| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
and
Andrea Klauser, MD*
Department of Radiology II, University Hospital Innsbruck, Anichstrasse 35, Innsbruck 6020, Austria*, e-mail: ferdinand.frauscher@uibk.ac.at
Department of Radiology, Thomas Jefferson University, Philadelphia, Pa
Editor:
We read with great interest the article by Dr Kuligowska and colleagues in the September 2001 issue of Radiology (1) about the accuracy of gray-scale and color Doppler ultrasonography (US) as predictors of prostate carcinoma. We would like to offer several comments on their contribution.
First, Dr Kuligowska and colleagues do not describe the location of targeted biopsies. This is a very important issue when color Doppler US-guided targeted biopsies for prostate cancer detection are evaluated. Changes of benign prostatic hyperplasia in the transition zone often demonstrate hypervascularity, which cannot be differentiated from the hypervascularity of malignant tissue. For this reason, many experts limit the targeted technique to peripheral zone specimens.
Second, systematic sextant and targeted biopsies in each subject were performed by the same radiologist. Since the targeted biopsy was performed before the sextant biopsy, a bias may be introduced into the positioning of the sextant biopsy specimens, which could artificially inflate the apparent accuracy of the sextant biopsy. In contrast to the present study, previous studies in which gray-scale and color Doppler US were compared for the detection of prostate cancer reported a higher positive predictive value for color Doppler US (2,3). It would be important to know the distribution of serum prostate-specific antigen (PSA) levels in the study population, since subjects with higher PSA values may be more likely to have larger cancers with detectable gray-scale abnormalities, as demonstrated in Figures 4 and 5. Doppler examination may be more useful in a screening population with only mildly elevated PSA levels.
Third, Dr Kuligowska and colleagues describe a sextant biopsy technique, demonstrated in Figure 2 that included "sampling of both the peripheral and central zones." In the text they state that "on average, two core biopsy samples of each of the six sextants" were obtained. The biopsy technique used in this study is unclear, and the terminology is imprecise. The central zone refers to a specific glandular zone that exists only above the level of the verumontanum (4,5) and cannot be distinguished from peripheral zone with conventional US. The classical sextant approach, as first described by Hodge et al (6), consists of parasagittal cores that pass through both peripheral and central zones in a parasagittal plane. The modified sextant, as described by Stamey (7), consists of laterally directed cores that pass almost entirely through the peripheral zone. The terminology "outer" and "inner" gland is preferred for the description of sonographic localization of cores, since "central zone" is easily confused with "central gland." Did the authors perform both a classical sextant and a modified sextant? Why were two cores obtained "on average" rather than in all cases? The authors should clarify the number of cores obtained from both the inner gland (transition zone) and their criteria for obtaining such specimens (ie, elevated PSA level >10 ng/mL and/or increased transition zone volume). Furthermore, it appears that cores from the outer and inner glands were placed into a single container; this is unfortunate since it precludes a separate analysis of outer and inner gland biopsy results.
Fourth, Dr Kuligowska and colleagues state that they used color Doppler US in their study but make no mention of power Doppler US. Figure 4b, however, looks like a power Doppler image. Power Doppler US is more sensitive to slow flow and is less angle-dependent than color Doppler US. One recent study, in which power Doppler US was used, demonstrated increased sensitivity in depicting blood flow, number, course, and continuity of vessels (8). Furthermore, it has been shown that foci of increased power Doppler flow may be 4.7 times more likely to contain cancer than adjacent tissues without flow (9). Obviously it is critical to know if this study used both color and power Doppler US. Although the present study used a US system with high-quality imaging, it should be noted that new high-frequency probes, which operate at a Doppler frequency of 9 MHz (most systems operate at a frequency of 6 MHz) may improve the results of color Doppler US for the detection of prostate cancer.
Finally, the application of multiple logistic regression to examine the independent effect of each screening method is problematic. Presumably, the unit of analysis for regression was the patient, since each patient is an independent observation. The outcome variable was the presence of prostate cancer. Predictor variables included gray-scale US, color Doppler US, and excess PSA. Findings from each patient were classified as normal or abnormal at gray-scale and color Doppler US, and these data were entered into the regression model. This regression model will not be used to consider whether the site of gray-scale or color Doppler abnormality corresponds to the site of cancer at biopsy, but only whether there is a sonographic abnormality at any location in the prostate, and whether the same patient has cancer. What if the sonographic abnormality did not correspond to the site of cancer? Furthermore, this type of analysis assumes equal weight for each observation (patient), even though the number of biopsies per patient were not equal. To address these issues, the analysis should include the information from individual biopsy sites. A conditional logistic analysis or generalized estimating equations approach would be more appropriate.
