BJU International (1999), 84, 37–42

Mortality and prostate cancer risk in 19 598 men aftersurgery for benign prostatic hyperplasiaC.D.J. HOLMAN, Z .S. WISNIE WSKI*, J .B. SEMMENS, I.L. ROUSE†and A.J . BASSCentre for Health Services Research, Department of Public Health, The University of Western Australia, East Perth, *Quality ofSurgical Care Committee, Royal Australasian College of Surgeons, Western Australian Branch, Nedlands, and † Health InformationCentre, Health Department of Western Australia, East Perth, Western Australia

Objectives To examine postoperative mortality and pros- than OP (rate ratio, RR, 1.20; 95% confidence interval1.08–1.34). The RR fell to 1.10 (0.99–1.23) aftertate cancer risk after the first prostatectomy for benign

prostatic hypertrophy over a 17-year period in a adjustment for comorbidity and to 1.07 (0.95–1.19)when accounting for nonlinearity. The relative sur-population-based cohort of men in Western Australia,

using improved methods to adjust for comorbidity. vival from the incidence of prostate cancer at 10 yearswas 103.7% after TURP and 104.5% after OP. ThePatients and methods The relative survival from death

and prostate cancer incidence was calculated against RR adjusted for age and calendar year was 1.44(0.94–2.21) for incidence and 1.37 (0.81–2.29) forthe background population rates. The outcomes of

transurethral resection of the prostate (TURP) and prostate cancer mortality.Conclusion There is at most a small and clinicallyopen prostatectomy (OP) were compared adjusting for

calendar year, age, admission type and comorbidity unimportant excess mortality risk from TURP; anydiCerence could be due to a protective eCect of OP onusing Cox regression. Fractional polynomials were

used to take account of nonlinearity in confounder the long-term risk of prostate cancer and a lower rateof repeat prostatectomy.eCects.

Results At 10 years, the relative survival was 116.5% Keywords Benign prostatic hyperplasia, prostatectomy,TURP, outcomes, mortality, prostate cancer, recordin TURP patients and 123.5% after OP. Adjusting

only for confounding by age, calendar year and linkageadmission type, TURP had a higher mortality rate

the procedure itself or other factors has not been fullyIntroduction

resolved. Similarly, the question of whether surgery forBPH aCects developing prostate cancer and whether theBPH is found at unselected autopsy in 40% of men aged

50 years and almost 90% of men over 80 years of age type of operation has any eCect has not been fullyevaluated [16,17].[1]. BPH causes significant urological morbidity and

mortality, with up to 10% of men undergoing prostatec- The present study used data from the Quality ofSurgical Care Project in Western Australia (WA) from atomy in their lifetime. Surgery is designed to remove the

occlusive adenoma whilst leaving the peripheral zone Health Services Research Linked Database [18]. Thiscontains population-based information linking all hospi-with its false capsule. Open prostatectomy (OP) is still

indicated for large glands, but has been supplanted by tal admissions, deaths and cancer registrations in a Stateof 1.7 million people over a 17-year period. These dataendoscopic TURP, which is now the preferred method

of therapy. were used to examine postoperative mortality and pros-tate-cancer risk after primary prostatectomy for BPH,TURP has been associated with a dramatic decline in

perioperative morbidity and mortality [2]. However, it is comparing open and endoscopic surgery.of concern that among 15 evaluations of surgical out-comes for BPH [2–16], four large population-based stud- Patients and methodsies found a consistently higher long-term mortality for

The WA-linked Database was used to extract all hospitalTURP than for OP [5,7,14,16]. Whether this apparentmorbidity data, death records and prostate cancer regis-eCect on late mortality was caused by patient selection,trations of men who had a hospital separation in1980–95 with mention of prostate surgery (ICD-9-CM60.0-6 [19] or ICPM 5600–5 [20]); a principal conditionAccepted for publication 9 March 1999

37© 1999 BJU International

38 C.D.J . HOL MAN et al.

of benign or malignant prostate disease (ICD-9-CM or Any mention of each of the 17 diagnostic categories ofthe Charlson Index on the hospital morbidity record ofICD-9 222.2, 600.0–602.9 [21]); or a death in 1980–95

from benign or malignant prostate disease; the file was either the index admission (using restricted codes to limitinclusion of complications [27]), or any prior hospitaldated 1 March 1997.

