Skin cancer: how accurate are local data?

BMJ 1995; 310 doi: (Published 25 February 1995) Cite this as: BMJ 1995;310:503
  1. Clive Richards, senior registrara,
  2. Hilary Richards, consultant in medical information technologyb,
  3. Derek Pheby, directorc
  1. a Department of Public Health Medicine, Bristol and District Health Authority, Bristol BS2 8EE
  2. b Clevedon, Avon BS21 7BX
  3. c South Western Regional Cancer Registry, Bristol BS8 2PR
  1. Correspondence to: Dr C Richards, Nottingham Health, Nottingham NG3 5AF.
  • Accepted 25 November 1994

Baseline information on the incidence of skin cancer for monitoring targets set in the Health of the Nation is based on data about malignant melanoma obtained from regional cancer registries.1 Levels of ascertainment of cancer by registries are variable.2 We undertook a retrospective survey to review the registration of skin cancer over 10 years (1984 to 1993). The South Western Regional Cancer Registry's data on skin cancer in residents of the Bristol and District Health Authority were compared with data obtained directly from the histopathology records of the hospital laboratories serving the district.

Subjects, methods, and results

The four NHS hospitals serving the Bristol and District Health Authority area have computerised histopathology systems, which were introduced around 1987. A printout of all cases of skin cancer was obtained from each hospital with a manual search for the years before computerisation. Two private hospitals were also asked to provide data. These were linked with patient identification details to provide the minimum dataset required for registration. The data were then matched against previous entries in the cancer registry.

A total of 365 additional cases of skin cancer treated at local hospitals were identified. These included 177 new registrations of malignant melanomas of the skin (code 172 of the International Classification of Diseases, ninth revision (ICD-9)); 102 related to local residents, giving an increase of 13% on the 790 cases already registered. Revised age standardised incidence rates were calculated for each of the years 1984 to 1993 (table). The private sector generated 23 of the 102 unregistered cases (23%). The remaining 188 additional registrations included diagnoses other than malignant melanoma of the skin (ICD-9 172) such as carcinoma in situ and lentigo maligna (ICD-9 232), dysplastic naevus or melanoma precursor (ICD-9 238), melanoma of the eye (ICD-9 190), melanoma of the male genitalia (ICD-9 187), and melanoma of the female genitalia (ICD-9 184).

Cancer registration for melanoma (ICD-9 172) in Bristol and District Health Authority, 1984–93

View this table:


Accurate ascertainment is a problem for all cancer registries. Computers in pathology departments can be useful sources of data when compiling a cancer registry, and their use forms one of the recommendations of the Review of the National Cancer Registration System (Alberman report).3 The observed shortfall in ascertainment for the diagnosis of malignant melanoma over recent years could be explained by a lag between diagnosis and registration or by increased use of outpatient treatment, thus reducing the value of inpatient diagnoses as a major data source for cancer registries. Other sources of error include the possibility of clinicians excising lesions without asking for a histopathological report and the variability of diagnosis between individual pathologists. The use of private care affecting cancer registration was also a concern of the Alberman committee.

The relatively low annual incidence of skin cancer makes monitoring of trends difficult even after an accurate register has been established. Bristol and District Health Authority with 830000 residents is the second largest health district in England and Wales,4 yet for melanoma the 95% confidence interval of the standardised rate ratio5 included unity for the change in incidence between 1984 and 1993 (1.27 (0.93 to 1.75)). It is therefore not possible to show a statistically valid increase over this 10 year period. The requirement of the Health of the Nation to monitor the change in annual incidence in each district health authority will need to be modified. Monitoring should be conducted in much larger populations (such as the new NHS regions) to show statistically robust trends.

We thank the staff of the South Western Region Cancer Registry, the secretaries to the consultant plastic surgeons working in Bristol, local pathology laboratories, and the Frenchay medical physics department.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract