Radiation dose has been reported to be an important determinant

538 The Relationship of Increasing Radiotherapy Dose to Reduced Distant Metastases and Mortality in Men with Prostate Cancer Rojymon Jacob, M.D. 1 Alexandra L. Hanlon, Ph.D. 2 Eric M. Horwitz, M.D. 1 Benjamin Movsas, M.D. 1 Robert G. Uzzo, M.D. 3 Alan Pollack, M.D., Ph.D. 1 1 Department of Radiation Oncology, Fox Chase Cancer Center, Philadelphia, Pennsylvania. 2 Department of Biostatistics, Fox Chase Cancer Center, Philadelphia, Pennsylvania. 3 Department of Urologic Oncology, Fox Chase Cancer Center, Philadelphia, Pennsylvania. BACKGROUND. The association of increasing radiotherapy (RT) dose with reduced biochemical failure (BF) is accepted widely. However, there is little direct evidence that dose escalation has an impact on distant metastasis (DM) or overall mortality (OM). These associations were examined in the current study. METHODS. The outcome of 835 patients who were treated at the Fox Chase Cancer Center (Philadelphia, PA) between 1989 and 1997 using 3-dimensional, conformal RT alone (median dose, 74 Gray [Gy]) was analyzed. Stepwise multivariate Cox proportional hazards regression analyses (MVAs) were performed with RT dose included as a covariate along with log-transformed initial pretreatment PSA level, Gleason score, palpation T status, age, and year of treatment (YOT), where indicated. To minimize the effect of YOT, an analysis was performed on a subgroup of 363 patients who were treated prior to 1994. RESULTS. With a median follow-up of 64 months, there were 220 PSA failures, 44 distant metastases, and 162 deaths. In MVA, RT dose (as a continuous variable) was a significant predictor for BF, DM, and OM. When YOT was included as a covariate, it was related strongly to all endpoints, and the correlations of RT dose with DM and OM were lost. When the effect of YOT was minimized by limiting the MVA to patients who were treated prior to 1994, RT dose again emerged as a significant predictor of DM. CONCLUSIONS. Escalation of RT dose reduced the rates of BF, DM, and OM significantly in patients with prostate cancer. The inclusion of YOT had a pronounced effect on these correlations that may confound interpretation. Cancer 2004;100:538 43. 2003 American Cancer Society. Supported in part by Grant CA-06927 from the National Cancer Institute and the Department of Defense and by United States Army Medical Research Grant PC020427. Address for reprints: Alan Pollack, M.D., Ph.D., Department of Radiation Oncology, Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, PA 19111-2497; Fax: (215) 728-2868; E-mail: a_pollack@fccc.edu The views expressed herein are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or the United States Department of Defense. Received July 22, 2003; revision received October 1, 2003; accepted October 20, 2001. KEYWORDS: prostate carcinoma, radiotherapy, dose escalation, metastasis, mortality. Radiation dose has been reported to be an important determinant of outcome in prostate cancer patients treated with radiotherapy (RT) in the prostate-specific antigen (PSA) era. 1 4 However, the effect of RT dose has been manifest primarily using biochemical failure (BF) as the endpoint. Although the pattern of a rising serum PSA level is regarded as an early predictor of distant metastasis (DM), 3,5 8 the relationship of BF to overall mortality (OM) is less well established. 9 11 Moreover, direct evidence that RT dose reduces DM or OM has not been conclusive. 3,12 17 In the current report, the association of RT dose to DM and OM was examined in a population of men with prostate carcinoma who were treated relatively uniformly with three-dimensional (3D)-conformal RT. Because, over the period of the study (1989 1997), stage mign 18 21 and Gleason score shift 22 24 occurred, both of which 2003 American Cancer Society DOI 10.1002/cncr.11927