The following case definition was used for the study admission with a separation date within 365 days of thedate of index admission, was counted as the presence ofof outcomes of surgery for BPH; a case was any man

aged �20 years who separated from an acute hospital a comorbid condition.Based on the frequencies of significant comorbid con-in WA in 1980–95, after a first prostatectomy for a

principal diagnosis of BPH (ICD-9-CM or ICD-9600). ditions, four diagnostic categories in addition to the 17comprising the Charlson Index were brought into theCases were restricted to those with no mention of prostate

cancer or bladder cancer in any previous admission, in method of comorbidity adjustment. These were otherischaemic heart disease and chest pain (ICD-9-CM orthe admission when the first prostatectomy occurred

(index admission) nor in any admission within 1 year ICD-9413–4, 786.5), cardiac dysrhythmias (427),hypertensive disease (401–5) and obesity (278). Wefrom index separation. Also excluded was any man who

had a prostate cancer registration preceding or occurring calculated a comorbidity summarisation score using thesum of the Cox regression coeBcients assigned to the 21within 1 year from index separation.

There were 19 598 cases that met the definition; the comorbid conditions [28,29]. In addition, the totalnumber of bed days ascribed to all previous hospital140 629 hospital morbidity records of the cases were

linked to 4845 death records and 355 prostate cancer admissions with a date of separation occurring within365 days of the index admission date was included as aregistrations, the latter occurring more than 1 year after

index separation. There were also 211 deaths in the second means of adjusting for comorbidity.Account was taken of the presence of nonlinearity incohort from prostate cancer. Linked death records were

available for the 17-year period 1980–96; all otherTable 1 Cumulative mortality and relative survival after surgerylinked records covered the 16-year period 1980–95.for benign prostatic hyperplasia, and cumulative incidence, relativeCumulative mortality at 31 and 93 days, and 1, 5survival and cumulative mortality from prostate cancer afterand 10 years was calculated using actuarial methodssurgery in Western Australia in 1980–95 according to procedure

[22]. Relative survival was estimated with a computer typeprogram developed in the Mayo Clinic [23], using theHakulinen method [24,25], and modified to include Closed procedures Open procedures

(18 464 TURPs) (1134 OPs)yearly age-specific mortality rates in WA males, to obtainexpected survival. Mortality rate ratios (RRs) comparing

BPHTURP with OP were obtained from Cox regression modelsCumulative mortality (%)[22]. Rate ratios were adjusted for calendar year, age,31 days 0.51 (0.05)* 0.62 (0.23)emergency vs elective admission type and comorbidity.93 days 1.53 (0.09) 1.59 (0.37)

The cumulative incidence and cumulative mortality 1 year 4.74 (0.16) 4.33 (0.60)from prostate cancer at 5 and 10 years after index 5 years 17.88 (0.31) 18.10 (1.18)admission were calculated using actuarial methods 10 years 31.77 (0.45) 34.97 (1.56)

Relative survival (%)[22]. Relative survival from cancer incidence was31 days 99.85 (0.05) 99.83 (0.23)calculated using the Mayo Program [23–25], with93 days 99.57 (0.09) 99.76 (0.38)yearly age-specific incidence rates of prostate cancer in1 year 99.54 (0.16) 100.85 (0.64)WA males used as the hazard rates in a comparable5 years 104.20 (0.39) 109.16 (1.57)

group of men. Expected survival was expressed relative 10 years 116.50 (0.78) 123.51 (2.96)to that at 1 year to adjust for the eCects of the initial Prostate cancerrisk-free period in the study population. Neither comor- Cumulative incidence (%)

5 years 1.13 (0.09) 0.63 (0.26)bidity nor admission type predicted prostate cancer10 years 3.29 (0.21) 2.43 (0.60)risk, and RRs comparing TURP with OP obtained fromRelative survival fromCox regression were adjusted only for calendar time

incidence (%)and age.5 years 101.66 (0.10) 101.85 (0.27)

Previous research has suggested the importance of 10 years 103.67 (0.22) 104.46 (0.66)adjustment for comorbidity in comparing postoperative Cumulative mortality (%)mortality between TURP and OP [2–16]. Diseases rep- 5 years 0.65 (0.07) 0.72 (0.27)

10 years 1.85 (0.16) 1.50 (0.44)resented in the Charlson Index of comorbidity (seeAppendix 1) were identified using the Dartmouth-

* Standard errors in parentheses.Manitoba adaptation for administrative data [26,27].

© 1999 BJU International 84, 37–42

RISK AFTER SURGERY FOR BENIGN PROSTATIC HYPERPLASIA 39

the relationships of mortality to the comorbidity sum- was explained because TURP patients, on average, wereyounger, had their surgery more recently and were moremarisation score and bed-day sum, using a system of

second-degree fractional polynomials to test 21 diCerent likely to have elective admissions than the OP patients.Once these factors were adjusted, there was a significantnonlinear models [30]. Best fit was obtained from xD+x

for the summarisation score and from x−D+lnx for the increase in postoperative mortality in the TURP group(1.20, 1.08–1.34). However, this result was reduced tobed-day sum; i.e. from fractional polynomials with val-

ences of (12, 1) and (−1

2, 0) [30]. Also included in the a level of no practical or statistical significance, first by

adjustment for comorbidity using the summarisationfinal model was a binary indicator term for no observedcomorbidity, consistent with the arguments of Robertson score and comorbid bed-days (1.10, 0.99–1.23), and

second, by taking account of nonlinearity in the eCectset al. [31]. The analysis of postoperative mortality usingCox regression excluded procedures in 1980, because it of age and comorbidity (1.07, 0.95–1.19; Table 2).

Another result shown in Table 2 is that the adjustedwas necessary to have 1 year of comorbidity dataavailable before surgery. postoperative mortality rate fell by 32% from 1981–84

to 1990–95, suggesting an overall improvement inanaesthetic and surgical risk for patients. However, even

Resultsafter adjustment for age, comorbidity, calendar year andprocedure type, patients admitted in an emergency hadThere were only small diCerences in postoperative cumu-

lative mortality, with the risk of death after TURP a little almost twice the postoperative mortality of those bookedas elective admissions for surgery.less than after OP (Table 1; Wilcoxon test, P=0.418).

These results do not take into account deaths expected Table 1 also shows comparisons of the cumulativeincidence of and cumulative mortality from prostateon the basis of general population experience. The

relative survival results are also shown in Table 1. At 5 cancer between TURP and OP. The risk of developing ordying from prostate cancer was higher after TURP thanyears, survival of all prostatectomy patients was signifi-

cantly better than in the general population of WA men. after OP, although the diCerence was not significance(Wilcoxon test, P=0.071 for incidence, P=0.847 forBy 10 years the relative survival was as high as 116.5%

in TURP patients and 123.51% after OP. There was little mortality). Even after adjusting for the mandatory 1-yearprostate cancer-free interval, included in the case defi-diCerence between TURP and OP in relative survival up

to 1 year, but thereafter the relative survival was higher nition, the relative survival from prostate cancer inci-dence in both procedure groups exceeded 100%after OP.

Cox regression analyses were used to compare the (Table 1), and the standard errors indicated a signifi-cantly lower risk than that expected on the basis ofmortality of patients undergoing closed and open pro-

cedures, adjusted for age, calendar year, admission type general population rates.The results of the Cox regression analyses of prostateand comorbidity (Table 2). Before adjustment, patients

undergoing TURP had a lower mortality rate than those cancer risk suggested empirically that the incidence ratewas 44% higher after TURP than after OP and thehaving OP (RR 0.87, 95% CI 0.78–0.97). However, this

Table 2 Cox regression of postoperative mortality after surgery for BPH in Western Australia in 1981–95; (n=18 747)*

Crude mortality rate ratio Adjusted† mortality rate ratioFactor (95% CI) (95% CI)

