Delay and survival in bladder cancer

Authors


D.M.A. Wallace, The Department of Urology, The Queen Elizabeth Hospital, Edgbaston, Birmingham, B15 2TH, UK.
e-mail: dmawallace@aol.com

Abstract

Objective To assess in detail and evaluate the effect on survival of delays in the diagnosis and treatment of cancer (which might lead to a worse prognosis), dividing the delay from onset of symptoms to first treatment into several components, comprising patient delay, general practitioner (GP) delay, and two or more periods of hospital delay.

Patients and methods Data were prospectively collected on 1537 new cases of urothelial cancer in the West Midlands from 1 January 1991 to 30 June 1992. Death information was obtained from the West Midlands Cancer Intelligence Unit and censored at 31 July 2000. The influence of delay times on survival was explored.

Results The median delay from onset of symptoms to GP referral was 14 days (Delay 1), from GP referral to first hospital attendance was 28 days (Delay 2), and from first hospital attendance to first transurethral resection of bladder tumour was 20 days (Delay 3). The median hospital delay (Delay 2 + 3) was 68 days and the median total delay (Delay 1 + 2 + 3) was 110 days. Patients with a shorter Delay 1 had a lower tumour stage and a 5% better 5-year survival. Patients with a shorter hospital delay had worse survival; total delay had no effect on survival.

Conclusions There was significantly better survival for patients referred to hospital within 14 days of the onset of symptoms. The relationship between delay and survival in bladder cancer is complex. Hospital delays may be influenced more by comorbidity than by the characteristics of the tumour. However, the adverse effects of delay seem to be most pronounced for patients with pT1 tumours.

Introduction

The clinical stage of the primary tumour is one of the most important determinants of survival in bladder cancer. The depth of tumour invasion into the bladder wall, the key component of stage, is clearly time-dependent. The delays encountered in the process of diagnosis and treatment of bladder cancer may lead to a worse prognosis, but this cannot be studied prospectively. Large studies of other malignancies provide conflicting evidence of the effect of delay on survival [1–6]. Bladder cancer is the fifth most common cancer in England and Wales, with 12 000 new cases and over 5000 deaths per year [7]. Death rates for muscle invasive-bladder cancer have not altered significantly over the last 30 years [8–10], and waiting times for the diagnosis and treatment of bladder cancer in England are poor compared with those for other malignancies [11,12].

The total delay from first onset of symptoms to first treatment can be divided into several components, comprising patient delay, GP delay and two or more periods of hospital delay. In the last 35 years several authors have studied these delays in detail for bladder cancer (Table 1) [11–18]. As can be seen from Table 1, patients usually present to their GP soon after the onset of symptoms (most within 1 week) and are rapidly referred to hospital [14–16,19]. However, there remains a significant proportion of patients who wait >4 weeks before consulting their GP (15%), and a significant proportion who have to wait for >4 weeks before their GPs refer them to hospital (12.5%) [12,18].

Table 1.  Median delays (in days) for patients with bladder cancer
YearRef.No. of
patients
Delay 1Delay 2Delay 3Total
delay
  1. Delay from: aonset of symptoms to first consultation with a GP; bGP consultation to hospital referral; cfirst outpatient clinic appointment to diagnostic cystoscopy; ddiagnostic cystoscopy to definitive treatment; eurgent referrals; fnon-urgent referrals; gDelay from first referral to first definitive treatment (Delay 2 + 3) in urgent referrals; hDelay from first referral to first definitive treatment (Delay 2 + 3) in non-urgent referrals.

1983[15]21272121105
1988[16]1007a/14b1828126
1991[17]5741426 48
1996[18]1862435c/55d
1996[18]1993121c/44d
2000[11]62720e/33f57g/82h

It is clear that hospital delay is the most significant cause of the total delay in the diagnosis and definitive treatment of patients with bladder cancer [11,12,14–18]. However, it is not clear how the individual components of the delays to treatment affect prognosis. It is important that there is a clearer idea of the effect of delay on survival, to develop new strategies to improve survival. The data presented here were collected prospectively to assess the effect of delay on clinical stage and outcome in bladder cancer, and was part of a surveillance study of occupation and urothelial cancer [20].

Patients and methods

In 1990 an Occupational Urothelial Tumour Unit was established in the West Midlands region. All consultant urologists, general surgeons and radiotherapists managing patients with urological diseases within the West Midlands were contacted and informed about the surveillance study. We attempted to collect data prospectively on all newly diagnosed cases of urothelial cancer in the West Midlands from 1 January 1991 to 30 June 1992. The completeness of these data was optimized by collecting from three sources: urologists (including general surgeons practising some urology), the West Midlands Cancer Intelligence Unit, and radiotherapy departments.

