<?xml version="1.0" ?> <tei> <teiHeader> <fileDesc xml:id="0"/> </teiHeader> <text xml:lang="en"> <p>S ystematic biopsy of the prostate represents the<lb/> only definitive diagnostic modality capable of<lb/> confirming malignancy among men with palpably<lb/> and ultrasonically undetectable lesions. However,<lb/> uniform biopsy approaches such as the quadrant or<lb/> sextant methods may carry sampling limitations in<lb/> view of the wide variation in gland sizes and<lb/> shapes. The same number of cores used for biopsy,<lb/> regardless of individual gland characteristics, may<lb/> lead to more extensive sampling of small glands<lb/> and less extensive, even possibly suboptimal sam-<lb/>pling in large glands. Such sampling differences are<lb/> likely to be accompanied by differences in biopsy<lb/> yield, as indeed demonstrated by Uzzo et al. <ref type="biblio">1</ref> and<lb/> confirmed by others. <ref type="biblio">2–5</ref><lb/></p> <p>Despite agreement suggesting that sextant pros-<lb/>tatic biopsy may lead to suboptimal sampling of<lb/> large glands, no clear recommendations regarding<lb/> potential modifications to this approach have been<lb/> made. To further explore the relationship between<lb/> prostate size and biopsy outcome, we developed a<lb/> computer model allowing experimentation with<lb/> several important parameters potentially affecting<lb/> biopsy yield. For this purpose we assessed various<lb/> peripheral zone sector biopsy methods relying on<lb/> 4, 6, 8, 10, or 12 systematically distributed cores.<lb/> We determined their yields in detecting a single<lb/> spherical tumor lesion of various sizes: 0.3 cc rep-<lb/>resenting a clinically insignificant <ref type="biblio">6,7</ref> tumor vol-<lb/>ume, 0.5 cc in size or corresponding to cutoff vol-<lb/>ume of a clinically significant tumor, as well as for<lb/> larger tumor sizes, namely, 0.9 and 1.4 cc. Each of<lb/> those tumor and biopsy combinations was simu-<lb/>lated in prostates 20, 40, 60, 80, and 100 cc in size.<lb/></p> <head>MATERIAL AND METHODS<lb/></head> <p>We used a custom-built fully computerized application re-<lb/>lying on two independent but interacting modules. The first<lb/> module creates a three-dimensional prostate containing tu-<lb/>mor lesions according to provided specifications. The second<lb/> module represents the biopsy method. Both modules are<lb/> FORTRAN-compiled (Absoft Corporation, Rochester Hills,<lb/> Mich) and designed to run on a MacIntosh 601 PowerPC pro-<lb/>cessor platform (Apple Computer, Cupertino, Calif).<lb/></p> <p>The prostate module generates a three-dimensional model<lb/> of the gland. The gland is defined by a spherical formula. The<lb/> three prostatic sphere planar lengths represent free parameters<lb/> and are user defined.<lb/></p> <p>For simplicity, only two zones are used, defining the entire<lb/> gland: the peripheral zone (PZ) and the anterior zone. The<lb/> periphery is crescent shaped. Similar to planar measurements,<lb/> the anteroposterior (AP) dimension of the periphery (thick-<lb/>ness of the PZ simulating its appearance on ultrasound image<lb/> in sagittal plane) represents a free parameter. Consequently,<lb/> for any total gland volume, a uniquely sized PZ can be defined.<lb/> Peripheral and total gland volume calculations are performed<lb/> by the software allowing definition of any spherical approxi-<lb/>mation of the gland and are accompanied by a defined variant<lb/> of the PZ.