David Quinn is the recipient of an NHMRC Neil Hamilton Fairley Fellowship and the Vincent Fairfax Family Foundation Fellowship from the Royal Australasian College of Physicians.
Prognostic significance of preoperative factors in localized prostate carcinoma treated with radical prostatectomy
Importance of percentage of biopsies that contain tumor and the presence of biopsy perineural invasion
Article first published online: 1 APR 2003
Copyright © 2003 American Cancer Society
Volume 97, Issue 8, pages 1884–1893, 15 April 2003
How to Cite
Quinn, D. I., Henshall, S. M., Brenner, P. C., Kooner, R., Golovsky, D., O'Neill, G. F., Turner, J. J., Delprado, W., Grygiel, J. J., Sutherland, R. L. and Stricker, P. D. (2003), Prognostic significance of preoperative factors in localized prostate carcinoma treated with radical prostatectomy. Cancer, 97: 1884–1893. doi: 10.1002/cncr.11263
- Issue published online: 1 APR 2003
- Article first published online: 1 APR 2003
- Manuscript Accepted: 16 DEC 2002
- Manuscript Revised: 4 DEC 2002
- Manuscript Received: 7 OCT 2002
- National Health and Medical Research Council of Australia (NHMRC)
- The Cancer Council New South Wales
- R. T. Hall Trust
- Freedman Foundation
- Leo & Jenny Leukaemia and Cancer Foundation of Australia
- St. Vincent's Clinic Foundation
- prostate carcinoma (PC);
- radical prostatectomy (RP);
- prostate biopsy;
- perineural invasion (PNI)
Predicting outcome for men with clinically localized prostate carcinoma treated with curative intent remains imprecise and further evaluation of accepted and potential predictive factors is needed.
The authors studied 696 men with localized prostate carcinoma diagnosed on transrectal biopsy and treated with radical prostatectomy at one institution between 1986 and 1999 to determine the relation between putative pretreatment prognostic factors and disease-free survival. Clinical stage, Gleason score, perineural invasion, number of biopsies containing tumor, and serum prostate specific antigen (PSA) were evaluated as predictors of extracapsular extension, seminal vesicle involvement, lymph node metastases, and surgical margin involvement as well as outcome after surgery. Kaplan–Meier method and Cox regression analyses were used to evaluate the contribution of different factors to adverse pathologic features and relapse.
At mean follow-up of 56.9 months (range, 1.0–177.9 months; median, 54.9 months), 26.1% (182 of 696 patients) of patients had developed a disease recurrence. Pretreatment serum PSA concentration, biopsy Gleason score, and clinical stage as well as number of biopsies involved with tumor as a percentage of the total number obtained were found to be independent predictors of outcome. In patients with PSA > 10 ng/mL, biopsy perineural invasion and percentage of biopsies containing tumor were found to independently predicted disease recurrent. Increased number of biopsies involved with tumor independently predicted extracapsular extension, margin involvement, seminal vesicle, and lymph node involvement.
This study demonstrated that the proportion of prostate biopsy cores containing tumor is an independent predictor of outcome after subsequent radical prostatectomy and suggested that perineural invasion has a predictive role in patients with a preoperative PSA > 10 ng/ml. Cancer 2003;97:1884–93. © 2003 American Cancer Society.
The preoperative prediction of outcome for patients with localized prostate carcinoma (PC) is impaired by undergrading and understaging of patients compared with operative findings. The precision in predicting outcome after therapy in patients with clinically localized PC may be greatly improved if further preoperative indicators of outcome can add information to that obtained routinely from clinical stage, biopsy Gleason score, and pretreatment serum prostate specific antigen (PSA) concentration.
Several North American groups have published data on cancer control after radical prostatectomy (RP). These cohorts now form the basis of recent studies on predicting outcome after surgery based on pretreatment factors. 1–5 In addition, a number of groups have published data on the preoperative prediction of adverse pathologic features at RP, including extracapsular extension, seminal vesicle involvement (SVI), lymph node (LN) metastases, and surgical margin involvement. 6–11 However, there are few studies that present predictive data for each of these adverse pathologic features concurrently with subsequent cancer progression in the same cohort.