Radiation Dose Escalation and Mortality/Jacob et al. 539 TABLE 1 Distribution of Patients by Gleason Score and T Stage after Subdivision by Initial Pretreatment Prostate-Specific Antigen Level ipsa Gleason score T stage 2 6 7 10 T1/T2a b 10 ng/ml 371 99 400 70 470 10 ng/ml 268 97 271 94 365 Total 639 196 671 164 835 ipsa: initial pretreatment prostate-specific antigen level. Total complicate retrospective analyses of treatment changes over time, the year of treatment (YOT) was included as a covariate. The effect of the inclusion of YOT on the relationship of RT dose to DM and OM was investigated. MATERIALS AND METHODS Patient Characteristics A total of 835 patients treated consecutively between April, 1989 and November, 1997 using 3D-conformal RT comprised the study cohort. Pathologic grading of diagnostic biopsy specimens was performed at Fox Chase Cancer Center (Philadelphia, PA). An initial pretreatment PSA (ipsa) level was available from all patients, and the median ipsa level was 9.0 ng/ml (range, 0.4 191.0 ng/ml). The median age of the study population was 69 years (range, 45 89 years). Table 1 shows palpation T stage and Gleason score subdivided by ipsa level. The majority of patients (n 470) had an ipsa level 10 ng/ml. Three hundred seventy-one patients had an ipsa level of 10 ng/ml and a Gleason score between 2 and 6. Treatment Patients were immobilized in a custom-made alphacradle cast, supine, before computed tomographicsimulation. The target volumes and technique of RT planning have been described previously. 25,26 In general, a four-field technique using megavoltage irradiation was used. For favorable-risk patients with a T1/T2 classification and a Gleason score of 2 6, only the prostate was treated. For the remaining intermediate- to high-risk patients, the prostate and periprostatic tissues were treated to 46 50 Gray (Gy), then a reduction was made to the prostate and seminal vesicles. A 1-cm planning target volume (PTV) margin was placed around the clinical target volume structures, and a 0.5-cm margin around the PTV typically was placed for penumbra. The median ICRU reference point dose was 74 Gy, with a range of 62.6 84.1 Gy (99% of patients received 68 Gy), with 5% variation across the PTV. None of the patients received neoadjuvant or adjuvant androgen deprivation as part of their treatment. Follow-Up and Endpoints Patients typically were seen at 3 months and then every 6 12 months. PSA values were obtained at 3 months and every 6 12 months thereafter. Radiologic evaluation, as part of the work-up for metastasis, was performed when indicated based on clinical assessment or PSA values. BF, DM, and OM were the endpoints for the study. The study population had a median follow-up of 64 months (range, 9 147 months). BF was defined according to the American Society of Therapeutic Radiology and Oncology (ASTRO) criteria, which include three consecutive rises in serum levels of PSA after reaching a nadir. 27 DM was defined as radiologic confirmation of hematogenous metastasis. Statistical Analysis Estimates of BF, DM, and OM were calculated using the Kaplan Meier product-limit method. 28 Univariate comparisons of outcome were accomplished using the log-rank test. 29 Multivariate analysis (MVA) using stepwise Cox proportional hazards regression models was used to evaluate predictors of outcome. 30 RT dose was used as a continuous variable in MVA and was dichotomized at 76 Gy for univariate comparisons. The 76 Gy cutoff point was chosen because most retrospective series have identified the greatest effects from doses at or above this level. 2,3,26,31 Other covariates for MVA included log-transformed ipsa level (continuous), T stage (T1/T2a,b vs. ), Gleason score (2 6 vs. 7 10), age (continuous), and YOT (continuous). Log-transformed ipsa(log[ipsa]) values were used because the log-transformation resulted in a Gaussian distribution. 6 To minimize the potential confounding effect of YOT as a covariate, the statistical analysis was repeated on the subgroup of 363 patients who were treated between 1989 and 1993; this population had mature follow-up (median, 92 months; range, 11 147 months) and was treated prior to an observed shift in Gleason scoring by pathologists that we determined had occurred after 1994. 32 RESULTS There were 220 BFs, 44 DMs, and 162 deaths (OM) within the study follow-up period. Kaplan Meier s of BF, DM, and OM at 8 years are shown in Table 2. On univariate analysis, Gleason score, ipsa level, and T stage were significant correlates of BF, DM, and OM. Doses of RT 76 Gy were associated with reduced rates of BF and OM. RT dose was not

