The accurate measurement of in vivo prostate volume has gained greater importance with the development of prostate brachytherapy. Prostate brachythera

Transcription

1 Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at Peter J. MacMahon, MRCPI Ann-Marie Kennedy, MSc Darra T. Murphy, MRCPI Michael Maher, FFR, RCSI Michelle M. McNicholas, FFR, RCSI, FRCR Modified Prostate Volume Algorithm Improves Transrectal US Volume Estimation in Men Presenting for Prostate Brachytherapy 1 Purpose: To evaluate the accuracy of the conventional transrectal ultrasonographic (US) prostate volume formula and determine whether a more accurate volume formula, calculated on the basis of prostate shape as observed at planimetry, can be described for the majority of prostate glands. ORIGINAL RESEARCH ULTRASONOGRAPHY 1 From the Department of Radiology, Mater Misericordiae University Hospital, Eccles Street, Dublin 7, Ireland (P.J.M., D.T.M., M. M. McNicholas); and Department of Radiotherapy, Mater Private Hospital, Dublin, Ireland (A.M.K., M. Maher). From the 2007 RSNA Annual Meeting. Received February 12, 2008; revision requested March 28; revision received June 25; accepted July 7; final version accepted July 30. Address correspondence to M. M. McNicholas ( michelle.mcnicholas@nbsp.ie ). Materials and Methods: Results: Conclusion: This retrospective study was institutional review board approved, with waiver of informed consent. A total of 138 consecutive patients who underwent prostate brachytherapy were included for analysis. Prostate volume was estimated by using the conventional prolate ellipsoid formula (length height width [ /6]). A reference standard for prostate volume was calculated by using planimetry. The mean prostate shape was assessed by using three-dimensional volume-rendering of the planimetric images. The prostate shapes were evaluated to determine the best-fit mathematic formula for accurate volume estimation. Statistical analyses were performed by using Pearson correlation, paired Student t test, Bland-Altman plots, and concordance correlation coefficient. Planimetric data showed the majority of prostate glands to be more bullet-shaped than ellipsoid. Only 13.3% of volumes determined by using the conventional prolate ellipsoid formula were within 10% of the planimetric volume. The prolate ellipsoid formula underestimated volume by 17% on average (95% confidence interval: 14%, 19%). A mathematic formula representing a bullet shape (length height width [ /4.8]) was determined to best represent the majority of prostate glands presented for brachytherapy; 75% of volumes were within 10% of planimetric volume by using this formula. Concordance correlation coefficient increased from 0.87 to Formula accuracy was particularly improved in prostate glands smaller than 55 cm 3 (P.14). A modified prostate volume formula that closely represents the shape of the prostate smaller than 55 cm 3 demonstrated improved volume measurement accuracy compared with the prolate ellipsoid formula used in men presenting for brachytherapy. RSNA, 2009 RSNA, 2009 Radiology: Volume 250: Number 1 January

2 The accurate measurement of in vivo prostate volume has gained greater importance with the development of prostate brachytherapy. Prostate brachytherapy is a form of radiation therapy where radioactive sources (seeds) are placed percutaneously by using a transperineal approach to the prostate gland under ultrasonographic (US) guidance to treat early prostate cancer. Patient selection for such therapy is dependent on the size of the prostate gland. Patients with large glands are frequently unsuitable for brachytherapy for technical reasons and because of an increased side-effect profile (1). Technical problems are largely a result of pubic arch encroachment over the prostate gland, and a cutoff size of 50 cm 3 is generally employed. Furthermore, in most cases, the radioactive seeds must be ordered in advance, and the number of seeds ordered is determined on the basis of the outpatient gland volume measurement. Recently, prostate-specific antigen density has been found to be a significant predictor of adverse pathologic features and biochemical recurrence in prostate cancer (2). It is calculated by dividing the serum prostate-specific antigen level by the prostate volume. Hence, it is dependent on accurate prostate volume measurement. Conventionally, the prostate gland volume is assessed as an outpatient procedure at the time of transrectal USguided biopsy retrieval by measuring three gland dimensions: the maximum length (L), the maximal height orthogonal to the length (H), and the maximal width (W). The volume is then calculated by using