<lb/></p> <p>The prostate model encloses a lattice defining a three-di-<lb/>mensional spatial coordinate system. The fineness of this lat-<lb/>tice grid represents another free parameter. Each coordinate<lb/> defines a potential tumor location. Therefore, the density of<lb/> the lattice affects the number of tumors that may be situated<lb/> within the PZ. To simplify our approach, we assumed tumor<lb/> distribution to be unifocal. Consequently, only one tumor is<lb/> present at any time within the gland. Furthermore, tumors are<lb/> not allowed to touch or protrude through the prostatic capsule<lb/> defining the external PZ boundary. Although restricted to the<lb/> PZ, lesions may protrude anteriorly. The origin of each spher-<lb/>ical tumor must, however, be confined to the PZ. Tumor ra-<lb/>dius represents another free parameter with a minimum of 1<lb/> mm. The maximum value depends on the size of the PZ where<lb/> the tumor is located.<lb/></p> <p>The biopsy module allows user definition of the number of<lb/> cores, their insertion sites, and angulation. The length of each<lb/> core was fixed at 15 mm, and its cross section was negligible<lb/> because the cores are defined by axial equations. Herein, we<lb/> tested biopsy methods relying on 2-, 4-, 6-, 8-, 10-, and 12-<lb/>sector biopsies. The insertion site for each core is defined for<lb/> the right half of the gland using latero-lateral (LL) and cranio-<lb/>caudal (CC) axis coordinates. The software then creates a<lb/> mirror image for the contralateral lobe. The AP coordinate is<lb/> automatically defined by the software on interaction with the<lb/> prostate module and depends on the shape of the previously<lb/> defined gland.<lb/></p> <p>AP and LL biopsy core angulations represent two additional<lb/> free parameters offered in the biopsy module. Values can<lb/> range from 0 to 90°. AP angulation of 0° results in the needle<lb/> pointing cranially along the vertical plane. A 30° AP angula-<lb/>tion, used in this study, approximates the usual needle angu-<lb/>lation achieved clinically with a side-firing probe. Similarly,<lb/> LL angulation of 0°, used here, results in needles pointing<lb/> anteriorly in the horizontal plane, whereas a 30° LL angulation<lb/> would have resulted in a 30° lateral deviation of the needle<lb/> trajectory.<lb/></p> <p>The prostate module defines the size and all possible tumor<lb/> locations for a given setup. It systematically modifies tumor<lb/> location from one three-dimensional coordinate of the lattice<lb/> to another, until all the possible locations have been ex-<lb/>hausted. Each new tumor location results in an interaction<lb/> with the biopsy module and a determination of whether the<lb/> tumor was or was not detected. A detection results from inter-<lb/>section of the tumor lesion by one or more of the biopsy cores<lb/> for a single combination of gland, tumor, and needle parame-<lb/>ters. The model calculates the detection rate defined by the<lb/> ratio of tumor placements that resulted in detection divided by<lb/> all possible tumor placements.