Completeness of invasive cervical cancer at the Cancer Registry of Norway

To the Editor

The traditional function of population-based cancer registries as providers of timely and relevant information on cancer incidence has expanded in the last decades to include research on the causes of cancer and their prevention. The quality of the data collected is critical in justifying its use in epidemiological research and decision-making. Completeness is a key indicator of quality and determines the extent to which all incident cancers in the population are included in the registry [1]. The Cancer Registry of Norway has, since 1952, systematically collected notifications on cancer for the Norwegian population. Recent evaluations indicate the Registry is close to complete, partly through compulsory reporting practices and the use of multiples sources of information [2], [3].

The aim of this study is to assess the completeness of both the registry data and paraffin block material from the major laboratories in Norway, with a view to assessing its completeness for research purposes. Cervical cancer was specifically selected given there are a number of published Nordic studies that have combined biological samples and registry data to assess risk factors for the disease (see for example [4], [5]), while there are rather few reports of data quality (see Discussion). Firstly, the completeness of cervical cancer at the Registry is assessed via an independent reascertainment of records in three pathology departments within a sample of time periods. Records of cancer cases identified during the audit are enumerated and matched against registry files so as to provide a quantitative estimate of completeness within the Registry. Secondly, the completeness of stored paraffin blocks at each of the pathology department is investigated and completeness estimated by similar means. The joint assessment is justified given the importance of the source material in preventative and screening research, and in translational research using biobank specimens.

Data and methods

All new cases of invasive cervical cancer (ICC) diagnosed in 1985 and 1999 were identified through the incidence registry at the Cancer Registry of Norway (CRN), the years selected in order to investigate the completeness of ICC at the Registry before and after introduction of the cervical screening program. To study the completeness of registration, CRN data (the registry source) was matched against diagnostic parameters at three selected pathology departments (the laboratory source) using the independent case ascertainment method [6]. Two of the selected departments (denoted Departments 1 and 2) were located at major public university hospitals in different regions of Norway, while the third (Department 3) was a large private laboratory. In combination, the departments diagnosed more than 70% of all ICC patients in Norway. Reports of ICC in 1985 and 1999 were extracted from the registration systems of Departments 1 and 2 on the basis of the morphology codes specified at each pathology department, and enumerated and matched with those available at the Registry. For Department 3, only reports for 2002 were available, and were selected using the same process; these were used as a proxy for the year 1999 (e.g. after the screening programme was introduced). All reports obtained from Department 1 and 3 were re-evaluated, while a random selection of reports was examined from Department 2.

Secondly, the completeness of paraffin blocks was assessed in the laboratories. A random selection of 86 blocks was included in the study. The archives of each laboratory were inspected to assess whether the paraffin blocks could be traced.

The matching of data from the different sources was based on the national 11-digit personal identification number system (PIN). The unique PIN is in use in all public health services and enables the matching of the laboratory and registry sources, including the identity digit (ID) of pathology reports and the ID of paraffin blocks.

Results

At the Registry, 348 and 283 cases were diagnosed in 1985 and 1999 respectively, of which 93% (in 1985) and 97% (in 1999) were based on a histological examination of the primary tumour. Autopsy was the single notification source for only two cancer cases, both diagnosed in 1985. None of the diagnoses in 1985 and 1999 were based on death certificates only. The histological profile of ICC cases was similar in 1985 and 1999, with about four-fifths squamous cell carcinomas (SCC), 15% adenocarcinoma and 5% other histological types (data not shown).

Comparing the numbers of reports and cases between the registry and laboratory sources (Table I), about 15% of the reports (43 of 284) were absent at the Cancer Registry in 1985, but this reduced to around 3% (8 of 253) in 1999. Although two diagnoses of ICC in 1985 obtained from the audit of Department 1 were not found in the incidence registry at the CRN, all other cases from the laboratory source could be matched to the Registry, giving a completeness estimate of ICC at CRN of 98.8% and 100% for the years of registration 1985 and 1999. In terms of the availability of blocks in each department, all blocks except one were found in the archives, suggestive of a high completeness proportion, estimated at 97.9%.

Table I.  Completeness of invasive cervical cancer (ICC) registrations at the Cancer Registry of Norway (CRN) and the completeness of paraffin blocks held in three pathological laboratories for the years 1985 and 1999.