For each new case diagnosed in this period the attending clinician was asked to supply data on a specially designed form. The data collected included the dates of onset of symptoms, first referral by the GP, first hospital appointment and first definitive treatment. The date of first definitive treatment was the date of the diagnostic TURBT. For patients with muscle-invasive bladder cancer, definitive treatment may have comprised cystectomy, radiotherapy, chemotherapy or a combination of these methods after the diagnostic TURBT. The TURBT was considered the first definitive treatment because this could be applied uniformly to all bladder cancers. Clinical details collected included the presence or absence of haematuria (macroscopic or microscopic), the number, size and type of tumours, and the findings of the bimanual examination. Pathological details and UICC TNM classification [21] were also collected. All forms were checked to ensure that TNM classification correlated with bimanual examination findings and tumour pathology. Where discrepancies occurred the consultant in charge of the patient's care was requested to clarify the details. Details of patient characteristics were also collected. In addition, patients were asked to complete a questionnaire on their smoking and occupational history. Data were returned to the Occupational Urothelial Tumour Unit and entered onto a computer database. All patients were notified to the West Midlands Cancer Intelligence Unit, who provided death information.

Delay times were calculated as follows: date of onset of symptoms to date of first referral by a GP=Delay 1; date of GP referral to date of first attendance at hospital=Delay 2; date of first hospital attendance to date of first treatment by TURBT=Delay 3; date of GP referral to date of first treatment=hospital delay (Delay 2 + 3); date of onset of symptoms to date of first treatment=total delay (Delay 1 + 2 + 3).

Statistical methods

Associations between the patient characteristics and median delay times were analysed using Pearson's chi-squared test for categorical data and the Mann–Whitney U-test for continuous data. Survival was calculated from the date of first treatment to the date of death or the censor date of 31 July 2000, using all-cause mortality. Survival curves were constructed using the method of Kaplan and Meier and the log-rank test was used to assess differences between groups [22]. Survival was compared in terms of demographic and tumour characteristics and delay times. A stratified survival analysis was used to test for differences within delay times, adjusting for tumour stage, and to test for smoking status, adjusting for delay times. Stepwise Cox-proportional hazards models [23] were applied to investigate the independent effect of age, sex, smoking status, haematuria, tumour stage, grade, type, size and number of tumours. This formed a base model which was used to adjust the effects of each delay. Hazard ratios with 95% CI and P values are presented.

Results

Data were collected for 1537 patients; the overall patient and tumour characteristics are shown in Table 2. The men had a median (interquartile range, IQR) age of 69 (62–76) years and women 71 (64–78) years. The initial treatment plan was recorded for 1451 patients (94%). Of these patients 71% were treated by TURBT and follow-up cystoscopy alone, 2% by TURBT and a course of intravesical chemotherapy, 22% by radiotherapy, 2% by cystectomy, 1% by chemotherapy, and 2% by other treatments (including palliation).

Table 2.  Overall patient and tumour characteristics (where recorded)
VariableNumber %
Sex (1537 responses, 100%)
 Male1146 (75)
 Female391 (25)
Haematuria at presentation (1188 responses, 77%)
 Macroscopic1032 (87)
 Microscopic70 (6)
 None86 (7)
Age, years (1537 responses, 100%)
 <60322 (21)
 61–70505 (33)
 71–80513 (33)
 >80197 (13)
Smoking status (1306 responses, 85%)
 Current or previous smoker1010 (77)
 Never smoked296 (23)
Occupational exposure (1287 responses, 84%)
 Known or suspected increased relative risk348 (27)
 No increased relative risk939 (73)
Tumour type (1460 responses, 95%)
 Papillary942 (65)
 Solid251 (17)
 Mixed267 (18)
Tumour number (1444 responses, 94%)
 Single1078 (75)
 2 or more366 (25)
Tumour size, cm (1421 responses, 92%)
leqslant R: less-than-or-eq, slant2575 (41)
 >2846 (59)
Tumour stage (1340 responses, 87%)
 pTa678 (51)
 pT1300 (22)
 T2–T4362 (27)
Grade (1391 responses, 91%)
 Well489 (35)
 Moderate534 (38)
 Poor and anaplastic368 (27)

In all, 785 patients (51%) had died at the censor date, of whom 720 (92%) had a known cause of death; 350 patients died from bladder cancer (49% of deaths), 135 from other cancers (19%), and there were 235 deaths (32%) from other causes. In 633 patients both stage and cause of death were known. There was a significant association between certified cause of death and tumour stage, indicating that patients with T2–4 tumours were more likely to die from bladder cancer than patients with pTa and pT1 tumours (Ptrend <0.001) (Table 3).

