula was applied to dimensions measured on final RP specimens to obtain a calculated specimen volume. For men with prostates weighing <55 g, we also calculated derived prostate volumes from both the TRUS and specimen dimensions using a formula that models the prostate as bullet shaped rather than ellipsoid (width height length π /4.8) [16 ]. Each volume variable was compared with its matched prostate weight. Absolute error in a volume variable was defined as the difference between volume and weight, with the percentage error being the error expressed as a percentage of the original volume. Spearman s rank correlation coefficients were calculated to compare volumes with weight, as some of the data were non-normally distributed and this would also allow comparison with previous studies. Concordance between variables was examined with Bland-Altman plots and concordance correlation coefficients for each volume variable compared with weight was calculated [17 19 ]. Statistical significance was defined as α <.5, and all tests performed were two-tailed. RESULTS In the present cohort of 236 men, the mean ( SD ) age was 61.1 (5.4) years and the mean ( SD ) body mass index was 27.5 (4.6) kg/m 2 ( Table 1 ). The median (interquartile range, IQR) prostate weight was 46 () g, whilst both TRUS volume (37 [IQR 19 ] ml) and specimen volume (32 [IQR 22 ] ml) had lower median values. Scatterplot examination of each volume variable against weight showed linear relationships with Spearman correlation coefficients of ρ.739 ( P <.1) for TRUS volume vs weight, and ρ.6 ( P <.1) for specimen volume vs weight ( Fig. 1 ). When we restricted the group to exclude outliers > ml and < ml, there were no improvements in the correlations (Table 2 ). However, comparison between specimen volume with weight consistently showed a marginally better correlation than the comparison between TRUS volume and weight, even when excluding outliers. All correlations between the two volume variables were modest. We then examined the actual differences between volume estimation and measured prostate weights in individual men across the cohort. Both volume variables tended to underestimate prostate weight and there THE AUTHORS 12

3 PROSTATE WEIGHT IS THE PREFERRED MEASURE OF SIZE FIG. 1. Scatterplot examination of correlations between TRUS volume, calculated specimen volume and prostate weight. (A) Correlation between TRUS volume and prostate weight; (B) Correlation between calculated specimen volume and prostate weight; (C) Correlation between TRUS volume and calculated specimen volume. A B C TRUS Prostate Volume, cc 1 Specimen Volume, cc 1 TRUS Prostate Volume, cc 1 1 Prostate Weight, g Prostate Weight, g Specimen Volume, cc 1 TABLE 3 Accuracy of TRUS volumes and calculated prostate specimen volumes compared with prostate weight TRUS volume vs prostate weight Specimen volume vs prostate weight TRUS volume vs specimen volume Number of patients Median (range, IQR) error, ml 9 ( 62 to 39, 13) 13.5 ( 52.3 to 27, 11.3) 4.8 ( 44.5 to 64.3, 15.7) 1 ml 129 (55) (34) 127 (54) ml 212 (9) 183 (78) 196 (83) Median (range, IQR) error, % 25 ( to, 38) 45 ( 386 to 44, 46) 12 ( 321 to, 42) 1% 51 (22) 25 (11) 49 (21) % 95 () 56 (24) 94 () Negative values are underestimations. was a wide range of individual error (Table 3 ). TRUS volume underestimated prostate weight by a median (IQR) of 9 (13) ml, although 55% were within 1 ml of the prostate weight. Specimen volume underestimated prostate weight by a median (IQR) of 13.5 (11) ml, but only 34% were within 1 ml. Although smaller prostates tended to account for the greatest range of errors, when glands <25 ml were excluded the number within 1 ml of prostate weight increased only to 58% for TRUS and 38% for specimen volume. In percentage terms, the range of individual error was similarly broad and only a minority of values fell within 1% of measured prostate weight (22% for TRUS, 11% for specimen volume). When error was defined as beyond % under or overestimation, the accuracy was not improved for many of the cases (% for TRUS and 24% for specimen volumes). Comparison between the two volume variables showed a surprising disparity. Although the median (IQR) difference was just 4.8 (15.7) ml, the disparity ranged from 45 ml underestimation to 64 ml overestimation. Just 54% were within 1 ml of each other, and only % were approximately concordant with an accepted accuracy within %. The examination of discrepancy between volume and weight variables with Bland- Altman plots and calculation of concordance coefficients showed large errors and poor agreement (Fig. 2 ). TRUS volume differed from prostate weight by 8.9 ml (95% confidence limit of agreement of 32.1 to 14.2), and had a concordance correlation coefficient of.689. The plot for specimen volume was similar with a bias of 13.4 ml (95% confidence limit of agreement of 34.5 to 7.7) and a concordance correlation coefficient of.613. In an attempt to improve accuracy by modelling smaller prostate glands as bullet shaped [ 16 ], we restricted the cohort to 165 men whose prostate glands were <55 g. We applied to their specimen and TRUS dimensions a formula that approximates the prostate gland as bullet shaped rather than ellipsoid and performed a similar analysis. Although there was an apparent 12 THE AUTHORS 12 59