We evaluated a cohort of 696 patients with localized PC diagnosed by transrectal prostate biopsy and subsequently undergoing surgery with intent of radical prostatectomy. The cohort received no neoadjuvant therapy. The aims were to determine the effect of preoperative factors including serum PSA, clinical stage, biopsy Gleason score, number and proportion of biopsy cores containing tumor as a fraction of all biopsy cores obtained, and the presence of perineural invasion (PNI) in the prostate biopsies in predicting adverse pathologic factors in the surgical specimen and disease-free survival.
MATERIALS AND METHODS
The study proceeded with the approval of the St. Vincent's Campus Research Ethics Committee. Between December 1986 and December 1999, 835 men with clinically localized carcinoma of the prostate underwent surgery for planned radical retropubic prostatectomy on the St. Vincent's Hospital Campus. Routine preoperative workup in addition to serum PSA, digital rectal examination (DRE) (as per 1992 tumor node metastasis [TNM] clinical stage regardless of year of surgery 12), and histologic diagnosis included pelvic computed tomography, transrectal prostatic ultrasound, and technetium bone scan. From February 1996 onward, patients with serum PSA concentrations < 10 ng/mL were not routinely subjected to bone scan. Of the original cohort, 139 patients were excluded from this analysis on the following bases: 50 men diagnosed with PC on the basis of prostate tissue obtained by transurethral resection, 83 men treated with neoadjuvant hormonal therapy; 4 with a histologic diagnosis other than acinar adenocarcinoma (3 cases of prostatic ductal predominant carcinoma and 1 case of sarcoma), and 2 men treated with perineal RP. Eight men had their surgery abandoned due to pelvic lymph node involvement detected on frozen section at surgery, and these patients were included in the study and considered as having recurred at time zero. The cohort for analysis was 696 men with acinar adenocarcinoma of the prostate primarily diagnosed by prostate biopsy and selected for intention to undergo retropubic RP with no preoperative treatment and a final censor date for follow-up of August 31, 2001. Surgery was undertaken by one of six urologists at the St. Vincent's Campus using similar RP technique. Recurrence-free survival was not significantly different between surgeons, when known prognostic factors were taken into account. Multivariate analysis examining the effect of one surgeon versus all others yielded hazard ratios close to 1 in most cases and P values ranging from 0.27–0.97 for individual surgeons. Greater than 97% of the patients were white. Patients given adjuvant radiation therapy (n = 43) were treated with a 4-field technique delivering 49.6–66.9 grays (Gy) in 25–30 fractions commencing 2.5–5.6 months after surgery. Exclusion of these patients from outcome analysis did not appear to alter any of the major conclusions of this study. Outcome after RP related to surgical pathologic features and preoperative serum PSA in our cohort has been reported recently. 13 That report included data on the patients described here but excluded patients where surgery was abandoned due to positive pelvic lymph nodes and included patients in whom the diagnosis of PC was made based on transurethral resection specimens.
Tissue Handling and Pathology Reporting
Each pathologist adhered to a similar processing and reporting pattern for prostate biopsy specimens. Prostate biopsy reports routinely included numbers of cores involved with tumor, Gleason score, 14 and the presence of PNI where evident. The proportion or percentage of biopsies containing tumor was calculated using the number of individual biopsies containing tumor divided by the total number of biopsies obtained. Data on the proportional length of individual biopsy cores involved with tumor were not available. PNI was defined as focal if one area was identified and as multifocal if two or more areas could be identified in the set of biopsy specimens. Each pathologist adhered to a similar processing and reporting pattern for RP specimens as recently described. 13 A surgical margin was considered positive if PC extended to an inked surface, including those at the prostatic apex.
Follow-Up and Data Acquisition
Clinical follow-up was undertaken every 3 months for 2 years and yearly thereafter. Serum PSA estimations using Hybritech equivalent assays were obtained every 3 months for 2 years and then every 6 months to 5 years and annually thereafter. The database was matched against the New South Wales State Cancer Registry to which all cancer diagnoses are notifiable by law. The Registry also collects information concerning death and cancer-specific death in cancer patients through a national clearinghouse for such information coordinated by the Federal Government of Australia. Data were collected prospectively from 1990, and all clinical notes were audited retrospectively by one author (D.I.Q.) to verify data.