<lb/></p> <p>Our analysis was twofold. First, we assessed the biopsy yield<lb/> for each combination of gland size and biopsy approach, by<lb/> using from 4 to 12 biopsy cores in prostates from 20 to 100 cc<lb/> in size. For each of the combinations, the needles were distrib-<lb/>uted in a fashion, in our opinion, representing the usual sys-<lb/>tematic biopsy core distribution. The positioning was defined<lb/> in transverse and sagittal projections on the basis of the num-<lb/>ber of sectors submitted to biopsy. If 4 needles were used, then<lb/> two similarly sized sectors were defined on each side of a plane<lb/> dividing the left from the right lobe. If 10 needles were used,<lb/> then in the same fashion each of the lateral lobes was divided<lb/> into three medially located sectors and two laterally situated<lb/> sectors. For the purpose of 12 biopsies, three medially and<lb/> three laterally situated sectors were defined for each left and<lb/> right lobe. Within each individual sector, the needle was po-<lb/>sitioned in a fashion allowing most central sampling, as is<lb/> usually done at the time of systematic prostatic biopsy.<lb/></p> <p>In the second part of our analysis, we assessed the detection<lb/> rate according to the density of sampling. Here, for each lesion<lb/> size, detection rates for all five gland size variants were used.<lb/> Thus obtained percentages demonstrate average detection<lb/> rates achieved with a given density of sampling, expressed in<lb/> cubic centimeters of total prostatic tissue for each biopsy core,<lb/> irrespective of gland size. We decided to define the sampling<lb/> density intervals as shown in <ref type="table">Table I</ref>.<lb/></p> <head>RESULTS<lb/></head> <p>We simulated five prostates: 20, 40, 60, 80, and<lb/> 100 cc in size. For each of the glands, tumor, and<lb/> biopsy variants, the computer performed between<lb/> 5868 and 33,799 tumor placements, each resulting<lb/> in one biopsy simulation. Each of the biopsy ap-<lb/>proaches was tested on all five different gland sizes<lb/> and this for all four tumor volumes. Overall, we<lb/> simulated more than 2 million patient biopsies.<lb/></p> <p>Our data demonstrate that positive biopsy rate<lb/> increases proportionately to increasing number of<lb/> cores used to sample the gland <ref type="table">(Table II)</ref>. For all<lb/> gland volumes and tumor sizes, a substantial in-<lb/>crease in positive biopsy rate can be observed with<lb/> increasing number of biopsy cores. Although one<lb/> could suspect near 100% detection of a relatively<lb/></p> <figure type="table">TABLE I. Averaged results from all simulations according to density of sampling (amount of<lb/> tissue per biopsy core) and according to lesion size ranging from 0.3 to 1.4 cc<lb/> Amount of Prostatic Tissue Sampled<lb/> per Biopsy Core (cc)<lb/> Average Detection Rate According to Lesion Size<lb/> (% of tumors detected)<lb/> 0.3 cc<lb/> 0.5 cc<lb/> 0.9 cc<lb/> 1.4 cc<lb/> 1.5–3.5<lb/> 32.6<lb/> 42.5<lb/> 53.7<lb/> 75.9<lb/> 3.6–7.5<lb/> 16.7<lb/> 25.0<lb/> 33.8<lb/> 43.1<lb/> 7.6–12.5<lb/> 10.7<lb/> 15.8<lb/> 21.7<lb/> 28.3<lb/> 12.6–25.0<lb/> 6.2<lb/> 9.8<lb/> 14.6<lb/> 20.6<lb/></figure> <p>large 1.4-cc lesion in a 20-cc prostate using 12 bi-<lb/>opsy cores, as many as 35% of lesions may remain<lb/> undetected. For the same tumor volume over half<lb/> of the tumors remained undetected when 10 bi-<lb/>opsy cores were used in a 40-cc gland.<lb/></p> <p><ref type="table">Table II</ref> demonstrates that regardless of the bi-<lb/>opsy approach or gland volume, between 4.3% and<lb/> 42.2% of tumors of questionable clinical signifi-<lb/>cance, <ref type="biblio">6,7</ref> 0.3 cc in size, may be detected. Similar to<lb/> detection of clinically significant lesions, detection<lb/> rate of minimal volume disease increases with in-<lb/>creasing amount of sampling. On the basis of the<lb/> above, the practicing urologist should note that<lb/> increased detection of clinically insignificant dis-<lb/>ease represents the trade-off for improved detec-<lb/>tion of clinically important lesions.