A high level of case ascertainment by cancer registries is essential to allow estimation of accurate incidence rates and survival. Nearly 20 years ago, researchers assessed the completeness and accuracy of registration of primary intracranial tumours (Scottish Cancer Registry [SCR]) compared to a database assembled in the context of a detailed incidence study carried out in the Lothian region of Scotland covering the period of diagnosis, 1989–1990.1 and 2 Disappointingly, SCR identified only 54% of cases, although the registry at that time did not attempt to collect information on ‘benign’ intracranial neoplasms which were included in the detailed incidence study. Even so, only 84% of neuro-epithelial tumours were identified by SCR, probably related in part to the fact that the cancer registry was not receiving neuropathology data from the regional neuro-oncology centre. An English study reported similar findings with only 52% of cases appearing in the regional cancer registry.3
Over time, access to diagnostic techniques has improved alongside improvements and changes in classification and clinical coding. Furthermore, SCR now receives neuropathology data from all neuro-oncology centres in Scotland, and has sought to collect information on benign tumours of the brain and spinal cord since the year of diagnosis, 2000. In light of these developments, we aimed to determine the completeness of ascertainment of primary intracranial tumours by SCR through independent/clinical case ascertainment in NHS Lothian for the period of diagnosis, 2011–2012.
Scottish Cancer Registry
SCR operates by bringing together predominantly electronic information from hospital patient administration systems including patient discharges from hospital (Scottish Morbidity Record 01), radiotherapy, oncology, and pathology records; death records from National Records Scotland; and private hospital records.4 All primary malignant and benign brain tumours are recorded on the SCR.
Inclusion and exclusion criteria
This retrospective cohort study was restricted to the period of diagnosis between 1 January 2011 and 31 December 2012 and limited to adults (age ≥18 years on the date of diagnosis) with a postcode within the NHS Lothian health board region (mid-year population ∼650,000). The date of diagnosis was taken as i) the date of the first abnormal imaging, or ii) the date of biopsy/resection. Patients in whom there was no neuro-radiology or histology were excluded, i.e. diagnosis of prolactinoma had to be supported by blood tests and imaging.
All suspected and histologically proven primary intracranial tumours (benign and malignant) of the brain and cranial nerves were counted, including primary central nervous system lymphoma. Meningeal, pituitary region and pituitary gland tumours were also included. Cerebral metastases, tumours of the spinal cord and spinal nerves, and recurrent intracranial tumours of any type were excluded.
Extraction from the SCR
Data were extracted for the study period for all records including the following anatomical site codes selected from the third edition of the International Classification of Diseases for Oncology (ICD-O-3): C70.0; C70.9; C71; C72.2; C72.3; C72.4; C72.5; C72.8; C72.9; C75.1; C75.2; and C75.3 (all behaviour codes).
Clinical case ascertainment
Three clinical sources were trawled as follows: i) neuro-oncology multidisciplinary team meeting (MDTM) minutes; ii) an endocrinology database; and iii) neuropathology records (sources i and ii are independent of the SCR data collection system). The neuro-oncology MDTM aims to discuss all intracranial tumours identified via any means including both benign and malignant tumours. The endocrinology database records all patients attending the endocrine outpatient clinics in the NHS Lothian region and each is assigned a diagnosis by a Consultant Endocrinologist. Both sources i and ii for the period 1 January 2011 – 31 March 2013 were reviewed manually by JRM. The neuropathology records hold information on all tissue samples analysed in the pathology system for NHS Lothian hospitals and an electronic extract was obtained using Systematized Nomenclature of Medicine codes matching those above. To ensure cases were true incident cases meeting the full inclusion criteria, each was cross-referenced with the patient's electronic secondary care medical record.
Clinically ascertained cases were reconciled against the SCR extract. We have previously quantified the extent of misclassification of incidence year in the Scottish Cancer Registration database.5 We did not regard misclassifications of incidence year as missed or ‘over-diagnosed’ cases as there is no reason to believe that such misclassification is other than random.
Completeness was defined as the proportion of intracranial tumours included in the SCR out of all those identified as diagnosed in residents of NHS Lothian area during the study period. Confidence intervals (95%) for completeness were calculated using the exact method.
Completeness was calculated for all intracranial tumours with secondary analysis of completeness by tumour morphology and tumour grade (see Supplementary File for coding definitions).
There were 320 records of primary intracranial tumours registered on the SCR for the period of interest and 264 clinical cases were ascertained. Fig. 1 shows the final ascertainment of clinical cases missing from the SCR.
Morling, J. R., Grant, R., & Brewster, D. H. (2016). Completeness of primary intracranial tumour recording in the Scottish Cancer Registry 2011-12. Public Health, 140, 278-281. https://doi.org/10.1016/j.puhe.2016.05.024