Table 3.  Certified causes of death by tumour stage in the 633 patients where both tumour stage and cause of death were known
CausepTa (%)pT1 (%)T2–4 (%)Total
Bladder cancer49 (21)68 (49)197 (75)314
Other cancer100 (43)40 (29)44 (17)184
Other causes83 (36)31 (22)21 (8)135
Total232139262633

Tumour stage was the most influential prognostic factor (P<0.001), patients with higher stage tumours having significantly worse survival (Fig. 1a). Patients with high grade tumours had poorer survival, as had patients with tumours of >2 cm (both P<0.001). There were no survival differences between number of tumours (single vs multiple) or between different degrees of haematuria (none, microscopic, macroscopic) (P=0.13 and 0.39, respectively). Survival by smoking status showed a borderline difference, suggesting that patients who were current or ex-smokers had poorer survival than those who had never smoked (P=0.04); this effect remained borderline when adjusting for stage of disease and for each of the delay times (P=0.08 and 0.04, respectively).

Figure 1.

Survival by: a, tumour stage (pTa, green solid line; pT1, red dashed line; T1–4, light green dotted line); b, Delay 1 (leqslant R: less-than-or-eq, slant14 days, green solid line; >14 days, red dashed line); and c, hospital delay (leqslant R: less-than-or-eq, slant68 days, green solid line; >68 days, red dashed line).

The median (IQR) Delay 1 was 14 (0–61) days (Fig. 2a). Patients with a longer delay were more likely to present with a higher stage tumour (P=0.04). Patients with an unknown haematuria status were more likely to have a shorter delay (P<0.001). No other patient or tumour characteristics showed a significant difference above or below the median delay (Table 4). Delay 1 had a significant effect on survival (Fig. 1b); patients with a delay of <14 days to referral had an improved survival of 5% at 5-years compared with those who had a delay of >14 days (P=0.02). Adjusting for tumour stage, there was a trend for patients with a shorter Delay 1 to have a better survival (P=0.06) (Table 5).

Figure 2.

The distribution of delay times for: a, Delay 1; b, Delay 2; c, Delay 3; and d, total delay. The median (IQR) delays are given in the text.

Table 4.  Patient characteristics by delay times (in days), n (%)
FactorGroupingNDelay 1Delay 2Delay 3Total delay
leqslant R: less-than-or-eq, slant14>14leqslant R: less-than-or-eq, slant28>28leqslant R: less-than-or-eq, slant20>20leqslant R: less-than-or-eq, slant110>110
  • *

    P<0.05.

Median age, years   70 69 70 69 69 70 70* 69
IQR   62–77 61–76 62–77 62–76 62–76 62–76 63–77 61–76
SexMale1146585 (77)534 (72)565 (75)557 (74)553 (73)569 (76)583 (77)*546 (72)
 Female391178 (23)203 (28)187 (25)195 (26)203 (27)179 (24)174 (23)208 (28)
Tumour stagepTa678357 (54)*317 (48)348 (52)326 (50)330 (51)344 (50)329 (50)346 (52)
 pT1300145 (22)153 (23)142 (21)157 (24)139 (22)160 (24)149 (22)149 (23)
 T2–T4362163 (24)187 (29)180 (27)172 (26)176 (27)176 (26)186 (28)169 (25)
Tumour size, cmleqslant R: less-than-or-eq, slant2575305 (43)264 (38)279 (40)292 (41)266 (38)*305 (43)271 (39)299 (42)
 >2846412 (57)426 (62)422 (60)418 (59)440 (62)400 (57)431 (61)410 (58)
HaematuriaMacro1017392 (51)*625 (85)610 (81)*410 (55)366 (48)*654 (88)525 (70)499 (66)
 Micro6933 (4)36 (5)30 (4)39 (5)21 (3)48 (6)31 (4)38 (5)
 None8534 (5)51 (7)50 (7)36 (5)45 (6)41 (5)39 (5)46 (6)
 Unknown329304 (40)25 (3)62 (8)267 (35)324 (43)5 (1)162 (21)171 (23)
SmokingNever289140 (21)149 (24)145 (23)144 (22)147 (23)142 (22)130 (20)*160 (25)
 Ever993512 (79)481 (76)496 (77)499 (78)491 (77)504 (78)518 (80)482 (75)
Table 5.  Survival by delay times stratified for tumour category
Delay N,
days
 NDead (N)% aliveO/EMedian [95% CI]
survival, years
% surviving at years
135
  1. – indicates limit not reached.