Procedure type, TURP/OP 0.87 (0.78–0.97) 1.07 (0.95–1.19)Calendar period1981–84 1.00 1.001985–89 0.95 (0.89–1.02) 0.87 (0.81–0.93)1990–94 0.64 (0.59–0.70) 0.68 (0.62–0.74)Age (per year) 1.090 (1.086–1.093) 1.075 (1.071–1.079)Admission type: emergency/elective 2.78 (2.61–2.95) 1.92 (1.80–2.04)Comorbidity summarisation score (per score unit) 2.72 (2.61–2.83) 2.07 (1.99–2.17)Comorbid bed days in preceding year (per bed day) 1.0030 (1.0027–1.0034) 1.0024 (1.0019–1.0030)

*Cases operated in 1980 were excluded to provide 1 year of lead time for the calculation of comorbidity indices.†Calendar period was adjusted for procedure type, age, admission type, comorbidity summarisation score and comorbid bed days. Theremaining factors were adjusted for procedure type, calendar year, age, admission type, comorbidity summarisation score and comorbidbed days. A binary indicator of no comorbidity was included throughout. Age, comorbidity summarisation score and comorbid bed dayswere adjusted using fractional polynomials.

© 1999 BJU International 84, 37–42

40 C.D.J . HOL MAN et al.

mortality rate from prostate cancer was 37% higher Scottish study found that the late death rate fromprostate cancer was greater in patients who underwent(Table 3). Both results were also consistent with either

no true eCect or more than twice the diCerence. TURP than OP [16].The results reported here are more consistent with

those from Connecticut [11], New York [15], OxfordDiscussion

[14] and Japan [13], and less consistent with those fromManitoba [5], Denmark [7], Northern California [10]These results suggest that the risk of postoperative

mortality in men undergoing TURP and OP for benign and Scotland [16]. The present final result is similar tothe adjusted RR of 1.03 (0.57–1.87) reported from aprostate disease is essentially similar. The best estimate

of the mortality RR comparing TURP with OP was 1.07 series of 252 patients treated at Yale-New Haven(Connecticut) Hospital [11], and the 1.05 reported for(0.95–1.19). These results also indicate that the risk of

mortality after TURP is no higher and perhaps slightly 1315 men undergoing surgery at the VeteransAdministration Medical Center in Brooklyn, New Yorkless than after OP in the immediate postoperative period.

The cumulative mortality at 31 days was 0.51% after [15]. The present study confirms an insignificant RR ofthis order in a much larger and population-based seriesTURP and 0.62% after OP.

Roos et al. [5], reporting from Manitoba, found that of 19 598 patients.Several confounding eCects were observed; TURPTURP had a higher long-term mortality than OP

(RR 1.45, 1.15–1.84). This finding was replicated in a patients were younger, had their operation more recentlyand were more often booked for elective surgery ratherDanish population-based study by Anderson et al.

(RR 1.19, 1.15–1.24) [7], and then in data from seven than admitted in an emergency. These attributes con-ferred a favourable mortality risk compared with OP, ashospital sites in the USA [8], and a health maintenance

organization in Northern California [10]. A plausible shown by the change in RR from 0.87 before adjustmentto 1.20 afterwards. However, TURP patients also had abiological mechanism for the eCect was not evident. The

Manitoba and Danish studies were adjusted for patient higher level of comorbidity, which confounded the eCectof procedure type in the opposite direction. Once comor-age and comorbidity, thus allowing for the observation

that men having TURP were, on average, less healthy bidity was adjusted for using the Charlson-based comor-bidity score and comorbid bed-days, the RR fell fromthan those undergoing OP [5,7].

Seagroatt, reporting from Oxford [14], found that 1.20 to 1.10. A further significant improvement in thegoodness of fit (P<0.001) and a further reduction inTURP had a higher 1-year mortality than OP (RR 1.81

in 1963–72, falling to 1.23 in 1981–85), but she the RR from 1.10 to 1.07 was achieved by modellingthe nonlinear relationships between the log (RR) andobserved also that the long-term mortality after TURP

was close to background population rates, and that there age, comorbidity score and comorbid bed-days. We con-tend that without the use of fractional polynomials towas no increase in 1-year death rates after prostatectomy

as TURP replaced OP. These observations, combined optimize the fit of the models to the eCects of theseexplanatory variables, the eCect of TURP on postopera-with results of hospital-based studies, which found no