In conclusion, we agree with the major conclusions of Dr Kuligowska and colleagues. Targeted biopsy based on gray-scale US, even coupled with color Doppler US, is not an adequate screening method for prostate cancer. However, interested readers should be provided with additional details regarding the biopsy technique and the patient population to evaluate these study findings.
REFERENCES
Department of Radiology, Boston University School of Medicine, 88 East Newton Street, Boston, MA 02118. e-mail: ewa.kuligowska@bmc.org
We wish to thank Drs Frauscher and Halpern for their comments regarding our recently published article (1), and we appreciate the opportunity to discuss and clarify several points of interest for readers.
First, in all cases, the targeted biopsies were limited to the peripheral zone. As authors of the letter point out, abnormalities in the transitional zone noted on gray-scale and/or color Doppler images cannot be used to differentiate between benign hyperplasia and malignancy. Therefore, we did not perform biopsy in lesions in the transitional zone. Lesions in the peripheral zone depicted either with gray-scale or color Doppler US were considered to be suspicious for malignancy, and two biopsies of each visible lesion were performed.
Second, although it is possible that bias could have been introduced by performing the targeted biopsy prior to the sextant biopsy, we do not believe this occurred. Care was taken to perform biopsy in each sextant without regard for any previous biopsy, with the caveat that obvious biopsy tracts were not reentered. Therefore, it is more likely that the bias would result in a lower accuracy for the sextant technique, since prior targeted biopsy sites were not directly reentered during the sextant biopsy. While Doppler examination may be helpful in the screening of patients with mildly elevated PSA levels, it is essential for identifying malignancies that are isodense, without any gray-scale abnormalities, and highly invasive with a high PSA level. Our results, which are similar to the results reported in other studies (2), show that color Doppler US improves the detection of prostate cancer by an additional 10%–20%. Our data also support the association of increased vascularity with a high Gleason score (P < .05).
Third, Drs Frauscher and Halpern correctly identify an error in terminology in the legend of Figure 2. Referral to sampling of the peripheral and central zones is imprecise. Correct terminology is sampling of the peripheral zone and central gland. The central gland refers to the transitional zone and central zone located in the periurethral region above the verumontanum. We agree that it would be clearer to the readers if we had described the sonographic location of cores, as "outer" (peripheral zone) and "inner" (central gland).
The biopsies were performed in a laterally directed approach to include primarily the peripheral zone (outer gland). The number of cores was limited to 10–12. When the volume of the transitional zone (inner gland) was massively enlarged, we obtained additional cores on each side to be certain that the stretched, thinned peripheral zone was completely sampled. A possible limitation of our study was to not separate the cores from the inner and outer gland into different containers. All biopsy samples were labeled according to location and divided into containers labeled for the left and right lobes of the prostate. All targeted biopsy smaples were placed in separate containers.
Fourth, in our study we used both color and power Doppler US as adjuncts to gray-scale US. We used a high-frequency 5–9-MHz end-fire transducer (Advanced Technology Laboratories, Bothell, Wash), as described in the article. Power Doppler US has a well-documented advantage over color Doppler US and was used to improve detection of vascularity.
Finally, the response of the Radiology statistical consultant follows regarding our statistical analysis.
REFERENCES
The authors of the letter seem to suggest that biopsies be used as the unit of analysis. This is a suggestion for a wholly different study focused on the biopsy specimen. The authors do correctly suggest that if multiple biopsies per patient had been the unit of observation, it would have been necessary to use more complex methods for analysis. However, even in studies where the more complex methods may be applied, it will still be necessary to return the reader’s interest back to the clinical application. Even in a biopsy-oriented study, it will be necessary to summarize the information as in table 2 and to summarize the results as Dr Kuligowska and colleagues have done. Thus, there is nothing problematic about the application of logistic regression to the patient-based information collected in this study.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