Delay 1 (P=0.02):
 leqslant R: less-than-or-eq, slant1476336052.80.929.48 [8.0, –]867264
 >1473739945.81.097.21 [6.0, 8.5]836859
Delay 1 by tumour stage (P=0.06):
pTaleqslant R: less-than-or-eq, slant1435711966.70.55978780
 >1431712959.30.67968676
pT1leqslant R: less-than-or-eq, slant141456952.40.89–[6.3, –]907766
 >141538246.41.047.26 [5.4, –]897159
T2–T4leqslant R: less-than-or-eq, slant1416312523.32.451.37 [1.1, 2.1]603427
 >1418715019.82.671.33 [1.0, 1.7]593326
Delay 2 (P=0.001):
 leqslant R: less-than-or-eq, slant2875241544.81.126.49 [5.6, 8.0]836657
 >2875234853.70.899.48 [8.5, –]877365
Delay 2 by tumour stage (P=0.001):
PTaleqslant R: less-than-or-eq, slant2834814857.50.70–[9.2, –]968575
 >2832610069.30.50968981
pT1leqslant R: less-than-or-eq, slant281427447.91.026.96 [5.0, –]877058
 >281577850.30.928.20 [6.4, –]907666
T2–T4leqslant R: less-than-or-eq, slant2818014817.82.911.15 [0.9, 1.6]553122
 >2817212925.02.231.61 [1.3, 2.3]653731
Delay 3 (P=0.47):
 leqslant R: less-than-or-eq, slant2075638049.71.038.07 [6.9, –]836859
 >2074838348.70.978.45 [6.9, –]877163
Delay 3 by tumour stage (P=0.29):
pTaleqslant R: less-than-or-eq, slant2033012462.40.64–[9.2, –]958575
 >2034412464.00.57988980
pT1leqslant R: less-than-or-eq, slant201396255.40.83–[7.0, –]907765
 >201609043.81.096.6 [5.4, –]887061
T2–T4leqslant R: less-than-or-eq, slant2017614219.32.931.1 [0.9, 1.3]552922
 >2017613523.32.251.9 [1.4, 2.4]653831
Hospital delay (P=0.001):
 leqslant R: less-than-or-eq, slant6875741245.51.126.59 [5.7, 8.1]826657
 >6874735153.00.899.48 [8.4, –]887365
Hospital delay by tumour stage (P=0.01):
pTaleqslant R: less-than-or-eq, slant6832613857.70.71–[9.1, –]958475
 >6834811068.40.51979081
pT1leqslant R: less-than-or-eq, slant681527451.30.94–[5.4, –]887360
 >681477846.91.007.83 [5.8, –]907365
T2–T4leqslant R: less-than-or-eq, slant6818414720.12.841.08 [0.9, 1.4]543023
 >6816813022.62.291.78 [1.3, 2.4]663829
Total delay (P=0.17):
 leqslant R: less-than-or-eq, slant11075739348.11.057.83 [6.3, –]826759
 >11075437250.70.958.62 [7.4, –]877263
Total delay by tumour stage (P=0.43):
pTaleqslant R: less-than-or-eq, slant11032912661.70.63968677
 >11034612264.70.57978879
pT1leqslant R: less-than-or-eq, slant1101496953.70.86–[6.3, –]887466
 >1101498245.01.077.02 [5.4, –]917360
T2–T4leqslant R: less-than-or-eq, slant11018614919.92.821.12 [0.9, 1.4]543124
 >11016913023.12.311.69 [1.3, 2.3]663729

The median Delay 2 was 28 (7–61) days (Fig. 2b). Patients known to have had macroscopic haematuria (n=1032) were more likely to have a shorter delay than those known to have had microscopic haematuria (n=70); patients with an unknown haematuria status were more likely to have a longer delay (P<0.001). There were no other significant differences in patient or tumour characteristics above or below the median delay (Table 4). Patients who had a shorter Delay 2 had a significantly worse survival (P=0.001; Table 5). Survival by Delay 2 after adjusting for tumour stage similarly showed that patients with a shorter Delay 2 had significantly worse survival (P=0.001; Table 5).

The median Delay 3 was 20 (0–50) days (Fig. 2c). Patients with a shorter Delay 3 were more likely to have larger tumours (>2 cm; P=0.04); patients with a longer Delay 3 were more likely to have macroscopic haematuria (P<0.001). Patients with an unknown haematuria status were more likely to have a shorter delay (P<0.001). No other patient or tumour characteristics differed significantly above or below the median delay (Table 4). Delay 3 had no effect on survival (P=0.47); this was also true after adjusting for tumour stage (P=0.29; Table 5).

The median hospital delay was 68 (34–118) days. Patients with a shorter hospital delay had a significantly worse survival, both overall (P=0.001; Fig. 1c) and adjusting for tumour stage (P=0.01; Table 5).

