Faecal occult blood screening for colorectal cancer at Nottingham: details of the verification process

Abstract

Cause specific mortality statistics derived from death certificates are highly dependent upon the accuracy of certification by the attending physician. In the Nottingham colorectal cancer screening trial, there were 12 624 deaths among the screening group and 12 515 among the control group during the period under consideration. There was no significant diVerence in all cause mortality rate (excluding deaths due to colorectal cancer) between the two study groups (rate ratio = 1.01, 95% confidence interval = 0.99 to 1.03). Disease specific mortality rates did not diVer significantly between the two groups either. Overall, the agreement between verified and certified cause of death was 86%. Using the certified cause of death would have resulted in an underestimation bias of 6.27% for colorectal cancer deaths. 2 Cause specific mortality statistics such as these are derived from the underlying cause of death (UCD) on death certificates, and are highly dependent upon the accuracy of certification. This depends upon the level of training and experience of the certifying doctor.3 Cancers are better recorded than other causes of death, but there are diVerences in recording among the various cancer sites. Some, like breast cancer, are better recorded while internal cancers have a higher chance of being missed.4 5 A small degree of inaccuracy may be due to the low necropsy rate. Verification of patient records allows the use of additional information in order to determine the cause of death more accurately. This is important within the setting of a randomised controlled trial where reduction in disease specific mortality is the end point being measured. This report describes the verification process used in the Nottingham trial. MethodsThe study design and methodology have been described in detail elsewhere.1 Briefly, 152 850 individuals aged 45 to 74 were randomly allocated to a screening group (invited to screening using Haemoccult FOB test kit every two years) or control group (no intervention). All deaths occurring from the date of entry into the study up to 30 June 1995 were analysed. The UCDs were coded from death certificates according to the International Classification of Diseases.6 Case notes were obtained for all deaths among study participants diagnosed with CRC, and those who had CRC or carcinomatosis (without specification of the primary tumour) on the death certificate. The initial verifier used a structured review form to classify deaths as definite, probable, possible, and unlikely CRC deaths; cause of death unknown/unverified; and deaths definitely not due to CRC. All cases in the probable, possible, and unlikely categories underwent a second review; a third reviewer coded any cases of disagreement. Causes of death were categorised as CRC (definite or probable CRC deaths) or "other causes" (deaths not due to CRC).Assuming the assessors' verdict to be correct, we calculated the proportions of deaths that had changed categories (from CRC to "other causes" and from "other causes" to CRC) as a result ...

Does the source of death information affect cancer screening efficacy results? A study of the use of mortality review versus death certificates in four randomized trials

“…To our knowledge, only two previous analyses have examined the impact of mortality review on the results of randomized trials of cancer screening [2,4]. The first considered data from the four randomized trials of mammography screening conducted in Sweden [4].…”

“…Also similar to our findings, in all but one instance in the Swedish trials mortality review tended to result in a larger estimate of the benefit of screening, and further there was one instance where the statistical significance of the results differed based on whether mortality review or death certificates were used to define endpoint. The second analysis considered data from the Nottingham trial of FOBT [2]. Death review in this trial was conducted for all deaths among known cases of colorectal cancer, as well as for those deaths for which colorectal cancer or carcinomatosis (with primary unspecified) was noted on the death certificate.…”

“…However, mortality review is time-consuming and expensive, and little is known regarding how findings from mortality review compare with those from death certificates alone. While investigators from several trials of colorectal [1,2], prostate [3], and breast cancer screening [4] have examined the accuracy of death certificates in reporting cancer-specific deaths, to our knowledge only two prior analyses have examined the actual impact of death certificate errors on trial results [2,4]. These analyses, from the four randomized trials of mammography screening conducted in Sweden [4] and from the Nottingham trial of fecal occult blood testing (FOBT) [2], reported similar estimates of screening efficacy when either mortality review data or death certificate data were used.…”

Nottingham trial of faecal occult blood testing for colorectal cancer: a 20-year follow-up

“…CRC mortality rates were calculated from the underlying cause as stated on the death certificate and the verified cause of death after case note review 6. The verification process was blind to whether the patient was in the control or intervention arm, although in around 5% of cases it was apparent from the notes that the patient had been screened.…”

Analysis of deaths occurring within the Nottingham trial of faecal occult blood screening for colorectal cancer

“…UCD as certified by the attending doctor was the basis of mortality classification, with one modification. At an earlier stage of the trial, independent assessors had reviewed case notes for all deaths where CRC or an unspecified primary tumour appeared to be a cause 11. Disagreements between certified causes and those verified by the assessors were evident, and it was concluded that certification under-reported certain or probable deaths from CRC.…”