the prolate ellipsoid formula: length height width [ /6] (3). Advance in Knowledge A modified prostate volume algorithm, which describes a bulletshaped prostate gland, as opposed to the conventional prolate ellipsoid shape, significantly improves volume measurement accuracy in men presenting for prostate brachytherapy. The number of radioactive seeds placed at brachytherapy is critically dependent on the prostate volume (4). The seeds must also be implanted accurately in a distribution that optimizes radiation dose, while avoiding overexposure of critical structures such as the urethra and rectum. Therefore, immediately prior to brachytherapy the prostate volume is accurately measured by using a planimetric transrectal US stepping method, which is considered the reference standard for measuring the prostate volume in vivo. Our prostate brachytherapy patients are predominately referred from outside institutions, having already undergone prostate volume estimation at the referring institution. We have noted a discrepancy between the outpatient transrectal US measurements and our prebrachytherapy planimetric measurement. This has, on occasion, led to termination of the brachytherapy procedure after the patient has received a general anesthetic, owing to the actual volume being too large for brachytherapy treatment. Apart from unnecessary exposure to general anesthetic and patient disappointment at the abandonment of the chosen form of treatment, there is also waste of the expensive radioisotope sources, which rapidly lose their optimum activity (5). An improvement in prostate volume estimation would also allow improved accuracy for prostate-specific antigen density measurements. Our purpose is to evaluate the accuracy of the conventional transrectal US prostate volume formula and determine whether a more accurate volume formula, calculated on the basis of prostate shape as observed at planimetry, can be described for the majority of prostate glands. Implication for Patient Care The current inaccurate method of prostate estimation can be improved simply and significantly, thus reducing inappropriate patient selection for prostate brachytherapy. Materials and Methods Patients Institutional ethics committee approval was obtained for this retrospective study, with waiver of informed consent. All patients referred for prostate brachytherapy as a treatment for localized prostate cancer in the last 2 years at our institution were included for analysis. All patients had been assessed as outpatients for an increased prostate-specific antigen level ( 4 ng/ml [ 4 g/l]). Patients were diagnosed with prostate carcinoma after transrectal US-guided core biopsy. All patients were staged as having localized disease and were suitable for iodine 125 brachytherapy. Sixty patients were retrospectively analyzed to assess accuracy of outpatient volume estimation by using the conventional prolate ellipsoid formula by comparing it with the volume calculated by using the planimetric stepping method. An analysis was made to determine the shape of the majority of glands and to derive a best-fit mathematic formula on the basis of the observed shape. Subsequently, the next 78 patients referred for brachytherapy were assessed by comparing the volume calculated in two methods from the same data set: by using the conventional prolate ellipsoid formula and by using a mathematic formula given the observed shape. The reference standard for prostate volume measurement was obtained at the time of brachytherapy by using a planimet- Published online /radiol Radiology 2009; 250: Author contributions: Guarantors of integrity of entire study, P.J.M., M. McNicholas; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, all authors; clinical studies, all authors; statistical analysis, P.J.M.; and manuscript editing, all authors Authors stated no financial relationship to disclose. 274 Radiology: Volume 250: Number 1 January 2009

3 ric stepping method, as described below. The actual outpatient transrectal US prostate volumes were not included for analysis as these data were obtained by several radiologists working at several institutions by using a variety of US equipment and uncertain techniques. All measurements were obtained from a single US machine (EUB 5500; Hitachi Medical, Tokyo, Japan), with an 8-MHz side-firing biplanar transrectal probe (EUP-U533; Hitachi Medical, Tokyo, Japan). This biplane probe incorporates both side-firing radial and linear mutually perpendicular transducers. The patient was placed on a table in the dorsolithotomy position and legs placed in stirrups. The reference prostate volume was computed with planar measurements obtained