The median total delay was 110 (62–209) days (Fig. 2d). Longer delays were significantly associated with women (P=0.05), younger patients (P=0.03), non-smokers (P=0.04) and patients with a low risk of occupational exposure (P=0.04). No other patient or tumour characteristics showed significant differences above or below the median delay (Table 4). The total delay had no effect on survival (P=0.17); this was also true after adjusting for tumour stage (P=0.43; Table 5).

For prognostic factors, there was no survival difference for sex (P=0.92), haematuria (P=0.39) and number of tumours (P=0.13), both in the log-rank analysis and Cox regression models. Cox regression models formed a base model consisting of age, tumour type, stage, smoking status, grade and tumour size (all inline imageP<0.001). Adjusting each delay by the base model showed no change between the unadjusted and adjusted values (Tables 5 and 6), apart from Delay 3 where there was a borderline difference suggesting that patients with shorter delays had a poorer survival.

Table 6.  Cox regression of delay times adjusted by a base model of independent factors
FactorGroupingCoefficientχ2 to remove/enterPHazard ratio
(95% CI)
A: Delay 1 adjusted by base model (n=1088)
Delay 1(leqslant R: less-than-or-eq, slant14, >14 days)0.17724.060.041.19 (1.01, 1.42)
AgeContinuous0.0540118.12<0.0011.06 (1.05, 1.07)
Tumour type(papillary, mixed, solid)0.260112.45<0.0011.30 (1.12, 1.50)
Tumour stage(pTa, pT1, T2–T4)0.260910.160.0011.30 (1.11, 1.52)
Smoking(never, ever)0.347910.500.0011.42 (1.15, 1.75)
Grade(well, moderate, poor)0.25029.790.0021.28 (1.10, 1.50)
Tumour size (cm)(leqslant R: less-than-or-eq, slant1, 1–leqslant R: less-than-or-eq, slant2, 2–leqslant R: less-than-or-eq, slant3, 3–leqslant R: less-than-or-eq, slant4, >4)0.10147.920.0051.11 (1.03, 1.19)
B: Delay 2 adjusted by base model (n=1090)
Delay 2(leqslant R: less-than-or-eq, slant28, >28 days)− 0.295011.17<0.0010.75 (0.63, 0.89)
AgeContinuous0.0532115.43<0.0011.06 (1.04, 1.07)
Tumour type(papillary, mixed, solid)0.262212.67<0.0011.30 (1.13, 1.50)
Tumour stage(pTa, pT1, T2–T4)0.287612.43<0.0011.33 (1.14, 1.56)
Smoking(never, ever)0.349310.570.0011.42 (1.15, 1.75)
Grade(well, moderate, poor)0.24139.250.0021.27 (1.09, 1.49)
Tumour size (cm)(leqslant R: less-than-or-eq, slant1, 1–leqslant R: less-than-or-eq, slant2, 2–leqslant R: less-than-or-eq, slant3, 3–leqslant R: less-than-or-eq, slant4, >4)0.10158.010.0051.11 (1.03, 1.19)
C: Delay 3 adjusted by base model (n=1090)
Delay 3(leqslant R: less-than-or-eq, slant20, >20 days)− 0.18384.340.040.83 (0.70, 0.99)
AgeContinuous0.0534116.14<0.0011.06 (1.05, 1.07)
Tumour type(papillary, mixed, solid)0.271213.65<0.0011.31 (1.14, 1.51)
Tumour stage(pTa, pT1, T2–T4)0.269311.09<0.0011.31 (1.12, 1.53)
Smoking(never, ever)0.350110.660.0011.42 (1.15, 1.75)
Grade(well, moderate, poor)0.260010.730.0011.30 (1.11, 1.52)
Tumour size (cm)(leqslant R: less-than-or-eq, slant1, 1–leqslant R: less-than-or-eq, slant2, 2–leqslant R: less-than-or-eq, slant3, 3–leqslant R: less-than-or-eq, slant4, >4)0.09496.970.0081.10 (1.03, 1.18)
D: Total delay adjusted by base model (n=1090)
Total delay(leqslant R: less-than-or-eq, slant110, >110 days)− 0.03140.130.720.97 (0.82, 1.15)
AgeContinuous0.0526113.30<0.0011.05 (1.04, 1.06)
Tumour type(papillary, mixed, solid)0.271313.61<0.0011.31 (1.14, 1.52)
Tumour stage(pTa, pT1, T2–T4)0.261410.290.0011.30 (1.11, 1.52)
Smoking(never, ever)0.33679.870.0021.40 (1.14, 1.73)
Grade(well, moderate, poor)0.24619.520.0021.28 (1.09, 1.50)
Tumour size (cm)(leqslant R: less-than-or-eq, slant1, 1–leqslant R: less-than-or-eq, slant2, 2–leqslant R: less-than-or-eq, slant3, 3–leqslant R: less-than-or-eq, slant4, >4)0.00956.970.0081.10 (1.03, 1.18)