In addition to the presence of extracapsular extension, surgical margin involvement, SVI, and LN metastases in the surgical specimen, the major outcome measure for this cohort study was failure of surgery to provide disease control. This was defined as a rise in serum PSA > 0.4 ng/mL with subsequent further rises or clinical local recurrence detected on digital rectal examination (DRE), confirmed by either a positive biopsy or subsequent rise in PSA. In addition, the institution of hormonal therapy intended to be indefinite or bilateral orchidectomy in the postoperative period was considered to represent conservative failure of surgery to control disease and, therefore, as disease recurrence (n = 24). The recurrence date was defined as the date of the first serum PSA reading ≥ 0.4 ng/mL, the first abnormal DRE, or the commencement of hormonal therapy in the respective groups. This definition of relapse is consistent with that used in reports on other large cohorts. 1, 5, 15–17
Data were evaluated for recurrence prediction using the Kaplan–Meier product limit method and log-rank test and by univariate and multivariate analysis in a Cox proportional hazards model. 18, 19 A limited subset analysis was undertaken dividing the cohort at inflection points for outcome described in other cohorts for clinical stage, preoperative PSA, and biopsy Gleason score. Contribution to adverse pathologic features was evaluated using a regression model. Correlations were evaluated using the chi-square test, Mann–Whitney U test, Kruskal–Wallis test, and the Student t test for paired data as appropriate. All statistical analyses were performed using Statview 4.5 software (Abacus Systems, Berkeley, CA). Statistical significance in this study was set at P < 0.05. All reported P values are two-sided.
Clinical and pathologic characteristics for the cohort are presented in Table 1. At a mean follow-up of 56.9 months (range, 1.0–177.9 months; median, 54.9 months), 26.1% (182 of 696 patients) of patients had developed disease recurrence.
|≤ 65 yrs||441||63.4||62.0||62.5|
|> 65 yrs||255||36.6||(40.7–76.7)|
|Biopsy Gleason score (n = 693)|
|No. of biopsies taken (n = 690)|
|No. of biopsies positive (n = 690)|
|Biopsy perineural invasion (n = 690)|
|Preoperative PSA (n = 688)|
|≥ 4.0–< 10||355||51.6||12.9||9.1|
|≥ 10–< 20||219||31.8||(1.0–194)|
|Extracapsular extension (n = 689)||307||44.6|
|Seminal vesicle involvement (n = 689)||94||13.6|
|LN metastases (n = 696)||24||3.4|
|Surgical margin involvement (n = 689)|
The mean Gleason score assigned to the biopsies was 5.96 (range, 3–10; median, 6) compared with that in the RP specimen of 6.39 (range, 3–10; median, 6). These data define a trend to underscoring of Gleason sum on biopsies compared with the final Gleason score assigned after surgery, which although statistically not significantly different may have clinical importance (P < 0.0001 for correlation).
An increasing percentage of the total number of biopsies positive for tumor correlated with clinical stage (P < 0.0001), pretreatment PSA (P = 0.01), Gleason score in either biopsy or RP (P < 0.0001), PNI in biopsies (P < 0.0001) and in the RP specimen (P = 0.0002), extracapsular extension (P = 0.0005), overall pathologic stage (P < 0.0001), positive surgical margins (P < 0.0001), and SVI (P < 0.0001) but not with LN metastases (P = 0.09).
PNI was reported in 10.4% of prostate biopsies and 58.3% of RP specimens. Although biopsy PNI markedly underpredicted RP PNI, there was a strong correlation between the two (P < 0.0001). Among preoperative factors, biopsy PNI was found to correlate with clinical stage (P = 0.03), biopsy Gleason score (P < 0.0001), and number of positive biopsies (P = 0.002) but not pretreatment PSA (P = 0.9). Biopsy PNI correlated with the extracapsular extensionECE (P < 0.0001), SVI (P = 0.02), positive surgical margins (P = 0.02), and RP Gleason score (P = 0.002) but not with LN metastases (P = 0.29).