<lb/></p> <p>When analyzed according to density of sampling<lb/> <ref type="table">(Table I)</ref>, maximum detection rates were obtained<lb/> when one biopsy core was taken for each 1.5 to 3.5<lb/> cc of prostatic tissue. Detection rates were nearly<lb/> halved when the sampling density was decreased<lb/> to the interval extending from 3.6 to 7.5 cc of pros-<lb/>tatic tissue for each biopsy core, which approxi-<lb/>mates sampling achieved with sextant biopsy in<lb/> glands ranging from 21.6 to 45 cc. When the lowest<lb/> sampling density was analyzed (12.6 to 25.0 cc of<lb/> prostatic tissue per biopsy core), approximating<lb/> quadrant biopsy in a 50 to 100-cc gland, between<lb/> 6.2% and 20.6% of lesions were detected. Such<lb/> yield represents between 19% and 27% of the per-<lb/>formance achieved with the highest sampling den-<lb/>sity.<lb/></p> <head>COMMENT<lb/></head> <p>Our results indicate that the positive biopsy rate<lb/> depends on the degree of gland sampling. Dou-<lb/>bling the number of biopsy cores can translate into<lb/> increasing the biopsy yield by as much as 100%.<lb/> Consequently, a volume-based algorithm deter-<lb/>mining the number of biopsy cores required for<lb/> adequate detection may provide optimal clinical<lb/> results. However, increased detection of significant<lb/> volume lesions is paralleled by increased detection<lb/> of small size lesions, of questionable clinical signif-<lb/>icance. Overdetection of low-volume disease may<lb/> represent an argument against increased sampling.<lb/></p> <p>A criticism of our study may relate to several of<lb/> its aspects. First, our results are derived from a<lb/> computer simulation and not from a clinical study.<lb/> We acknowledge that a computer simulation of a<lb/> clinical situation is and likely always will be lim-<lb/></p> <figure type="table">TABLE II. Biopsy yield for spherically shaped tumor lesions according to gland size and number<lb/> of sectors subjected to biopsy<lb/> Prostatic<lb/> Size (cc)<lb/> Peripheral<lb/> Zone AP<lb/> (mm)<lb/> Number of<lb/> Simulations<lb/> Biopsy Yield (% of tumors detected)<lb/> 4<lb/> Sectors<lb/> 6<lb/> Sectors<lb/> 8<lb/> Sectors<lb/> 10<lb/> Sectors<lb/> 12<lb/> Sectors<lb/> Lesion 0.3 cc in volume (clinically insignificant)<lb/> 20<lb/> 5.3<lb/> 5,868<lb/> 19.4<lb/> 27.2<lb/> 35.0<lb/> 33.0<lb/> 42.2<lb/> 40<lb/> 7.1<lb/> 17,868<lb/> 6.2<lb/> 8.6<lb/> 15.5<lb/> 19.8<lb/> 25.4<lb/> 60<lb/> 8.4<lb/> 19,632<lb/> 8.3<lb/> 12.6<lb/> 13.5<lb/> 15.0<lb/> 21.3<lb/> 80<lb/> 8.6<lb/> 26,566<lb/> 5.0<lb/> 8.6<lb/> 10.7<lb/> 12.1<lb/> 21.0<lb/> 100<lb/> 10.2<lb/> 33,799<lb/> 4.3<lb/> 6.7<lb/> 11.5<lb/> 10.3<lb/> 13.7<lb/> Lesion 0.5 cc in volume (borderline clinically significant)<lb/> 20<lb/> 5.3<lb/> 4,915<lb/> 27.3<lb/> 37.2<lb/> 45.4<lb/> 41.4<lb/> 50.9<lb/> 40<lb/> 7.1<lb/> 16,119<lb/> 8.2<lb/> 11.4<lb/> 24.4<lb/> 28.4<lb/> 37.4<lb/> 60<lb/> 8.4<lb/> 17,769<lb/> 13.4<lb/> 19.9<lb/> 22.3<lb/> 24.5<lb/> 32.1<lb/> 80<lb/> 8.6<lb/> 24,332<lb/> 8.6<lb/> 13.3<lb/> 17.8<lb/> 15.2<lb/> 30.0<lb/> 100<lb/> 10.2<lb/> 31,216<lb/> 6.9<lb/> 10.4<lb/> 15.2<lb/> 16.1<lb/> 20.9<lb/> Lesion 0.9 cc in volume (clinically significant)<lb/> 