Discussion

When this study was conducted a close liaison was established with the West Midlands Cancer Registry to ensure that we were notified promptly of all new registrations. The number of patients accrued in this study was ≈85% of the total registrations of bladder cancer for this period for the West Midlands. Data might not have been collected from patients if they were treated privately, too ill to complete the questionnaire, or if they were diagnosed by clinicians other than urologists. These data represent the largest prospective study to date of delay and survival in bladder cancer. However, it is incomplete in parts. Tumour data and simple patient data (age, sex) were easily obtained from the Cancer Registry and are complete, but clinically based data (presence or absence and degree of haematuria) and more detailed epidemiological data (smoking status, risk of occupational exposure) relied upon clinicians and patients to complete questionnaires and some data are incomplete. Because of the time elapsed since the study commenced (10 years) and the many centres involved (some of which have subsequently undergone significant reorganisation) we did not attempt to obtain the missing data.

In comparison with the BAUS audit data of 5839 newly presenting bladder cancers (BAUS Section of Oncology, unpublished data, 2000) the present series shows a higher proportion of patients with pTa tumours (51% vs 34%) with a resulting lower proportion of pT1 tumours (22% vs 32%) and T2–T4 tumours (27% vs 33%). Gulliford et al.[16] had proportions of patients with stages pTa, pT1 and T2–T4 tumours of 35%, 20% and 21%, respectively, although the stage was unclassified in 24%. In addition, the present series has a male : female ratio of 3 : 1 and a similar proportion of current or previous smokers (77%) as in other large studies [24]. The age and sex distribution of the present population are similar to the national figures reported for England and Wales [7]. We therefore consider the present patients to be representative of the bladder cancer population of the whole UK.

Haematuria status was recorded for 1188 patients, representing 77% of the total. At presentation, 87% of these patients had macroscopic haematuria, 6% had microscopic haematuria, and 7% had no haematuria. This is in agreement with the commonly cited value that 80–90% of patients with bladder cancer present with macroscopic haematuria and is similar to the findings of Payne [13] (90%) and Wallace and Harris (91%) [14]. Interestingly, the percentages of micro- and macroscopic haematuria have remained unchanged since those studies were carried out in the 1950s and 1960s, despite the subsequent widespread use of urine dipstick testing. This highlights the importance of macroscopic haematuria as the dominant presenting symptom of bladder cancer, which usually has a very clearly defined onset and can be easily recalled. However, the presence or absence of haematuria and the degree of haematuria (micro- or macroscopic) had no effect on survival. Other symptoms in patients with bladder cancer include ‘cystitis-like’ symptoms in the absence of a UTI, which have a less well-defined onset and may be responsible for delays in referral [14–17]. Mommsen et al.[15] highlighted these patients as a high-risk group and recommended that patients aged >50 years who present with symptoms of cystitis with negative urine culture should have their urine examined cytologically.

There was an unclear relationship between smoking and survival. Overall, current and ex-smokers had a poorer survival than those who had never smoked, but this difference was only just significant (P=0.04). However, within stage there was no clear trend and the results were not significant. Similarly for Delay 1, current or ex-smokers had a worse survival in each delay category (above or below the median Delay 1), but again these results were only just significant (P=0.04). Despite this unclear relationship between smoking and survival from bladder cancer in these patients, smoking is the single most important cause of urothelial cancer [24].

Sorahan et al.[20] showed that an important proportion of patients presenting with urothelial tumours are likely to have had occupational exposure to urothelial carcinogens, but that this does not affect their prognosis. In the present series the 27% of patients who had worked in an occupation with a known or suspected increased relative risk of urothelial malignancy had no difference in survival from the remaining patients (P=0.99).

In the present series there was a large proportion of patients with noninvasive papillary tumours (pTa) of grade 1 and 2; these patients have a 5-year survival similar to the normal population and present with macroscopic haematuria at a slightly higher frequency than those with muscle-invasive tumours (T2–T4). In patients with pTa tumours of grades 1 and 2, bladder cancer is not likely to affect the 5-year survival and this group can act as an internal control to assess how comorbidity might be affecting delay and survival. The presenting symptoms, diagnosis and initial management of these superficial tumours by TURBT are the same as for invasive cancers, and the risk of progression to muscle-invasion is low (4–20%) [25,26].