Kaplan–Meier and Cox analyses for each of preoperative serum PSA, biopsy Gleason score, clinical stage, percentage of biopsies involved with PC, and the presence of PNI in the biopsy specimen showed that each of these factors predicted outcome if considered alone (Figs. 1, 2; Table 2). Fewer total number of biopsies taken at each diagnostic biopsy episode correlated with a poorer outcome (Table 2, P = 0.013). In looking for inflection points for outcome based on the percentage of biopsies containing tumor, there was a significant change in prognosis when > 50% of biopsies taken at 1 diagnostic procedure were positive for cancer (Table 2, P = 0.0003) (Fig. 2B).
|Variable||Subgroup||RR (95 % CIs)a||P value|
|Clinical stage (n = 696)||cT1||1.0|
|≥ cT2||1.97 (1.38–2.81)||0.002|
|Other||2.18 (1.50–3.18)||< 0.0001|
|cT3||6.62 (4.22–10.38)||< 0.0001|
|Preoperative serum PSA (n = 688)||Continuous||1.019 (1.015–1.023)||< 0.0001|
|≤ 10 ng/ml||1.0|
|> 10 ng/ml||2.37 (1.75–3.22)||< 0.0001|
|≤ 20 ng/ml||1.0|
|> 20 ng/ml||3.86 (2.78–5.38)||< 0.0001|
|Biopsy Gleason score (n = 693)||Continuous||1.354 (1.201–1.526)||< 0.0001|
|≥ 8||3.58 (2.45–5.24)||< 0.0001|
|Percentage of biopsies positive for cancer (n = 690)||Continuous||1.013 (1.007–1.018)||< 0.0001|
|> 50%||1.74 (1.29–2.34)||0.0003|
|Total no. of biopsies taken (n = 690)||Continuous||0.888 (0.809–0.975)||0.013|
|No. of biopsies positive for cancer (n = 690)||≤ 4||1.0|
|≥ 6||2.70 (1.59–4.60)||0.0002|
|Biopsy perineural invasion (n = 690)||No||1.0|
Multivariate analysis of pretreatment factors in our cohort demonstrated that, in addition to preoperative serum PSA, biopsy Gleason score. and clinical stage, the percentage of biopsies involved with PC is an independent predictor of disease recurrence after prostatectomy, whereas the presence of perineural invasion is not (Table 3A). Multivariate analysis demonstrated that neither the number of biopsies containing cancer as a percentage of the total nor the presence of perineural invasion in the biopsies were predictive of outcome in cases with preoperative PSA ≤ 10 ng/mL. (Table 3, Bi). However, in those patients with a PSA > 10 ng/mL, both of these factors were found to independently predict outcome (Table 3, Bii). Further subset analyses did not define significant difference between subgroups defined by Gleason score or clinical stage in our cohort data (data not shown).
|Variable||Subgroup||RR (95 % CIs)a||P value|
|Clinical stage||≥ cT2 vs. cT1||1.46 (1.01–2.12)||0.046|
|Preoperative serum PSA||Continuous||1.016 (1.012–1.020)||< 0.0001|
|Biopsy Gleason score||Continuous||1.284 (1.129–1.461)||0.0001|
|Percentage of biopsies positive||Continuous||1.007 (1.001–1.013)||0.015|
|Biopsy perineural invasion||Yes vs. no||1.40 (0.90–2.16)||0.13|
|Preoperative PSA ≤ 10 ng/ml|
|Clinical stage||≥ cT2 vs. cT1||2.25 (1.04–4.84)||0.04|
|Biopsy Gleason score||Continuous||1.470 (1.171–1.845)||0.0009|
|Percentage of biopsies positive||Continuous||1.002 (0.992–1.013)||0.65|
|Biopsy perineural invasion||Yes vs. no||1.487 (0.571–3.875)||0.42|
|Preoperative PSA > 10 ng/ml|
|Clinical stage||≥ cT2 vs. cT1||1.22 (0.72–2.07)||0.45|
|Biopsy Gleason score||Continuous||1.215 (1.016–1.453)||0.03|
|Percentage of biopsies positive||Continuous||1.015 (1.007–1.023)||0.0002|
|Biopsy perineural invasion||Yes vs. no||2.120 (1.258–3.573)||0.005|
The percentage of number of biopsies positive for cancer also independently predicted a number of adverse pathologic features in the RP specimen including extracapsular extension, surgical margin involvement, SVI, and LN metastases (Table 4). The presence of PNI in the specimen was not independently predictive of these features (Table 4). The only preoperative factor to maintain independent predictive significance when included with pathologic variables from the RP specimen was preoperative serum PSA (data not shown, see Quinn et al. 13).