4 HONG ET AL. TABLE 4 Accuracy of TRUS volumes and calculated prostate specimen volumes in small prostate glands (under 55 g) using formula for bullet shape compared with ellipsoid TRUS volume Specimen volume Ellipsoid Bullet Ellipsoid Bullet Number of patients Median (range, IQR) error, ml 7.8 ( 43 to 39, 11).1 ( 41 to 55, 14) 12 ( to 27, 9.8) 5.6 ( 37 to 44, 12) 1 ml 1 (61) 115 (7) 63 (38) 1 (61) ml 156 (95) 152 (92) 143 (87) 159 (96) Median (range, IQR) error, % 26 ( to, ).6 ( 3 to 68, 32) 5 ( 386 to 44, 52) 19 ( 286 to 55, ) Range 1% 32 (19) 56 (34) 15 (9) 38 (23) % 65 (39) 13 (62) 37 (22) 71 (43) Concordance coefficient (compared with weight) Negative values are underestimations. improvement across many parameters, the actual rates and range of individual error remained high, the Bland-Altman plots showed large discrepancies between volume and weight, and concordance correlation coefficients remained poor (Table 4, Fig. 3 ). DISCUSSION Accurate measures of prostate gland size from RP specimens are important in the clinical and academic settings. The accuracy of prostate volume estimations compared with measured weight or volume has been the subject of several studies, but these have involved either small cohorts, only explored one algorithm of prostate volume estimation, lack relevance in the RP cohort, or limit analysis to correlations rather than concordance at an individual level [6,7,12,16 ]. Many recent studies involving RP cohorts still use TRUS volumes or calculated specimen volumes to represent prostate gland size [4,9,1,13 15 ] and we sought to determine the general comparability of calculated volumes when compared with prostate weight and to see if this might be improved by alterations to the mathematical modelling. In the present study of 236 men from a contemporary RP cohort, we found that although prostate volumes calculated from both TRUS and the ex vivo specimen correlated well with prostate weight, they were both individually inaccurate with poor FIG. 2. Bland-Altman plots comparing error values of calculated prostate volumes compared with prostate weight. (A) TRUS-volume error compared with prostate weight; (B) Specimen volume compared with prostate weight. ULA, upper confidence limit of agreement; LLA, lower confidence limit of agreement. A 95% ULA 14.2 Bias % LLA concordance overall. Calculated specimen volumes underestimated weight by a median (IQR) of 13.5 (11.3) ml or 45% (46%) whilst TRUS volumes underestimated weight by a median (IQR) of 9 (13) ml or 25% (38%). These figures were similar to those described by Sajadi et al. [7 ] in their series of 497 men examining TRUS volumes, and suggests that although the use of one experienced US operator provided minor improvement in the accuracy of TRUS volume, it was not a major factor in overall accuracy. Perhaps more significant was the large range of individual error between the volume variables and weight, with TRUS values having the wider range of percentage error compared with specimen volume. With B 95% ULA 7.7 Bias % LLA 34.5 error defined as beyond 1% agreement, only 11% of calculated specimen volumes were in approximate concordance with weight, and only 22% of TRUS volumes were in this category. Although specimen volume correlated more strongly with prostate weight than did TRUS volume, and it might be expected that measurement on an excised specimen is less prone to subjective operator error than in situ estimation of prostate dimensions, the ex vivo volume variable did not appear to be better than TRUS volume overall. This is a similar finding to a series by Rodriguez et al. [6 ] and given that one pathology provider and one sonographer were used in this study for greatest consistency, this suggests that the greatest source of variation 12 THE AUTHORS 12

5 PROSTATE WEIGHT IS THE PREFERRED MEASURE OF SIZE B 95% ULA 12.1 Bias % LLA 26.4 Bias % LLA 29.6 D 95% ULA 24.1 Bias % LLA 21.8 between calculated volumes and measured prostate weight arises from the limitations of the mathematical formula used in the former. The idea that the accuracy of calculated prostate volumes is most limited by the commonly used prolate ellipsoid formula is not new [6], but the reasons that underlie this are seldom discussed. The formula models the prostate as an ellipsoid solid and is inflexible to variations in shape. Recent experience with in situ imaging in 138 men suggests that prostate shape is quite variable once the gland reaches a size of 55 ml and therefore any fixed formula disregarding shape is prone to error, particularly with larger prostates [16]. In that same study, the authors reported that prostates <55 ml were consistently bullet-shaped rather than ellipsoid and showed significant improvements in TRUS volume estimations using an adjusted formula. However, whilst our present attempt to improve estimations using the same formula in prostate glands of <55 ml resulted in some improvements, it failed to 95% ULA 3.7 Bias.85 95% LLA 2. 