20<lb/> 5.3<lb/> 4,280<lb/> 34.3<lb/> 45.5<lb/> 54.4<lb/> 47.9<lb/> 57.3<lb/> 40<lb/> 7.1<lb/> 14,246<lb/> 10.6<lb/> 14.3<lb/> 35.2<lb/> 38.3<lb/> 49.7<lb/> 60<lb/> 8.4<lb/> 15,792<lb/> 20.2<lb/> 28.8<lb/> 32.1<lb/> 35.8<lb/> 44.3<lb/> 80<lb/> 8.6<lb/> 21,950<lb/> 13.3<lb/> 19.7<lb/> 26.2<lb/> 18.1<lb/> 36.4<lb/> 100<lb/> 10.2<lb/> 28,454<lb/> 10.4<lb/> 15.3<lb/> 18.6<lb/> 22.9<lb/> 28.3<lb/> Lesion 1.4 cc in volume (clinically significant)<lb/> 20<lb/> 5.3<lb/> 3,474<lb/> 43.7<lb/> 56.9<lb/> 63.9<lb/> 56.0<lb/> 64.9<lb/> 40<lb/> 7.1<lb/> 12,251<lb/> 13.4<lb/> 18.3<lb/> 46.7<lb/> 48.8<lb/> 62.0<lb/> 60<lb/> 8.4<lb/> 13,960<lb/> 28.9<lb/> 38.5<lb/> 42.2<lb/> 46.3<lb/> 56.5<lb/> 80<lb/> 8.6<lb/> 19,598<lb/> 19.6<lb/> 27.2<lb/> 35.2<lb/> 21.9<lb/> 41.9<lb/> 100<lb/> 10.2<lb/> 25,667<lb/> 14.6<lb/> 21.0<lb/> 22.4<lb/> 30.1<lb/> 35.8<lb/> KEY: AP anteroposterior diameter.<lb/></figure> <p>ited by its inability to reproduce all possible vari-<lb/>ables and interactions. However, we would like to<lb/> emphasize the importance of laboratory modeling,<lb/> because it provides valuable information otherwise<lb/> difficult to obtain within a reasonable time. Fur-<lb/>thermore, because of its relative simplicity com-<lb/>pared with clinical experimentation, modeling al-<lb/>lows focusing on a single factor while controlling<lb/> other variables. On an example basis, we con-<lb/>trolled all variables such as the size of the gland,<lb/> the size and shape of the peripheral zone, the num-<lb/>ber, distribution and angulation of biopsy needles<lb/> and focused entirely on the effect of tumor size on<lb/> the biopsy yield. Furthermore, only under experi-<lb/>mental conditions can the true yield of a given bi-<lb/>opsy method be appreciated. In real life the num-<lb/>ber of true negatives is unknown, and reported<lb/> positive biopsy rates may represent an overesti-<lb/>mate of the real, yet unknown, detection rate.<lb/></p> <p>Finally, an important critique may relate to tu-<lb/>mor distribution within the PZ. Because the model<lb/> development is ongoing and this study depicts its<lb/> initial phase, we used a grossly simplified, system-<lb/>atic tumor distribution. Furthermore, unlike in re-<lb/>ality, all tumors were of spherical shape. Again, we<lb/> realize that the majority of cancerous prostates<lb/> harbor multicentric lesions of various shapes. In<lb/> addition to systematic distribution of tumors, the<lb/> physical limitations of the design contributed to<lb/> findings unlikely representative of reality.<lb/></p> <p>Close examination of the number of simulations<lb/> performed for each gland variant reveals one of the<lb/> limitations of our model: the number of tumors<lb/> originating from within the PZ is not necessarily in<lb/> proportion to the increase in the zone and gland<lb/> size. A 100% increase in size as seen between 20<lb/> and 40 cc yielded approximately a threefold in-<lb/>crease in the number of tumors <ref type="table" >(Table II)</ref>. How-<lb/>ever, a 50% gland size increase, such as from 40 to<lb/> 60 cc, yielded at most a 10% increase. The subse-<lb/>quent increases in the number of tumors were well<lb/> above 10% and were more in proportion to the<lb/> increase in gland size (from 60 to 80 cc and from 80<lb/> to 100 cc). This apparent limitation is related to the<lb/> rigid conditions of our design: all tumors must re-<lb/>main nonpalpable and confined within the pros-<lb/>tatic capsule. For selected tumor volume and gland<lb/> volume combinations, the model is unable to fit as<lb/> many tumors as would intuitively be expected,<lb/> given the constant increment in the AP dimension<lb/> of the PZ compared with the AP dimension of the<lb/> gland.