540 CANCER February 1, 2004 / Volume 100 / Number 3 TABLE 2 Relationship of Various Prognostic Factors to 8-Year Kaplan Meier Estimates of Biochemical Failure, Distant Metastasis, and Overall Mortality No. of patients BF (%) DM (%) OM (%) Gleason score 2 6 639 30 5 25 7 10 196 44 a 16 a 38 a T stage T1/T2a b 671 31 5 25 164 46 a 16 a 37 a ipsa (ng/ml) 0.0 9.9 470 22 6 23 10.0 19.9 245 35 6 32 20 120 73 a 15 a 33 a RT dose (Gy) 76 480 38 8 32 76 355 29 a 6 b 22 a BF: biochemical failure; DM: distant metastases; OM: overall mortality; ipsa; initial pretreatment prostate-specific antigen level; RT: radiotherapy; Gy: Gray. a P 0.05. b P 0.17. TABLE 3 Multivariate Cox Proportional s Analysis using Biochemical Failure as the Endpoint a RT dose Continuous 0.1098 0.0001 0.90 Log(iPSA) Continuous 0.7157 0.0001 2.05 Gleason score 2 6 vs. 7 10 0.6385 0.0001 1.89 T stage T1/T2a b vs. 0.4014 0.013 1.49 RT: radiotherapy; Log(iPSA): log-transformed initial prostate specific antigen level. a Two hundred twenty of 835 patients in the analysis experienced biochemical failure. related to DM in univariate analysis (P 0.17). For patients with tumors T3 and Gleason scores 7, there was a trend toward a lower DM rate with doses 76 Gy (P 0.08). The lack of association between RT dose and DM in univariate analysis may have been caused by the cut-point that was chosen (76 Gy was used because previous studies have indicated that this is an important threshold level) and by an unequal distribution of other factors. MVAs that incorporated RT dose as a continuous variable were performed to address these issues. Tables 3 5 show the results of the MVAs without YOT included as a covariate. RT dose was significantly associated with BF, DM, and OM. The higher the RT dose, the lower the risk of BF, DM, or OM. The other significant covariates were log(ipsa), T stage, and TABLE 4 Multivariate Cox Proportional s Analysis using Distant Metastasis as the Endpoint a RT dose Continuous 0.1600 0.005 0.85 Log(iPSA) Continuous 0.4139 0.011 1.51 Gleason score 2 6 vs. 7 10 1.2506 0.0001 3.49 T stage T1/T2a b vs. 1.2078 0.0002 3.35 RT: radiotherapy; Log(iPSA): log-transformed initial prostate-specific antigen level. a Forty-four of 835 patients had distant metastasis. TABLE 5 Multivariate Cox Proportional s Analysis using Overall Mortality as the Endpoint a RT dose Continuous 0.0668 0.013 0.94 Gleason score 2 6 vs. 7 10 0.4167 0.019 1.52 T stage T1/T2a b vs. 0.4877 0.006 1.63 Age Continuous 0.0659 0.0001 1.07 RT: radiotherapy. a There were 162 deaths overall among the 835 patients in the analysis. TABLE 6 Multivariate Cox Proportional s Analysis using Distant Metastasis as the Endpoint and Including the Year of Treatment as a Covariate a Gleason score 2 6 vs. 7 10 1.3308 0.0001 3.78 T stage T1/T2a b vs. 1.1123 0.0004 3.04 YOT Ordinal 0.4622 0.0001 0.63 YOT: year of treatment. a Forty-four of 835 patients had distant metastasis. Gleason score. Age was related significantly to OM, but not to BF or DM. When YOT was included in the MVAs as a continuous variable, it was associated significantly with BF (P 0.0001), DM (P 0.0001), and OM (P 0.0001). RT dose remained related independently to BF (not shown) but not to DM (Table 6) or OM (Table 7). Log(iPSA) continued to be related significantly to BF but not to DM or OM. Because YOT was significantly related to RT dose (P 0.0001 analysis of variance) and, therefore, could

Radiation Dose Escalation and Mortality/Jacob et al. 541 TABLE 7 Multivariate Cox Proportional s Analysis using Overall Mortality as the Endpoint and Including the Year of Treatment as a Covariate a Gleason score 2 6 vs. 7 10 0.4174 0.0179 1.52 T stage T1/T2a b vs. 0.3560 0.0443 1.43 YOT Continuous 0.1909 0.0001 0.83 Age Continuous 0.0594 0.0001 1.06 YOT: year of treatment. a There were 162 deaths overall among the 835 patients in the analysis. TABLE 8 Multivariate Cox Proportional s Analysis using Distant Metastasis as the Endpoint in the Subset of 363 Patients who were Treated before 1994 a RT dose Continuous 0.2115 0.0019 0.81 