95% ULA 18 Bias % LLA % ULA 6.5 C FIG. 4. Bland-Altman plot showing nearequivalence between prostate weight with true gland volume calculated by water displacement (Data from Varma and Morgan [12], Spearman ρ.98, concordance coefficient.998). ULA, upper confidence limit of agreement; LLA, lower confidence limit of agreement. A FIG. 3. Bland-Altman plots of error values from calculated prostate volumes for glands < 55 g comparing the prolate-ellipsoid formula with the bullet-shape formula. (A) TRUS volume errors (ellipsoid); (B) Specimen volume errors (ellipsoid); (C) TRUS volume errors (bullet-shaped); (D) Specimen volume errors (bullet-shaped). ULA, upper confidence limit of agreement; LLA, lower confidence limit of agreement. improve the large range of individual error and poor concordance with weight overall. It is likely that no single formula applied to three dimensions will produce highly accurate estimations of prostate volume due to the variable shape of the prostate. The gold-standard measure of prostate volume in the RP setting is the direct measurement of water displacement by the gland. This avoids the issue of gland shape and provides a consistent, one-dimensional measurement similar to the way prostate weight is determined. However, as water displacement is not routinely performed and the utility of cohorts with long follow-up times is often mandated by the lengthy clinical course after RP [], it therefore remains an ideal but largely unavailable measure of prostate size in contemporary RP datasets. Instead, prostate weight has been shown to be an excellent surrogate of true prostate volume. In a prospective study of excised prostate glands, Varma and Morgan [12] showed that prostate weight was virtually equivalent to true volume determined by water displacement. Over a wide range of glands weighing from 29 to 158 g, they reported an error rate for weight from 5.4 to 5.6% and absolute error ranged from 2 to 3 ml (Fig. 4). In addition, they showed that calculated specimen volume was a far inferior predictor of true volume with error ranging from 52.8 to 5%. The accuracy and consistency of prostate weight arises from its reliance upon a simple, reproducible and unidimensional measurement. It is routinely reported and the data is therefore readily available alongside specimen dimensions and TRUS volumes in RP cohorts. The large errors and poor concordance between calculated volumes and prostate weight in the present study and the work of others [6,7] suggests that prostate weight is the most accurate and precise available measure of prostate size in the RP setting. To this end, we would recommend that it be used in preference over calculated volumes as a standalone variable or as a correction factor in PSAD or percentage tumour involvement, as the high prevalence of inaccuracies in calculated specimen volume data might adversely influence the findings of studies that use these variables by increasing the probability of type 2 statistical errors. In clinical trials or studies where the prostate gland remains in situ, measurements able to account for prostate shape are likely to prove better than calculated volumes. Early work using planimetric measures of prostate volume with MRI have already shown greater precision [21 23], and the use of 12 THE AUTHORS 12 61

6 HONG ET AL. three-dimensional US probes for TRUS are reportedly providing better gland volume estimations [24 26 ]. However, even if these new volume methods establish a role in the pre-rp setting, the long follow-up times required in prostate cancer research and the time taken for new technologies to reach widespread clinical use will mean that accurate prostate volume data will take some years to populate clinical databases of RP cohorts. Despite its potential limitations that include the absence of blood and other fluids in the specimen, the tissue shrinkage after fixation and the potential confounding effects of the presence of malignancy, prostate weight will continue to serve as an excellent measure of prostate size for some time. In conclusion, measures of prostate gland size are important in RP patients independently or as correction factors for derived variables such as PSAD or percentage tumour involvement. The present study suggests that TRUS volume and calculated specimen volume are both equally inaccurate as measures of prostate size regardless of the mathematical formula used. Prostate weight should instead be used in preference as a variable or correction factor and this should be considered when appraising studies that involve large retrospective RP