excess mortality after risk adjustment [11,13,15], sup- tive mortality would have been over-estimated.The present result is consistent with a possible protec-ported a view that the apparent higher long-term mor-

tality after TURP was unlikely to be caused by the tive eCect of OP against the future development ofprostate cancer, as reported from Scotland [16]. Alongoperation itself, but rather as the consequence of OP

patients being fitter than those having TURP. with other investigators, we have also found a reductionby about a half in the rate of repeat prostatectomy afterHowever, Hargreave et al. [16] have since published

results from a large population-based study in Scotland, OP compared with TURP [2,3,5,8,10]. It is possible thatthese associations confer a small survival advantage,showing the RR of late mortality after TURP compared

with OP to be 1.15 (1.11–1.19) after adjusting for because OP patients are less likely to die from prostatecancer and less likely to experience a further period ofcomorbidity and age. Moreover, on the question of

prostate cancer risk after surgery for BPH [17], the postoperative risk after another prostatectomy.

Prostate cancer eCect measure Crude rate ratio (95% CI) Adjusted* rate ratio(95% CI)

Incidence rate ratio: TURP/OP 1.33 (0.87–2.03) 1.44 (0.94–2.21)Mortality rate ratio: TURP/OP 1.17 (0.70–1.95) 1.37 (0.81–2.29)

*Procedure type was adjusted for calendar year and age, the later as a fractional polynomial.

Table 3 Cox regression analysis of prostatecancer risk after surgery for BPH inWestern Australia in 1980–95 (n=19 598)

© 1999 BJU International 84, 37–42

RISK AFTER SURGERY FOR BENIGN PROSTATIC HYPERPLASIA 41

In other results, we have found that the removal of use of VLAP in a clinical rather than experimentalsetting, but note that long-term follow-up data areall prostate cancer deaths from the analysis and cen-

soring of follow-up at the date of a repeat prostatectomy lacking [35]. It will be important therefore to undertakepostmarketing surveillance of laser prostatectomy usingfurther reduced the adjusted RR, firstly to 1.05

(0.94–1.18) and then to 1.04 (0.93–1.17). We cannot population-based record linkage, as has occurred withTURP.be confident that TURP patients experience a higher rate

of prostate cancer and a higher rate of postoperativemortality due to repeat prostatectomy based on the

Acknowledgementspresent data alone. However, if these conjectures aretrue, they could explain a very small and clinically This research was supported by the National Health and

Medical Research Council of Australia. The Healthunimportant diCerence in the long-term mortality risks.Whether possible diCerences in risks of this magnitude Services Research Linked Database was supported by the

Lotteries Commission of Western Australia. We areand degree of uncertainty should be discussed withpatients is a moot point, although the possible need for indebted to Mr Michael Lawrence-Brown, Professor

David Fletcher and fellows of the WA Branch of thea repeat prostatectomy after TURP should be explained.The absolute level of postoperative mortality observed Royal Australasian College of Surgeons for their collabor-

ation in the Quality of Surgical Care Project.in WA after prostatectomy is comparable with that inother health systems [9]. The decrease of about one-third in the rate of postoperative death in 1990–95

Referencescompared with 1981–84 may reflect improvements inanaesthetic and surgical technique, especially given that 1 Berry SJ, CoCey DS, Walsh PC, Ewing LL. The development

of human benign prostatic hyperplasia with age. J Urolthe result was adjusted for diCerences over time in the1984; 132: 474–9age distribution of patients, their circumstances of

2 Mebust WK, Holtgrewe HL, Cockett ATK, Peters PC, Writingadmission and their levels of comorbidity. The long-termCommittee. Transurethral prostatectomy: immediate andsurvival of all prostatectomy patients was higher thanpostoperative complications. A cooperative study of 13in the general population of WA men. This suggests theparticipating institutions evaluating 3,885 patients. J Urol

presence of a selection eCect, which has excluded high-1989; 141: 243–7

risk males from surgery. 3 Wennberg JE, Roos N, Sola L, Schori A, JaCe R. Use ofLike Hammarsten et al. [17], we found no evidence of claims data systems to evaluate health care outcomes.

