Time from diagnosis to surgery and prostate cancer survival: a retrospective cohort study

The South West Public Health Observatory (SWPHO) provided us with an anonymised dataset of English cancer registry records linked to inpatient Hospital Episode Statistics (HES) and the Office of the National Statistics (ONS) mortality databases. Prostate cancer was defined as having a tumour classified in the International Classification of Diseases (ICD) version 10 as C61.

From all patients who were registered in the population-based cancer registries, patients diagnosed between January 1, 1996 and December 31, 2009, who were 15 years or older at the time of diagnosis, and who had surgical resection with curative intent within 6 months of diagnosis were included in the study. The age criterion was to allow comparability between our results and recent published survival estimates [11, 12]. The 6-month cut-off was based on clinician perceptions on acceptable time from diagnosis to surgery not influenced by conditions necessitating delay (D. Gillatt, personal communications). Patients diagnosed with secondary cancers, in situ cancers, or diagnosed via death certificates only (DCO) or through autopsy were excluded.

A total of 22,152 men with prostate cancer met these criteria. From these, we excluded men with time from diagnosis to surgery of over 6 months (n = 4,171), and a further 938 patients with negative or zero post-operative survival times (follow-up). After all exclusions, we were left with 17,043 patients in the final sample.

The time from diagnosis to first curative surgery was defined as the time (in months) between the date of cancer diagnosis (as recorded in the registry database) and the date of the first curative resection (as recorded in HES). The date of diagnosis is defined by the cancer registries as the date of the first event or event of higher priority (if recorded within three months of the first event) among the following: histological or cytological confirmation, admission to the hospital or first consultation at the outpatient clinic because of the malignancy, or date of death [13]. In more than 99% of patients, diagnosis was confirmed through histology of the primary tumour. Time between diagnosis and surgery were categorized into 0–3 and 4–6 months.

Curative prostate cancer resections were defined as total / radical prostatectomy (M611), perineal prostatectomy (M614), open excision of prostate (M618), and prostatectomy NEC (M619), based on the Office of Population Censuses and Surveys (OPCS) Classification of Interventions and Procedures [10].

Post-operative survival was defined as the number of days between the date of the first curative resection and the date of outcome (death or censoring). Follow-up was censored at five- and ten- years, or at the end of the study period, which was December 31, 2009.

Other covariables in the study were age, region of residence, ethnicity, tumour differentiation, level of deprivation and period of cancer plan implementation. Age at cancer diagnosis was categorized as 15–54, 55–64, 65 years and above. Geographical region was defined as the patient’s region of residence at the time of diagnosis. Ethnicity was self-reported ethnicity, as recorded in the HES database [14, 15]. This was categorized as White and non-White, and could not be further subdivided due to the small number of cases in ethnic groups other than White. Only ethnicity codes in 2005 to 2009 were used as these were deemed most complete (SWPHO, personal communication) [15], so ethnicity was coded as “unknown” prior to 2005. Analyses looking specifically at the effect of ethnicity on the association of time between diagnosis and surgery with survival were limited to patients diagnosed between 2005 and 2009.

Tumour differentiation refers to cell differentiation at the time of tumour biopsy and was classified as well-, moderately-, poorly- and un-differentiated. The implementation period of the UK Department of Health targets was defined as ‘prior to implementation’ (1996–2000), and ‘after implementation’ (2001–2009). Level of deprivation was derived from the income component of the 2007 Index of Multiple Deprivation (IMD) [16], and was computed for small geographical areas known as Lower Super Output Areas (LSOAs; mean population = 1500 people) [17]. Quintiles based on English IMD scores were computed, with the first quintile designated as the least deprived.

The median time from diagnosis to surgery by each of the covariables were computed. Using univariable and multivariable linear regression, coefficients reflecting the additional days for each category compared to the reference category were determined for each covariable. All covariables were controlled for in the multivariable analysis. For all regression analyses, we used multiple imputation using chained equations (ICE) to account for missing data on grade and deprivation quintile [18, 19]. A total of 20 complete data sets were constructed to reduce sampling variability from the imputation process [20], and the results were combined using Rubin’s rules [18, 19].

Complete estimates of relative survival (where all men diagnosed between 1996 and 2009 were included, regardless of whether they had full five-year or partial follow-up) [21], were computed using the STRS command in STATA, version 12 [22]. Relative survival is a measure of survival, having accounted for underlying mortality rates. It is the ratio of the observed survival of cancer patients to the probability of survival that would have been expected if patients had experienced the same survival probability as the general population [23]. Survival probabilities were estimated at intervals of 6 months in the first year, then yearly up to 10 years. We used age-, region- and calendar year-specific UK life tables for males [24] to account for the differences in the underlying mortality and used the Ederer II method [23] to determine expected survival.

To investigate the effect of time between diagnosis and surgery on survival, the data were stratified according to categories of time to surgery (0–3 and 4–6 months) and relative survival estimates were visually inspected. To account for waiting time paradox (wherein patients offered surgery within few weeks of diagnosis could be presenting more severe manifestations of the disease), a sensitivity analyses was done using more refined time intervals (1, 2, 3–4 and 5–6 months). We limited the analysis to computing relative survival estimates because the excess mortality in the study population is negative (i.e. the mortality in our study population is lower than that of the general population) which causes model convergence problems (P. Dickman and P. Lambert, personal communication).

This project was approved by the Faculty of Medicine and Dentistry Committee for Ethics, University of Bristol (101153), the NHS South Central – Berkshire B Research Ethics Board (11/SC/0387) and the National Information Governance Board (NIGB, ECC 7-02(d)/2011).

We made use of cancer registry data that were provided to us in anonymised form. The use of this data is regulated by the Confidentiality Advisory Group (CAG, formerly NIGB) of the Health Research Authority and does not require individual patient consent.