cohorts. ACKNOWLEDGEMENTS M.K.H.H. is supported by scholarships from the Royal Australasian College of Surgeons and from the National Health and Medical Research Council, Australia CONFLICT OF INTEREST None declared. REFERENCES 1 Uhlman MA, Sun L, Stackhouse DA et al. Tumor volume, tumor percentage involvement, or prostate volume: which is predictive of prostate-specific antigen recurrence? Urology 1 ; 75 : Ramos CG, Roehl KA, Antenor JA, Humphrey PA, Catalona WJ. Percent carcinoma in prostatectomy specimen is associated with risk of recurrence after radical prostatectomy in patients with pathologically organ confined prostate cancer. J Urol 4 ; 172 : Radwan MH, Yan Y, Luly JR et al. Prostate-specific antigen density predicts adverse pathology and increased risk of biochemical failure. Urology 7 ; 69 : Magheli A, Rais-Bahrami S, Trock BJ et al. Prostate specific antigen versus prostate specific antigen density as a prognosticator of pathological characteristics and biochemical recurrence following radical prostatectomy. J Urol 8 ; 179 : Chung BI, Tarin TV, Ferrari M, Brooks JD. Comparison of prostate cancer tumor volume and percent cancer in prediction of biochemical recurrence and cancer specific survival. Urol Oncol 2 11 ; 29 : Rodriguez E Jr, Skarecky D, Narula N, Ahlering TE. Prostate volume estimation using the ellipsoid formula consistently underestimates actual gland size. J Urol 8 ; 179 : Sajadi KP, Terris MK, Hamilton RJ et al. Body mass index, prostate weight and transrectal ultrasound prostate volume accuracy. J Urol 7 ; 178 : Jones TD, Koch MO, Bunde PJ, Cheng L. Is prostate-specific antigen (PSA) density better than the preoperative PSA level in predicting early biochemical recurrence of prostate cancer after radical prostatectomy? BJU Int 6 ; 97 : Song C, Seo S, Ahn H et al. Percent tumor volume predicts biochemical recurrence after radical prostatectomy: multi-institutional data analysis. Int J Clin Oncol 11 [Epub ahead of print ]. DOI: 1.17/s Hayashi N, Urashima M, Ikemoto I et al. Prostate-specific antigen adjusted for total prostatic tumor volume as a predictor for outcome after radical prostatectomy. Prostate Cancer Prostatic Dis 7 ; 1 : 5 11 Watanabe H, Igari D, Tanahashi Y, Harada K, Saito M. Measurements of size and weight of prostate by means of transrectal ultrasonotomography. Tohoku J Exp Med 1974 ; 114 : Varma M, Morgan JM. The weight of the prostate gland is an excellent surrogate for gland volume. Histopathology 1 ; 57 : Eastham JA, Kattan MW. Disease recurrence in black and white men undergoing radical prostatectomy for clinical stage T1-T2 prostate cancer. J Urol ; 163 : Ates M, Teber D, Gozen AS et al. Do tumor volume, tumor volume ratio, type of nerve sparing and surgical experience affect prostate specific antigen recurrence after laparoscopic radical prostatectomy? A matched pair analysis. J Urol 7 ; 177 : Sfoungaristos S, Perimenis P. PSA density versus risk stratification for lymphadenectomy-making decision in patients with prostate cancer undergoing radical prostatectomy. Int Urol Nephrol. 11 ; 43 : MacMahon PJ, Kennedy AM, Murphy DT, Maher M, McNicholas MM. Modified prostate volume algorithm improves transrectal US volume estimation in men presenting for prostate brachytherapy. Radiology 9 ; 25 : Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics 1989 ; 45 : Hilson A. Bland-Altman plot. Radiology 4 ; 231 : Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986 ; 1 : 37 1 Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999 ; 281 : Rahmouni A, Yang A, Tempany CM et al. Accuracy of in-vivo assessment of prostatic volume by MRI and transrectal ultrasonography. J Comput Assist Tomogr 1992 ; 16 : Lee JS, Chung BH. Transrectal ultrasound versus magnetic resonance imaging in the estimation of prostate volume as compared with radical prostatectomy specimens. Urol Int 7 : 78 : Jeong CW, Park HK, Hong SK, Byun SS, Lee HJ, Lee SE. Comparison of prostate volume measured by transrectal ultrasonography and MRI with the actual prostate volume measured after radical prostatectomy. Urol Int 8 : 81 : Kanao K, Kikuchi E, Nakashima J et al. Three-dimensional ultrasonography in THE AUTHORS 12

7 PROSTATE WEIGHT IS THE PREFERRED MEASURE OF SIZE evaluation of benign prostatic hyperplasia. Int J Urol 4 ; 11 : Hoffmann AL, Laguna MP, de la Rosette JJ, Wijkstra H. Quantification of prostate shrinkage after microwave thermotherapy: a comparison of calculated cell-kill versus 3D transrectal ultrasound planimetry. Eur Urol 3 ; 43 : Chin JL, Downey DB, Elliot TL et al. Three dimensional transrectal ultrasound imaging of the prostate: initial experience with an emerging technology. Can J Urol 1999 ; 6 : 7 6 Correspondence: Matthew Hong, Department of Surgery, University of Melbourne, Royal Melbourne Hospital, Grattan St Parkville, VIC 35, Australia. m.k.hong@ausdoctors.net Abbreviations : RP, radical prostatectomy ; PSAD, PSA density ; IQR, interquartile range. 12 THE AUTHORS 12 63