Some of the studies reported previously (Table 1) analysed the effect of delay on survival [13–18]. Payne [13] showed that patients with a 0–2 month symptom history (haematuria in 90%) had the best crude 3-year survival rate (57%), followed by those with a history of geqslant R: gt-or-equal, slanted2 years (48%) [13]. Interestingly, the groups faring worst had symptom histories of 3–8 months (37%) and 9–23 months (44%) [13]. Similarly, Wallace and Harris showed that for patients with muscle-invasive bladder cancers treatment within 1 month after the onset of symptoms resulted in a better survival rate than when treatment was delayed by 1–6 months (60% vs 25% 3-year crude survival rate) [14]. With yet further delay beyond 6 months the prognosis improved [14]. However, for superficial bladder cancers, delay in initiating treatment had little effect on the 3-year survival rate [14]. It was concluded that the biological potential of a growth was more important in determining prognosis than delay [14]. Mommsen et al.[15] showed no significant influence of delay on survival from bladder cancer in Denmark, although in patients with T1 and T2 tumours there was a tendency towards longer survival with shorter total delay, but the trend was not significant. Gulliford et al.[17] reported a trend towards worsening prognosis with increasing duration of symptoms before referral, but again the association was not significant (P=0.06), although they had incomplete data in a third of their patients. However, patients with the shortest hospital delay and shortest total delay had a worse prognosis than those treated after longer delays, with further analysis suggesting that patients with severe disease were given priority for treatment [17]. There was little evidence to suggest that the prognosis deteriorated with increasing delay [17].

In the present patients the duration of Delay 1 was very similar to that reported by others [14,16], with patients typically presenting early and being referred to hospital rapidly. Hospital delay (Delay 2 + 3) and total delay (Delay 1 + 2 + 3) have remained virtually unchanged over the last 35 years and remain poor, especially when compared with other malignancies [11–18]. This is despite large increases in the number of urologists during this time and the development of rapid diagnostic services for the investigation of haematuria. Delay 1 had a significant effect on tumour stage and survival; a longer Delay 1 resulted in higher stage tumours and worse survival. Within the stage groups those with a longer Delay 1 had a significantly worse survival and this difference was most pronounced for patients with pT1 tumours.

The delay from first hospital referral to first hospital attendance, and the delay from first hospital attendance to first treatment (Delay 2 and Delay 3, respectively) have a complex relationship with survival. Patients with a shorter Delay 2 had a significantly poorer survival even after adjusting for tumour stage (both P=0.001) and independent prognostic factors (P<0.001). Delay 3 had a similar but borderline effect on survival but only after adjusting for the independent prognostic factors, with those patients with a shorter Delay 3 having a worse survival (P=0.04). This difference was not significant when adjusting for tumour stage alone (P=0.27), except for the small subgroup of patients presenting with pT1 tumours who appeared to have significantly better survival with a shorter Delay 3. Although subgroup analysis should be treated with caution, this is an interesting group of patients for urologists. Gulliford et al.[17] also showed that patients with shorter hospital delays had worse survival. When Delays 2 and 3 are taken together to give the hospital delay the findings are the same; patients with a shorter hospital delay, both overall and within stage, had a significantly worse survival. This suggests that other factors, such as comorbidity, were used to select patients for early hospital consultation; comorbidity may then have subsequently determined outcome, and not bladder cancer. Clearly, for Delay 2, urologists have to prioritize patients for consultation solely based upon information provided by the GP; comorbid factors may thus be used to influence this prioritization; Macarthur et al.[12] eluded to this in 1985. After the first hospital consultation urologists have additional clinical information, e.g. the degree of haematuria and the findings at flexible cystoscopy, on which to base decisions about the priority of treatment, although comorbidity may still have an effect on that decision. For example, patients with large solid-looking tumours at flexible cystoscopy may take priority for early TURBT and conversely, patients with small papillary tumours at flexible cystoscopy may have been considered as requiring TURBT less urgently. This may explain why patients with a shorter Delay 3 had significantly larger tumours and a tendency to worse survival. In addition, a rapid change in a patient's symptoms between the initial presentation to the GP and the first hospital consultation may prompt rapid definitive treatment, as may be seen in breast cancer [6].

Interestingly, patients with an unknown haematuria status were more likely to have a shorter Delay 1, longer Delay 2 and shorter Delay 3. Haematuria status had the lowest recording rate (77%) of all the variables assessed (Table 2). The reasons for this and the different delays in this group are not immediately apparent, but there was no significant difference in the total delay.