Gleason score 2 6 vs. 7 10 1.2531 0.0021 3.50 T status T1/T2a b vs. 1.5452 0.0001 4.69 RT: radiotherapy. a Twenty-eight of the 363 patients had distant metastasis. confound the association of RT dose to patient outcome, a subgroup of patients treated between 1989 and 1993 (n 363) was investigated. In this subgroup, there were 132 BFs, 28 DMs, and 116 deaths (OM). Median follow-up was 92 months. In MVA, RT dose, Gleason score 7 10, and T3 category were related significantly to DM (Table 8); YOT was not significant and was not included in the final model. The only significant predictor for OM was age. DISCUSSION Many retrospective and prospective studies have confirmed a reduction in BF as a consequence of increasing RT dose. 1 4 However, there is little evidence of the impact of RT dose on DM and OM. In the University of Texas M. D. Anderson Cancer Center s randomized radiation dose escalation prostate carcinoma trial, which compared a 70 Gy to 78 Gy, 3 the DM rate was lower for patients with a PSA level 10 ng/ml who were treated with 78 Gy (2% vs. 12% at 6 years; P 0.056). No relation between RT dose and OM was seen in that study, because there were few deaths attributable to prostate carcinoma. Also, in a matched-pair analysis that compared 74 Gy 74 Gy by Hanks et al., 15 RT dose was associated significantly with DM and cause-specific death for patients who were matched for ipsa, Gleason score, and T stage. Kestin et al. 16 also did a matched-pair analysis and found that higher biologic RT doses were associated with reduced OM; however, those authors did not find any relation of dose with DM or cause-specific death. In one of the few studies wherein MVA was performed, Valicenti et al. 17 found that RT dose was an independent predictor of cause-specific death and OM for patients who had Gleason scores of 8 10. That analysis involved patients who were treated in Radiation Therapy Oncology Trials in the pre-psa era; and, although there was a range of doses used, 87% of patients were treated with conventional doses 70.2 Gy. No adjustment was made for time-related changes, such as stage mign or differences in follow-up. To our knowledge, no previous PSA-era multivariate study of patients who were treated mainly with conventional or greater RT doses has investigated the impact of dose on DM or OM and included a covariate (e.g., YOT) to account for such time-related concerns. The data shown here reveal a strong association between increasing RT dose and reduced DM and OM. Although the MVAs documented that RT dose was independent of T stage, Gleason score, and ipsa, there are concerns about reliability. Over the course of the study (1989 1997), there were notable changes in patient attributes. The use of PSA and new or improved imaging modalities in screening (e.g., ultrasound and endorectal coil magnetic resonance imaging) and transrectal, ultrasound-guided, sextant or greater biopsies contributed to earlier diagnosis. These features, combined with changes in the American Joint Committee on Cancer staging system, such as the inclusion of imaging and biopsy findings, have resulted in stage mign. 18 21 A prostate cancer patient with T2b disease today is not the same as a patient with T2b disease 5 10 years ago. Another, perhaps even more significant factor that obscures retrospective studies is the shift in Gleason scoring that has occurred over the study period. 22 24 Recently, Chism et al. 32 found that pathologists consistently are assigning higher Gleason scores in diagnostic prostate biopsy specimens today compared with the scores assigned prior to 1994. Thus, a tumor may have been assigned a Gleason score of 6 in the early 1990s and a score of 7 in the late 1990s. Such time-related changes would improve the outcome of patients with Gleason 6 and Gleason 7 disease artificially over time. Distinguishing between the effect of RT dose, which also was altered over time, and follow-up differences (patients who received higher RT doses gen-

542 CANCER February 1, 2004 / Volume 100 / Number 3 erally were treated more recently), 33 stage mign, 18 21 Gleason score shift, 22 24 and possibly other factors affecting patient outcome is not possible without the inclusion of a covariate that adjusts for these changes. This is the reason for the inclusion of YOT as a covariate. The influence of the time-dependent changes that occurred in follow-up, stage classification, and Gleason scoring should be negated by YOT; however, because RT dose is associated significantly with YOT, any impact of RT dose on outcome also would be blunted. Indeed, when YOT was included in the MVAs, RT dose no longer was associated with DM or OM. There still was a strong relationship between RT dose and BF. One way to limit the influence of time-related factors is to perform the MVAs on patients who were treated over a shortened time span. When only patients treated between 1989 and 1993, inclusive, were considered, it was found that RT dose was related to DM. Under those conditions, YOT was not significant in MVA and was not included in the final model (Table 8). These data provide much more convincing evidence that RT dose is effective at reducing DM. The inability to document an association between RT dose and OM in this cohort probably was related to the limited number of patients; death due to intercurrent disease in patients with prostate cancer is common, necessitating large numbers of patients when OM is used as the endpoint. In summary, increasing RT dose improved outcomes in patients with clinically localized prostate carcinoma. The association with BF is undisputed; however, the association with DM and OM has been less certain. Our data show that RT dose also is related significantly to DM. We were only able to show an association between RT dose and OM without correcting for time-related changes (exclusion of YOT as a covariate); a larger patient cohort will be needed to establish this association more conclusively. REFERENCES 1. Lyons J, Kupelian P, Mohan D, Reddy C, Klein EA. Importance of high radiation doses (72 Gy or greater) in the treatment of Stage T1 T3 adenocarcinoma of the prostate. Urology. 2000;55:85 90. 2. Zelefsky MJ, Fuks Z, Hunt M, et al. High dose radiation delivered by intensity modulated conformal radiotherapy improves the outcome of localized prostate cancer. J Urol. 2001;166:876 881. 3. Pollack A, Zagars GK, Starkschall G, et al. Prostate cancer radiation dose response: results of the M. D. Anderson Phase III randomized trial. Int J Radiat Oncol Biol Phys. 2002;53:1097 1105. 4. Hanks GE, Hanlon AL, Epstein B, Horwitz EM. Dose response in prostate cancer with 8 12 years follow-up. Int J Radiat Oncol Biol Phys. 2002;54:427 435. 5. Zagars GK, Pollack A. The fall and rise of prostate-specific antigen. Kinetics of serum prostate-specific antigen levels after radiation therapy for prostate cancer. Cancer. 1993;72: 832 842. 6. Pollack A, Zagars GK, Kavadi VS. Prostate specific antigen doubling time and disease relapse after radiotherapy for prostate cancer. Cancer. 1994;74:670 678. 7. Zagars GK, Pollack A. Kinetics of serum prostate-specific antigen after external beam radiation for clinically localized prostate cancer. Radiother Oncol. 1997;44:213 221. 8. Pound C, Partin A, Eisenberger M, Chan D, Pearson J, Walsh P. Natural history of progression after PSA elevation following radical prostatectomy. JAMA. 1999;281:1591 1597. 9. Sandler HM, Dunn RL, McLaughlin PW, Hayman JA, Sullivan MA, Taylor JM. Overall survival after prostate-specificantigen-detected recurrence following conformal radiation therapy. Int J Radiat Oncol Biol Phys. 2000;48:629 633. 10. Kwan W, Pickles T. In regard to Kupelian et al.: impact of biochemical failure on overall survival after radiation therapy for localized prostate cancer in the PSA era. IJROBP 2002;52:704 711. Int J Radiat Oncol Biol Phys. 2002;54:1577 1579; author reply, 1579. 11. Kupelian PA, Buchsbaum JC, Patel C, et al. Impact of biochemical failure on overall survival after radiation therapy for localized prostate cancer in the PSA era. Int J Radiat Oncol Biol Phys. 2002;52:704 711. 12. Shipley WU, Verhey LJ, Munzenrider JE, et al. Advanced prostate cancer: the results of a randomized comparative trial of high dose irradiation boosting with conformal protons compared with conventional dose irradiation using photons alone. Int J Radiat Oncol Biol Phys. 1995;32:3 12. 