<lb/></p> <p>Another limitation due to a model design prob-<lb/>lem is evidenced by the fact that in select condi-<lb/>tions, the tumor detection rate decreased despite<lb/> an increase in the number of cores used for biopsy<lb/> and despite unchanged tumor and gland parame-<lb/>ters. Such a paradoxical finding can be seen in <ref type="table" >Ta-<lb/>ble II</ref>, where in a 20-cc gland a 10-sector biopsy<lb/> yields 33% detection, compared with 35% and 42%<lb/> for 8 and 12 cores, respectively. Here, in our opin-<lb/>ion, the unexpectedly low yield achieved with 10<lb/> cores compared with 8 or 12 cores is due to one or<lb/> more excessively laterally positioned needles. Such<lb/> lateral positioning results in the core transecting a<lb/> thinner PZ than if placed more medially. Because<lb/> fewer tumors can be positioned within a thinner,<lb/> lateral PZ, the detection rate decreases.<lb/></p> <p>We acknowledge that, in reality, the number of<lb/> detectable tumors does not decrease on the most<lb/> lateral aspect of the PZ. As such for it to be closer to<lb/> real life, our model should demonstrate a steady<lb/> detection rate when more laterally positioned nee-<lb/>dles are used. We also acknowledge that the num-<lb/>ber of possible tumor locations should increase in a<lb/> more proportionate manner to the increase in<lb/> gland volume than that demonstrated in our<lb/> model. However, we do not believe that these rep-<lb/>resent major limitations, because our goal is not to<lb/> advocate adoption of our biopsy algorithms but to<lb/> emphasize the need and stimulate discussion re-<lb/>garding possible improvements in currently used<lb/> biopsy strategies. We emphasize that the impor-<lb/>tance of our findings resides in demonstrating that<lb/> detection rate is likely to significantly increase if<lb/> more than six biopsies are performed in bulky<lb/> glands. Although we recognize the experimental<lb/> nature of our results, we currently obtain one bi-<lb/>opsy core for at least every 5 cc of prostatic tissue.<lb/> This translates into a biopsy algorithm, as shown in<lb/> <ref type="table">Table III</ref>.<lb/></p> <head>CONCLUSIONS<lb/></head> <p>Herein, we introduced a novel tool for assess-<lb/>ment of sector prostatic biopsy. We believe that<lb/> our data, although bound by inherent limitations<lb/> of computer modeling, confirm previous clinical<lb/> suspicion that the extent of sampling determines<lb/> its yield. On the basis of our results, we believe that<lb/></p> <figure type="table">TABLE III. Recommended number of biopsy cores, according to gland size<lb/> Gland size (cc)<lb/> 0–30<lb/> 30.1–40<lb/> 40.1–50<lb/> 50.1–60<lb/> 60.1–70<lb/> 70.1–80<lb/> >80.1<lb/> Recommended number of<lb/> biopsy cores<lb/> 6<lb/> 8<lb/> 10<lb/> 12<lb/> 14<lb/> 16<lb/> 18<lb/></figure> <p>the number of biopsy cores required for adequate<lb/> detection should not be determined in a uniform<lb/> fashion but should follow a gland volume-based<lb/> algorithm. Although our results are preliminary,<lb/> they warrant further analysis aimed at defining the<lb/> optimal approach to sector biopsy of the palpably<lb/> and ultrasonically benign gland.</p> </text> </tei>