by using 5-mm step sections with the probe mounted in a stepping device (EXII; CIVCO, Kalona, Iowa). The circumference of each prostate image was traced by a staff radiologist (M. M. McNicholas, with 15 years experience in prostate imaging). The US software computed the volume from the measurements of each section (Variseed, version 7.1; Varian Medical Systems, Charlottesville, Va). The three prostate dimensions (length, height, and width) used in the prolate ellipsoid formula were measured independently by three radiologists (P.J.M., D.T.M., and M. M. McNicholas, each with 2 15 years experience) on the US images obtained at planimetry. The maximal length, the maximal height measured orthogonal to the sagittal length, and the maximal width were measured by using the images reconstructed from planimetric data. The conventional transrectal US volume estimation was calculated by using the prolate ellipsoid formula. This conventional volume calculation was compared with the reference standard volume as assessed at planimetry. Shape Analysis The prostate glands in the initial cohort of 60 patients were volume rendered in three dimensions by the planimetry software (Fig 1a) and visual assessment of the rendered prostate shape was performed (Fig 2a). To establish the optimal volume formula, various geometric three-dimensional shapes (ie, prolate ellipsoid, cylinder, cube, and cone) were combined in an effort to replicate the rendered prostate shape. Mathematic formulas were derived for each shape that could be simplified to a three-dimensional planar measurement technique (length height width). Shape combinations that were assessed included: one-half ellipsoid plus cone, one-half ellipsoid plus cube, two-thirds ellipsoid plus cylinder, two-thirds ellipsoid plus cone, one-third ellipsoid plus cylinder, and one-third ellipsoid plus Figure 1 cube. For each combination, a single mathematic volume formula was derived and compared with the reference data set by using the statistical tests described below. A single three-dimensional shape composed of a cylinder and one-half of a prolate ellipsoid (a bullet shape [Fig 2b]) demonstrated the closest statistical concordance and is presented here. We observed this to be the case for glands of up to about 55 cm 3 in volume. Above this volume, we observed that the shape of the glands was more unpredictable. The bullet formula for determining Figure 1: (a) Sample of volume-rendered US images of prostate glands measuring less than 55 cm 3 from initial patient cohort. (b) Prostate glands more than 55 cm 3 have unpredictable shapes secondary to asymmetric lobar enlargement. Figure 2 Figure 2: (a) Three-dimensional rendering of prolate ellipsoid shape assumed for prostate to use with conventional volume formula. (b) New bullet shape combines cylinder and one-half prolate ellipsoid. Radiology: Volume 250: Number 1 January

4 prostate shape was then assessed in the next 78 patients referred for brachytherapy to validate accuracy. Most (55 of 60) of these patients had glands that were less than 55 cm 3 ; a volume of less than 50 cm 3 at outpatient transrectal US was one of the referral criteria. Statistical Analysis We examined the correlation between prolate ellipsoid volumes and planimetry by using the Pearson correlation coefficient, paired t test, linear regression, and Bland-Altman plots (6,7). We additionally used the concordance correlation coefficient as a method to compare two measurements (8). Interobserver reliability was demonstrated by using linear regression, Bland-Altman plots, and concordance correlation. Results The mean prostate volume, as calculated by using the conventional prolate ellipsoid formula with planimetric data sets in the initial 60 patients, was cm (standard deviation) (range, cm 3 ). The reference standard volume calculated by using the stepping method was cm (range, cm 3 ). The mean difference between techniques was 5.87 cm 3 (95% confidence interval: 5.01, 6.73). Thus, Figure 3 in this cohort, the prolate ellipsoid formula underestimated planimetric volume by 17% on average (95% confidence interval: 14%, 19%). Only 13% (eight of 60) of volumes recorded by using the prolate ellipsoid formula were within 10% of the reference planimetric volume. There was a significant difference between the measurement methods as assessed by using the paired Student t test (P.0001). When comparing the data for correlation, r 0.969, r (Fig 3a). The sample concordance correlation coefficient is in this cohort. The Bland-Altman plot demonstrates the discrepancy between the measurement techniques with a mean difference (bias) of 5.9 and a 95% confidence limit of agreement