1985
Department	Number of ICC cases (reports) in laboratory source	Number of ICC cases (reports) in registry source (CRN)	Estimated completeness of ICC cases at CRN
1	31 (78)	31 (54)	100%
2	142 (206)	140 (187)	98.6%
3*	–	–	–
Total	173 (284)	171 (241)	98.8%
	Number of selected blocks in laboratory sources	Number of blocks found in archives	Estimated completeness of selected blocks in laboratory source
1	15	15	100%
2	11	11	100%
3	12	12	100%
Total	38	38	100%
1999
Department	Number of ICC cases (reports) in laboratory source	Number of ICC cases (reports) in registry source (CRN)	Estimated completeness of ICC cases at CRN
1	35 (39)	35 (38)	100%
2	86 (190)	86 (185)	100%
3**	19 (24)	19 (22)	100%
Total	140 (253)	140 (245)	100%
	Number of selected blocks in laboratory sources	Number of blocks found in archives	Estimated completeness of selected blocks in laboratory source
1	22	21	95.5%
2	14	14	100%
3	12	12	100%
Total	48	47	97.9%

*Not available in Department 3.

**Year 2002 for Department 3.

Discussion

Data quality and its continued assessment are fundamental concerns to a functioning registry, and in this communication we have focussed on cervix cancer and its completeness. Given the burgeoning etiologic and clinical research involving information from biological samples linked to population-based cancer registries, there is an increasing need to jointly evaluate the quality of the sources. Evidently, the combination of cancer registry data and corresponding biological material will only be of optimal scientific value if both sources can be considered to have a high degree of completeness.

Cancer of the cervix has tended to be, in relative terms, less reported than other cancers in certain registries [7], [8], although this study confirms the recent review that cervical cancer registrations at the CRN can be considered close-to-complete [2]. The estimated completeness was above 98% in both 1985 and 1999, and the marked decline in the number of apparently-missed reports is in line with documented improvements in registration practices including efficient multiple-source reporting and trace-back [2], [3].

Only one of 86 paraffin blocks in the selected pathology departments was apparently missing. Almost all pathology departments in Norway store their paraffin-embedded tissues in archives, and there is a huge amount of potential research material stored within each department. The three centres from which completeness of paraffin blocks were estimated to process over two-thirds of all cervix cancer diagnoses in Norway.

The independent case ascertainment method used here is a commonly-applied approach to providing quantitative estimates of completeness [1]. Records of cancer cases identified during an audit of three pathology laboratories over two periods were enumerated and matched against the files at the CRN. Unmatched cases were followed back to ascertain their reportability, and the percentage of cases apparently missed were estimated. A Danish study has used a similar approach based on three radiotherapy centres and one major gynaecology hospital [9], and a US report has matched medical records from pathology and radiation oncology departments in hospitals to case listings (from SEER) to identify unreported cases [8].

A Canadian study has attributed high rates of false-positive diagnosis of ICC at the British Columbia Cancer Registry to the lack of a procedure to downgrade the first diagnosis in the light of information subsequently recorded at the registry [10]. The conclusion was that the accuracy of the recorded data could be further improved by establishing procedures for collecting and recording all the pathological information generated during routine diagnosis and treatment of registered cases and by instituting quality control procedures for checking the reliability of the coding practices. Unfortunately, the pathology departments in Norway examined in this study do not link cancer reports to cancer cases. Each department files reports and therefore cannot downgrade cytology reports when later biopsies confirm earlier cancer diagnoses. This can result in an overestimate of cancer incidence using this data source, and for research purposes, data at the CRN is timely and may be considered of superior quality.

There are certain drawbacks in this investigation, particularly the rather small number of cases enumerated and matched, which in turn could lead to imprecise estimates of completeness. However the departments diagnose almost three-quarter of all ICC patients in Norway, and the data are not, to our knowledge, differentially incomplete; on this assumption we may therefore consider their aggregation as reasonably nationally representative. In addition, the results are in line with previous positive evaluations of the registry material and we may reasonably conclude that there is likely a high degree of completeness of cervical cancer at the CRN. Coupled with the high completeness estimates found for the biological samples (paraffin blocks) in selected laboratories, this report provides encouraging support for cancer research, and hopefully also reinforces the need for joint evaluations of biological and registry material as we further integrate biobanking and cancer registration activities.

Acknowledgements

This study was undertaken as part of the Cancer Control using Population-based Registries and Biobanks (CCPRB) project, funded by the EU Sixth Framework Programme (FP6-2002-LifeSciHealth Contract No. 503465).The authors would like to thank the Pathology Departments participating in this study: the Department of Pathology, Haukeland University Hospital, Bergen; the Laboratory for Pathology AS, Oslo and the Department of Pathology, The Norwegian Radiumhospital, Oslo. The work was partially undertaken while EB and ST were employed at the Cancer Registry of Norway.