an increase in prostate cancer risk after surgery for BPH JAMA 1987; 257: 933–6relative to general population rates, despite speculation 4 Roos NP, Ramsey EW. A population-based study of

prostatectomy: outcomes associated with diCering surgicalthat surgical trauma may disseminate or promote theapproaches. J Urol 1987; 137: 1184–8growth of cancer cells. However, as noted by Guess [32],

5 Roos NP, Wennberg JE, Malenka DJ et al. Mortality andthe exclusion of patients incidentally found to havereoperation after open and transurethral resection of theprostate cancer upon histological examination of tissueprostate for benign prostatic hyperplasia. New Eng J Medremoved from a supposedly benign enlargement of the1989; 320: 1120–4prostate, produces a selection bias towards men who are

6 Roos LL, Fisher ES, Sharp SM, Newhouse JP, Anderson G,free of latent prostate cancer, because no equivalent

Bubolz TA. Postsurgical mortality in Manitoba and Newexclusion can be made in the general population. England. JAMA 1990; 263: 2453–8

Importantly, the present study provides confirmation 7 Andersen TF, Bronnum-Hansen H, Sejr T, RoepstorC C.that TURP is a safe procedure for the surgical treatment Elevated mortality following transurethral resection of theof BPH in a large and recently assembled population- prostate for benign hypertrophy! But why? Medical Care

1990; 28: 870–81based cohort of patients followed for up to 17 years.8 Taylor Z, Krakauer H. Mortality and reoperation followingTURP is established as the standard technique for the

prostatectomy: outcomes in a Medicare population. Urologyvast majority of glands requiring operation. Laser ther-1991; 38 (Suppl): 27–31apy for BPH, and in particular visual laser assisted

9 Stephenson WP, Chute CG, Guess HA, Schwartz S, Lieberprostatectomy (VLAP) has been compared directly withM. Incidence and outcome of surgery for benign prostaticTURP in randomized controlled trials with follow-uphyperplasia among residents of Rochester, Minnesota

periods of up to 12 months [33–35]. The results have1980–87. Urology 1991; 38 (Suppl): 32–42

shown comparable medium-term outcomes, although 10 Sidney S, Quesenberry CP, Sadler MC, Cattolica EV, Lydickearly irritative symptoms are more common after VLAP EG, Guess HA. Reoperation and mortality after surgical[35]. Peak medical advisory groups, such as the National treatment of benign prostatic hypertrophy in a largeHealth and Medical Research Council of Australia, have prepaid medical care program. Medical Care 1992; 30:

117–25recommended that there are suBcient data to permit the

© 1999 BJU International 84, 37–42

42 C.D.J . HOL MAN et al.

11 Concato J, Horwitz RI, Feinstein AR, Elmore JG, SchiC SF. data: diCering perspectives. J Clin Epidemiol 1993; 46:1075–9Problems of comorbidity in mortality after prostatectomy.

JAMA 1992; 267: 1077–82 28 Miettinen OS. Stratification by multivariate confounderscore. Am J Epidemiol 1976; 104: 609–2012 Thorpe AC, Cleary R, Coles J, Vernon S, Reynolds J, Neal

DE. Deaths and complications following prostatectomy in 29 Lemeshow S, Tepes D, Avrunin JS, Pastiden IT. Predictingthe outcome of intensive care unit patients. J Am Stat Ass1400 men in the Northern Region of England. Br J Urol

1994; 74: 559–65 1988; 83: 348–5630 Royston P, Altman GD. Regression using fractional poly-13 Koshiba K, Egawa S, Ohori M, Uchida T, Yokoyama E,

Shoji K. Does transurethral resection of the prostate pose nomials. Appl Statistics 1994; 43: 429–6731 Robertson C, Boyle P, Hsieh C-C, Macfarlane GJ,a risk to life? 22-year outcome. J Urol 1995; 153: 1506–9

14 Seagroatt V. Mortality after prostatectomy: selection and Maisonneuve P. Some statistical considerations in theanalysis of case-control studies when exposure variablessurgical approach. Lancet 1995; 346: 1521–4are continuous measurements. Epidemiology 1994; 5:15 Crowley AR, Horowitz M, Chan E, Macchia164–70RJ. Transurethral resection of the prostate versus open

32 Guess HA. Benign prostatic hyperplasia: antecedents andprostatectomy: long-term mortality comparison. J Urolnatural history. Epidemiol Rev 1992; 14: 131–531995; 153: 695–7

34 Anson K, Nawrocki J, Buckley J et al. A multicenter,16 Hargreave TB, Heynes CF, Kendrick SW, Whyte B, Clarkerandomized, prospective study of endoscopic laser ablationJA. Mortality after transurethral and open prostatectomyversus transurethral resection of the prostate. Urologyin Scotland. Br J Urol 1996; 77: 547–531995; 46: 305–1017 Hammarsten J, Andersson S, Holmen A, Hogstedt B, Peeker

35 National Health and Medical Research Council. ClinicalR. Does transurethral resection of a clinically benignPractice Guidelines. The Management of Uncomplicatedprostate gland increase the risk of developing clinicalLower Urinary Tract Symptoms in Men. December 1996.prostate cancer? A 10-year follow-up study. Cancer 1994;Canberra: Australian Government Publishing Service, 199774: 2347–451

33 Cowles RS, Kabalin JN, Childs S et al. A prospective18 Semmens JB, Lawrence-Brown MMD, Fletcher DR, Rouserandomized comparison of transurethral resection to visualIL, Holman CDJ. The Quality of Surgical Care Project: alaser ablation of the prostate for the treatment of benignmodel to evaluate surgical outcomes in Western Australiaprostatic hyperplasia. Urology 1995; 46: 155–60using population-based record linkage. Aust NZ J Surg

1998; in press19 National Coding Centre. Australian Version of The

AuthorsInternational Classification of Diseases, 9th Revision, ClinicalC.D.J. Holman, PhD, Professor.Modification (ICD-9-CM). Sydney: National Coding Centre,Z.S. Wisniewski, FRACS, Consultant Urologist.1995J.B. Semmens, PhD, Research Fellow.20 World Health Organization. International Classification ofI.L. Rouse, PhD, General Manager.Procedures in Medicine. Geneva: World HealthA.J. Bass, PhD, Senior Research Fellow.Organization, 1978Correspondence: Dr C.D.J. Holman, Centre for Health Services21 World Health Organization. Manual of the InternationalResearch, Department of Public Health, The University ofStatistical Classification of Diseases, Injuries and Causes ofWestern Australia, East Perth, Western Australia.Death. Geneva: World Health Organization, 1977

22 Norusis MJ. SPSS for Windows. Base System User’s GuideRelease 5.0. Chicago: SPSS Inc., 1992 Appendix

23 Therneau T, Sicks J, Bergstralh E, OCord J. Expectedsurvival based on hazard rates. Rochester, Minnesota: May The 17 diagnostic categories used in the Charlson indexClinic Section of Biostatistics Technical Report #52, March of comorbidity are: myocardial infarction; congestive1994 cardiac failure; peripheral vascular disease; cerebrovas-

24 Hakulinen T. On long term relative survival rates. J Chron cular disease; dementia; chronic pulmonary disease;Dis 1977; 30: 431–43 rheumatological disease; peptic ulcer disease; mild liver

25 Hakulinen T, Abeywickrama KH. A computer program disease; diabetes (mild to moderate); diabetes withpackage for relative survival analysis. Computer Programs

chronic complications; hemiplegia or paraplegia; renalin Biomedicine 1985; 19: 197–207

disease; any malignancy, including lymphoma and leu-26 Charlson ME, Pompei P, Ales KL, McKenzie CR. A newkaemia; moderate or severe liver disease; metastatic solidmethod of classifying prognostic comorbidity in longitudinaltumour; and acquired immunodeficiency syndrome. Thestudies: development and validation. J Chron Dis 1987;present comorbidity summarisation score included all of40: 373–83these categories plus other ischaemic heart disease,27 Romano PS, Roos LL, Jollis JG. Adapting a clinical

comorbidity index for use with ICD-9-CM administrative cardiac dysrhythmias, hypertensive disease and obesity.

© 1999 BJU International 84, 37–42