of 12.4 to 0.6 (Fig 3b). Our subjective assessment of threedimensional rendered data showed that the prostate shape demonstrated variation as the volume increased secondary to asymmetric lobar enlargement (Fig 1b). Below 55 cm 3, prostate glands tended to have a consistent shape. Above 55 cm 3, there was more variation in shape and less predictability in calculated volume. The mathematic best-fit volume formula for prostate glands smaller than 55 cm 3 is composed of onehalf of a prolate ellipsoid and onehalf of an elliptical cylinder (Fig 2b). This shape resembles a bullet, with the flat part corresponding to the base and the rounded end corresponding to the apex. For such a bullet shape, the volume formula computes as: length height width ( /4.8), or length height width When the bullet formula is applied to the same cohort of 60 patients, the mean prostate volume is cm (range, cm 3 ). The mean difference in volume measurement between the new bullet formula and planimetry is reduced to 1.46 cm 3 (95% confidence interval: 0.39, 2.53). Seventyfive percent (45 of 60) of volumes calculated by using the bullet formula are now within 10% of the planimetric volume. However, there is still a significant difference between the volumes obtained with the two methods by using the paired t test (P.008). When we analyzed only the prostate glands smaller than 55 cm 3 (55 of 60), there was no significant difference between the groups as assessed by using the paired t test (P.14). The Pearson correlation coefficient did not change significantly when compared with the prolate ellipsoid formula (r 0.967, r ) (Fig 4). However, the Lin concordance correlation coefficient increased from 0.87 to 0.95 when switching to the bullet formula. We subsequently applied the bullet Figure 3: (a) Scatterplot of volume data compares conventional prolate ellipsoid volume calculations with reference standard volume at planimetry. Identity line represents perfect agreement for data. (b) Bland-Altman diagram of same data set. Data should plot close to identity line if good agreement is seen between techniques. 276 Radiology: Volume 250: Number 1 January 2009

5 Figure 4 Figure 4: (a) Scatterplot with linear fit compares bullet formula to planimetry for all 60 patients initially evaluated. (b) Bland-Altman diagram of a. (c) Scatterplot of prostate glands measuring less than 55 cm 3 by using bullet formula. (d) Bland-Altman diagram of c. formula to the next 78 patients referred for brachytherapy. In this group, the Lin concordance correlation coefficient increased from 0.83 to Data plotted closer to the identity line (45 ) when the bullet formula was used (Fig 5). By using the prolate ellipsoid formula in this cohort, 17% (13 of 78) were within 10% of the planimetric volume. However, by using the bullet formula, 67% (52 of 78) were within 10% of the planimetric volume (Table). Interobserver measurement accuracy of prostate dimensions from planimetric data are demonstrated in the scatterplot in Figure 6. The concordance correlation coefficient for two independent observers is high at Discussion Currently, prostate volume is calculated at the time of outpatient prostate transrectal US-guided biopsy. While the accuracy of volume calculations by using a three-dimensional measurement method (length height width) will always be limited when compared with a planimetry-based method, it is a measurement that is used to clinically assess a patient s suitability for prostate cancer treatment; thus, its accuracy is important. There are a number of factors that have to be taken into account when assessing the capability of a method to help estimate actual prostate gland size. The re- Radiology: Volume 250: Number 1 January

6 Figure 5 producibility of measurements is clearly important and an essential factor in the use of the prolate ellipsoid formula (9). Prostate gland dimensions, determined by using nonplanimetric transrectal US, is the principal source of error that contributes to decreased precision. It can be minimized by employing a strict measurement technique; in particular, ensuring that the height of the gland is always measured in the sagittal plane orthogonal to the length measurement, reducing any so-called salami effect (10). This effect predominately occurs when the prostate height is measured from the axial view. The prostate section that is defined by the probe depends on the operator s angle relative to the long axis, which is extremely difficult to judge on an axial view alone. We have demonstrated that the prolate ellipsoid formula is not accurate at predicting prostate volume even when using measurements obtained from planimetric US