|Pathologic feature||Predictive variable||RR (95% CIs)b||P value|
|Extra capsular extension (n = 674)c|
|Biopsy Gleason score: (continuous)d||1.170 (1.064–1.287)||0.001|
|Percentage of biopsies positive: (continuous)d||1.007 (1.002–1.011)||0.002|
|Pretreatment serum PSA: (continuous)d||1.007 (1.002–1.011)||0.01|
|Positive surgical margins (n = 672)c|
|Biopsy Gleason score: (continuous)d||1.119 (1.021–1.227)||0.016|
|Percentage of biopsies positive: (continuous)d||1.005 (1.001–1.009)||0.02|
|Pretreatment serum PSA: (continuous)d||1.006 (1.001–1.011)||0.02|
|Seminal vesicle involvement (n = 672)c|
|Biopsy Gleason score: (continuous)d||1.566 (1.324–1.851)||< 0.0001|
|Percentage of biopsies positive: (continuous)d||1.013 (1.006–1.021)||0.0006|
|Pretreatment serum PSA: (continuous)d||1.011 (1.005–1.018)||0.0008|
|Lymph node metastases (n = 688)c|
|Clinical stage: T3 vs. ≤ T2||6.99 (2.85–17.15)||< 0.0001|
|Percentage of biopsies positive (continuous)d||1.024 (1.009–1.039)||0.002|
|Lymph node metastases, excluding cT3 (n = 653)c|
|Preoperative PSA > 10 vs. ≤ 10 ng/ml||3.56 (1.13–11.20)||0.03|
|Percentage of biopsies positive: > 50 percent vs. ≤ 50 percent||2.79 (1.01–7.70)||0.05|
The data presented in this article confirm the importance of preoperative serum PSA, clinical stage, and biopsy Gleason score in predicting adverse pathologic features and disease recurrence in patients undergoing radical prostatectomy for localized PC. The significance of factors as predictive parameters in this setting has been demonstrated by many other authors. 1, 7, 20–24
The data also demonstrated that the percentage of diagnostic biopsies involved with PC is a predictor of adverse pathologic features at, and recurrence after, surgery for clinically localized PC. Of note, the percentage of biopsies involved with tumor proved to be a strong independent predictor of adverse pathologic findings when modeled with other preoperative variables, whereas PNI did not. The effect of percentage of biopsies involved was independently predictive overall but became statistically nonsignificant when the model was restricted to patients with a preoperative PSA ≤ 10 ng/mL. Both PNI and percentage of biopsies containing tumor were independent predictors of recurrence in patients with PSA > 10 ng/mL, suggesting a greater effect for these factors in patients at higher risk. However, this was not borne out in subset analyses in patients with higher stage or Gleason score.
Recently, D'Amico et al. have demonstrated the prognostic importance of the number of biopsies involved with PC treated with radical prostatectomy. 4, 25 Using strata based on preoperative serum PSA, clinical stage, and biopsy Gleason score, they demonstrated that the percentage of positive prostate biopsies added clinically significant information regarding time to PSA failure after surgery in a cohort of 960 patients. 4 The model was validated in an independent set of 823 patients. 4 Data from two groups demonstrated that either the total number of biopsies or the percentage of biopsies positive for tumor is predictive of extraprostatic extension (EPE) at RP. 26, 27 Our data suggest that the proportion of biopsies positive for tumor maybe a geographically generalizable predictor of adverse pathologic features and of recurrence and that predictive models incorporating this variable may represent an improvement on those that do not.