13. Fiveash J, Hanks G, Roach M, et al. 3D conformal radiation therapy (3D CRT) for high grade prostate cancer: a multiinstitutional review. Int J Radiat Oncol Biol Phys. 2000;47: 335 342. 14. Magrini SM, Cellai E, Pertici M, et al. Radical radiotherapy of localised prostate cancer: the relationship between radiation dose and survival. Cancer Radiother. 1998;2:351 358. 15. Hanks G, Hanlon A, Pinover W, Horwitz E, Schultheiss T. Survival advantage for prostate cancer patients treated with high dose 3D conformal radiation. Cancer J Sci Am. 1999;5: 152 158. 16. Kestin LL, Martinez AA, Stromberg JS, et al. Matched-pair analysis of conformal high-dose-rate brachytherapy boost versus external-beam radiation therapy alone for locally advanced prostate cancer. J Clin Oncol. 2000;18:2869 2680. 17. Valicenti R, Lu J, Pilepich M, Asbell S, Grignon D. Survival advantage from higher-dose radiation therapy for clinically localized prostate cancer treated on the Radiation Therapy Oncology trials. J Clin Oncol. 2000;18:2740 2746. 18. Amling CL, Blute ML, Lerner SE, Bergstralh EJ, Bostwick DG, Zincke H. Influence of prostate-specific antigen testing on the spectrum of patients with prostate cancer undergoing radical prostatectomy at a large referral practice. Mayo Clin Proc. 1998;73:401 406. 19. Hankey BF, Feuer EJ, Clegg LX, et al. Cancer surveillance series: interpreting trends in prostate cancer Part I: evidence of the effects of screening in recent prostate cancer incidence, mortality, and survival rates. J Natl Cancer Inst. 1999;91:1017 1024. 20. Jhaveri FM, Klein EA, Kupelian PA, Zippe C, Levin HS. Declining rates of extracapsular extension after radical prostatectomy: evidence for continued stage mign. J Clin Oncol. 1999;17:3167 3172.

Radiation Dose Escalation and Mortality/Jacob et al. 543 21. Ung JO, Richie JP, Chen MH, Renshaw AA, D Amico AV. Evolution of the presentation and pathologic and biochemical outcomes after radical prostatectomy for patients with clinically localized prostate cancer diagnosed during the PSA era. Urology. 2002;60:458 463. 22. Schellhammer PF, Moriarty R, Bostwick D, Kuban D. Fifteen-year minimum follow-up of a prostate brachytherapy series: comparing the past with the present. Urology. 2000; 56:436 439. 23. Smith EB, Frierson HF Jr., Mills SE, Boyd JC, Theodorescu D. Gleason scores of prostate biopsy and radical prostatectomy specimens over the past 10 years: is there evidence for systematic upgrading? Cancer. 2002;94:2282 2287. 24. Chism DB, Hanlon AL, Troncoso P, Horwitz EM, Pollack A. Impact on outcome of the Gleason score shift. Int J Radiat Oncol Biol Phys. 2002;54:265 266. 25. Hanks GE, Hanlon AL, Schultheiss TE, et al. Dose escalation with 3D conformal treatment: five year outcomes, treatment optimization and future directions. Int J Radiat Oncol Biol Phys. 1998;41:501 510. 26. Horwitz EM, Hanlon AL, Pinover WH, Anderson PR, Hanks GE. Defining the optimal radiation dose with three-dimensional conformal radiation therapy for patients with nonmetastatic prostate carcinoma by using recursive partitioning techniques. Cancer. 2001;92:1281 1287. 27. Cox J, Grignon D, Kaplan R, Parsons J, Schellhammer P. Consensus statement: guidelines for PSA following radiation therapy. Int J Radiat Oncol Biol Phys. 1997;37:1035 1041. 28. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:447 457. 29. Mantel N. Evaluation of survival data and two new rank order statistics arising in its considen. Cancer Chemother Rep. 1966;50:163 170. 30. Cox DR. Regression models and life tables. J R Stat Soc B. 1972;34:187 220. 31. Pollack A, Smith L, von Eschenbach A. External beam radiotherapy dose-response characteristics of 1127 men with prostate cancer treated in the PSA era. Int J Radiat Oncol Biol Phys. 2000;48:507 512. 32. Chism D, Hanlon AL, Troncoso P, Horwitz EM, Pollack A. The Gleason score shift: score four and seven years ago. Int J Radiat Oncol Biol Phys. 2002;56:1241 1247. 33. Vicini F, Kestin L, Martinez A. The importance of adequate follow-up in defining treatment success after external beam irradiation for prostate cancer. Int J Radiat Oncol Biol Phys. 1999;45:553 561.