images that were all performed by a single experienced operator in identical conditions (patient in lithotomy position receiving general anesthetic). To our knowledge, this is the first study to employ rigorous statistical techniques and methods to minimize confounding factors in prostate volume measurement. There are a number of published studies recommending the prolate ellipsoid method as a useful alternative to planimetry in outpatient transrectal US prostate volume measurement (3,11, 12). In particular, Littrup et al (3) used a similar biplanar probe for their threedimensional measurements in a cohort of patients with prostate volumes ranging from 20 to 100 cm 3. These previous studies demonstrated good Pearson correlation of transrectal US volumes with planimetry or radical prostatectomy volumes. However, the Pearson correlation coefficient only demonstrates whether a linear association exists between two variables. There will inevitably be good linear association between tests that measure the same variable, but this does not mean that there is good agreement between two measurements. In our study, we used two additional statistical tests. The Bland-Altman plot is specifically used to assess the agreement between two methods of clinical measurement. The concordance correlation coefficient published by Lin (8) has been shown to be a robust method at establishing the strength of agreement between two measurements. In contrast to the Pearson correlation coefficient, it contains a measure of how far a best-fit line deviates from the identity line through the origin. In this respect, it is superior to the Pearson correlation coefficient in this type of study. Of note, prostatectomy specimens are not a standard of reference in a study of prostate volume, as there is considerable prostate tissue shrinkage ex vivo (10,13). While the volume calculated at stepping planimetry is prone to some error owing to the use of 5-mm acquisition increments, it is regarded as an accurate method of measuring in vivo volumes (14). The use of reconstructed multiplanar images obtained at the time of planimetry allows for an accurate method of deriving dimensional measurement free from temporal- and technique-derived confounding factors. In our study, we calculated the transrectal US volumes from the height, length, and width measurements obtained at planimetry. To eliminate possible confounding variables, such as interobserver variation and volume change between outpatient transrectal US and brachytherapy treatment, the outpatient transrectal US volumes (obtained by numerous observers with different US equipment) were not used. In addition, any potential for discrepancy in prostate dimension measurement owing to patient positioning is eliminated in this study by comparing volumes while the patient remains in one position (in this case, dorsolithotomy). Our study demonstrates that the prolate ellipsoid formula underesti- Figure 5: Scatterplots of second patient cohort (n 78) referred for brachytherapy show (a) conventional prolate ellipsoid formula compared with planimetry and (b) bullet formula compared with planimetry. 278 Radiology: Volume 250: Number 1 January 2009

7 Overview of Statistical Results Prostate Parameter Mean Difference SD (ml) P Value* r Value Lin Concordance First 60 patients Prostate 55 cm 3 Prolate ellipsoid vs planimetry Bullet shape vs planimetry Prostate 55 cm 3 Prolate ellipsoid vs planimetry Bullet shape vs planimetry Next 78 patients Prostate 55 cm 3 Prolate ellipsoid vs planimetry Bullet shape vs planimetry Prostate 55 cm 3 Prolate ellipsoid vs planimetry Bullet shape vs planimetry Note. SD Standard deviation. * Calculated by using Student t test. Figure 6 Figure 6: Scatterplot of bullet volume calculations from two independent observers. Plotting of measurements close to identity line shows low interobserver error. mates the planimetric prostate volume by 17%, supporting the results of Nathan et al (15), who showed that planimetric volumes were 17% greater than those of outpatient transrectal US measurements. By examining the rendered prostate shape, we observed that the prolate ellipsoid shape underestimates the size of the prostate gland and thus intrinsically underestimates the prostate volume by using the measurements obtained. Our bullet formula compensates for this inaccuracy. Its application to the same prostate dimensions significantly improved the ability of a formulaic method to more accurately predict planimetric volume. This is demonstrated with improved Bland- Altman plots (Figs 5 and 6) and, in particular, improved Lin concordance correlation coefficient from 0.87 to We have validated the bullet formula by confirming its accuracy when compared with planimetry in the second set of 78 patients referred for brachytherapy. Overall, between 66% and 75% of prostate volumes are estimated within 10% of the planimetric volume by using the bullet formula, as opposed to between 13% and 17% by using the prolate ellipsoid method. More importantly, for glands that were less than 55 cm 3 in volume, the bullet formula is highly accurate. This is relevant for patients considering brachytherapy, since 50 cm 3 is the approximate cutoff volume of suitability for this form of treatment. Whereas the prostate volume is more than cm 3, patients are usually deemed unsuitable for brachytherapy, meaning that inaccuracy in precise measurement above this volume is not clinically important. We recognize that the measurements used to calculate volume in this study with the patient lying in the lithotomy position are likely to be more accurate than measurements obtained at outpatient transrectal US when the patient is lying in a lateral position. Nonetheless, the bullet formula significantly improved agreement with the reference standard planimetric volume, as calculated by using the stepping method. Thus, it can be inferred that the bullet formula should also improve the accuracy of the outpatient transrectal US measurement. We also recognize that a single operator, experienced in prostate brachytherapy, is likely to make more accurate measurements than a radiologist who performs occasional outpatient transrectal US. Nonetheless, if the bullet formula significantly improved accuracy when the measurements were obtained by an experienced operator, it should do the same for the less-experienced radiologist. This study demonstrated an inherent error in the prolate ellipsoid technique by underestimating the size of the base of the prostate gland. Any volume Radiology: Volume 250: Number 1 January

8 calculation made on the basis of three measurements can only estimate actual size. However, if the calculation is important, it must be as accurate as possible. We have demonstrated that a significant improvement in prostate volume calculations is possible by using a volume formula computed to better represent the majority of prostate shapes in men suitable for prostate brachytherapy. We suggest that the current volume calculation: length height width ( / 6) (or, length height width 0.52) be replaced by the bullet formula: length height width ( /4.8) (or, length height width 0.65), especially in prostate glands of less than 55 cm 3 in volume. References 1. Langley SE, Laing R. Prostate brachytherapy has come of age: a review of the technique and results. BJU Int 2002;89(3): Radwan MH, Yan Y, Luly JR, et al. Prostatespecific antigen density predicts adverse pathology and increased risk of biochemical failure. Urology 2007;69(6): Littrup PJ, Williams CR, Egglin TK, Kane RA. Determination of prostate volume with transrectal US for cancer screening. II. Accuracy of in vitro and in vivo techniques. Radiology 1991;179(1): Al-Qaisieh B, Brearley E, St Clair S, Flynn A. A study of a pretreatment method to predict the number of I-125 seeds required for prostate brachytherapy. Int J Radiat Oncol Biol Phys 2006;65(1): Armpilia CI, Dale RG, Coles IP, Jones B, Antipas V. The determination of radiobiologically optimized half-lives for radionuclides used in permanent brachytherapy implants. Int J Radiat Oncol Biol Phys 2003;55(2): Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1(8476): Hilson A. Bland-Altman plot [letter]. Radiology 2004;231(2): Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics 1989; 45(1): Bazinet M, Karakiewicz PI, Aprikian AG, et al. Reassessment of nonplanimetric transrectal ultrasound prostate volume estimates. Urology 1996;47(6): Dähnert WF. Determination of prostate volume with transrectal US for cancer screening [letter]. Radiology 1992;183(3): Terris MK, Stamey TA. Determination of prostate volume by transrectal ultrasound. J Urol 1991;145(5): Kimura A, Kurooka Y, Hirasawa K, Kitamura T, Kawabe K. Accuracy of prostatic volume calculation in transrectal ultrasonography. Int J Urol 1995;2(4): Jonmarker S, Valdman A, Lindberg A, Hellström M, Egevad L. Tissue shrinkage after fixation with formalin injection of prostatectomy specimens. Virchows Arch 2006;449(3): Miyashita H, Watanabe H, Ohe H, Saitoh M, Oogama Y, Iijima S. Transrectal ultrasonotomography of the canine prostate. Prostate 1984;5(4): Nathan MS, Seenivasagam K, Mei Q, Wickham JE, Miller RA. Transrectal ultrasonography: why are estimates of prostate volume and dimension so inaccurate? Br J Urol 1996;77(3): Radiology: Volume 250: Number 1 January 2009