Biopsy PNI as a predictor of outcome has been evaluated by a number of investigators with variable results. De la Taille et al. found that biopsy PNI was of independent predictive value for recurrence in 24% of 319 men with PC; was independently prognostic when modeled with pretreatment variables such as clinical stage, PSA, and Gleason score; but lost independence when modeled with pathologic variables principally because it correlated with pathologic stage in a manner similar to that noted in our cohort. 28, 29 This finding runs counter to those of Egan et al., 30 who found in their series of 349 patients that PNI was not an independent factor in determining outcome. The detection of biopsy PNI coupled with tumor with a Gleason score of ≥ 7 has been reported to have a 94% specificity for extracapsular disease at prostatectomy. 31, 32 However, other workers have found, as we did overall, that although PNI on biopsy does predict extracapsular extension at RP, its predictive potential becomes nonsignificant when modeled with biopsy Gleason score, serum PSA, and/or other biopsy features. 30, 33, 34 Stone et al. found that a positive seminal vesicle biopsy, PNI on biopsy, and PSA > 20 ng/mL each independently predicted lymph node involvement in 212 patients with PNI found in 17% of patients. 35 D'Amico et al. 36 recently demonstrated that biopsy PNI provided additional prognostic information when modeled with preoperative PSA, biopsy Gleason score, and percentage of positive prostate biopsies but only in patients who were otherwise at low risk of disease recurrence. A deficit of low-risk criteria patients in our cohort meant that comparable analysis was not possible, and our data suggested utility in patients with preoperative PSA > 10 ng/mL. Rubin et al. 27 reported that biopsy PNI was predictive of EPE in multivariate models but only when factors related to proportional involvement of core biopsies were excluded. Whether modification of surgical technique to resect the neurovascular bundle in patients with PNI on biopsy reduces recurrence rates is unclear. 37
The spread of tumor cells along nerves and vessels is an important mechanism of PC progression. 38 Recent work demonstrates that perineural invasion by PC is associated with a reduced apoptotic index 39 and increased tumor volume. 40 The underlying mechanistic bases for these relations are unclear but may relate to selective neural affinity in cells with apoptotic resistance or the ability to invade the prostate stroma or to neural paracrine factors that inhibit apoptosis and/or stimulate disease progression. 41, 42 PNI in the RP specimens from our cohort was common and prognostic, but this effect was lost in multivariate analysis largely because of the close relation between PNI and pathologic stage variables. 13 Of note, there are few reports of PNI in RP specimens as a prognostic factor 36, 43, 44 despite the finding that reporting of its presence has been recommended. 45, 46 Some reports have suggested that presence of PNI or the volume of PC involved with PNI may predict PSA recurrence. 43, 44, 47 However, these studies suffer from varying methodologic flaws including retrospectivity and lack of central pathology review. In addition, the incidence of RP PNI varies greatly from 39 to 75% between different cohorts. 43, 44, 47 Despite this, and acknowledging the need for well-designed prospective studies focused on PNI, if these data are considered with the incremental prognostic effects of unifocal and multifocal PNI seen in both biopsy and RP specimens in our cohort (Fig. 2C and Quinn et al. 13), they suggest that increasing quantities of PNI predict recurrence. Furthermore, our data suggested that biopsy PNI may be of value in the preoperative delineation of higher-risk patients who have a PSA > 10 ng/mL. Given this, other preoperative assessments of PNI such as imaging for neurotrophic factors in the prostate before surgery represent a potential mechanism by which understaging could be minimized.
Taken together, these data suggested that the percentage of biopsies involved with PC and the presence of perineural invasion have an important relation with disease recurrence after the therapy of early-stage disease. These parameters are readily applicable to everyday clinical and pathologic practice but need further testing in larger cohorts and modeling in predictive nomograms before they can be more widely utilized. 5, 17, 48
The current study data and those of other investigators suggested that the percentage of biopsies involved with PC and the presence of biopsy PNI at the time of diagnosis in clinically localized disease have the potential to refine and improve the predictive accuracy of models designed to quantify risk of disease recurrence in individual patients. This requires investigation and further delineation in large prospective cohorts but may represent a basis on which patients could be offered novel neoadjuvant, adjuvant, or alternative treatment strategies in the context of clinical trials. In the interim, pathologists should consider reporting routinely on the proportion of biopsies involved by tumor and the presence of perineural invasion in those biopsies.
The authors acknowledge the contribution of Anne Maree Haynes and Luis Winoto for data collection and analysis and Dr. John Finlayson for data acquisition and pathologic standardization.
- 19Regression models and life tables. J R Stat Soc. 1972; 34: 187–189..
- 26Predicting prostate carcinoma volume and stage at radical prostatectomy by assessing needle biopsy specimens for percent surface area and cores positive for carcinoma, perineural invasion, Gleason score, DNA ploidy and proliferation, and preoperative serum prostate specific antigen: a report of 454 cases. Cancer. 2001; 91: 2196–2204., , , et al.
- 45Practice protocol for the examination of specimens removed from patients with carcinoma of the prostate gland. A publication of the Cancer Committee, College of American Pathologists. Task Force on the Examination of Specimens Removed From Patients With Prostate Cancer. Arch Pathol Lab Med. 1994; 118: 779–783., , .