The first TURBT was defined as the time of first ‘definitive treatment’, as this could be applied uniformly to the whole series. This would be the date of true definitive treatment for all patients with pTa tumours, most patients with pT1 tumours and some patients with T2 tumours; 71% of 1451 patients where the treatment plan was known were treated by TURBT and follow-up cystoscopy alone. For muscle-invasive tumours (T2–T4) the initial TURBT is a staging procedure only and ‘true’ definitive treatment usually requires radiotherapy or cystectomy. The time of cystectomy is easily defined, but for radiotherapy this is more difficult because of the prolonged nature of the treatment. In addition, at the time of this study patients were being recruited into trials of neoadjuvant chemotherapy, with chemotherapy being administered for several weeks before radical cystectomy or radiotherapy [10]. Again, defining the time of ‘true’ definitive treatment is difficult, especially defining a time that is biologically equivalent for widely differing treatment regimens, although current practice for calculating waiting times classifies the date of first definitive treatment as the first day of treatment with radiotherapy or neoadjuvant chemotherapy. We did not attempt to collect this information for the present patients prospectively or retrospectively, as it would be very difficult to extract it after such a long period. However, these are important issues that should be studied prospectively.

Women, younger patients and non-smokers had a significantly longer total delay, as reported by others [12,15], and may reflect a lower index of suspicion of bladder cancer in these groups of patients. The total delay had no significant effect on survival. However, with stage, patients with a shorter total delay had a worse survival, except for patients with pT1 tumours where the reverse was true. Again, comorbidity may have had an effect, although the benefits of early curative treatment for patients with pT1 tumours may have outweighed this effect.

Clearly, patients with pT1 tumours had a significantly poorer outcome than those with pTa tumours (62% vs 78% 5-year survival) and Table 3 shows that the rates of death from bladder cancer, other cancers, and other causes for patients with pT1 tumours were exactly between the rates for patients with pTa tumours and muscle-invasive tumours (T2–T4). With stage the patients with pT1 tumours who had a longer Delay 1 had a significantly worse survival (59% vs 66% 5-year survival). In addition, contrary to the trends described above where patients with a longer Delay 3 and longer total delay tended to have a better outcome, patients with pT1 tumours, a longer Delay 3 and longer total delay had a worse outcome (61% vs 65%, and 60% vs 66% 5-year survival, respectively). This highlights the importance of this group of tumours; they should be targeted for rapid and aggressive treatment before further invasion and the formation of regional metastases. Clinically it is not possible to distinguish noninvasive tumours (pTa) from tumours with stromal invasion (pT1) without histology, as both appear exophytic and have predominantly papillary growth patterns. Both are managed by initial TURBT and additional intravesical therapy is prescribed for those with poor prognostic factors or frequent recurrences. The use of review pathology and repeat cystoscopy and biopsy within 2–6 weeks to re-stage patients with pT1 tumours is currently being advocated, as up to 29% may be upstaged to tumours that require more radical therapy, e.g. cystectomy or radiotherapy [27,28]. However, at the time of this study (1991–92) most urologists in the UK would have managed patients with even high-grade pT1 tumours by TURBT alone [29]. By the time muscle is invaded tumour biology rather than treatment delay may be a more important determinant of survival. Therefore, with early invasive disease (stages pT1 and T2a) hospital delay may have a greater effect on survival. These cases cannot be selected on clinical grounds alone and therefore strategies of healthcare need to be developed to ensure rapid referral and treatment of all patients with symptoms of bladder cancer. In the UK, government guidelines for the urgent referral of patients with suspected cancer came into effect for urology in December 2000, and include all adults with macroscopic haematuria and those with microscopic haematuria aged >50 years. Realistic targets must be set not just for the time to initial treatment by TURBT, but for definitive treatment for all invasive tumours.

In conclusion, survival is significantly better for those patients with bladder cancer referred to hospital by their GPs within 14 days of the onset of symptoms. In addition, hospital delays for bladder cancer have changed little over the last 35 years and may be influenced more by comorbidity than by the characteristics of the tumour, thus showing poorer survival for patients with shorter hospital delays. Despite the worse survival overall for patients seen and treated fastest by hospitals, the adverse effects of delay seem to be most pronounced for patients with pT1 tumours. While strategies for shortening the delay to definitive treatment for all invasive bladder cancers (both pT1 and muscle-invasive) are urgently needed, the analysis of the effects of delay must consider all confounding factors such as comorbidity.

Acknowledgements

We thank all participating clinicians and their staff from within the West Midlands Region and the West Midlands Cancer Intelligence Unit for their support. Professor David Kerr and Professor Michael Richards provided invaluable critical feedback on the manuscript. The work was funded by a grant from the Health and Safety Executive and supported by the CRC Trials Unit and The Institute of Occupational Health, The University of Birmingham. R.T. Bryan is funded by a Medical Research Council Research Training Fellowship.

Ancillary