This study included residents of the four former Danish counties of North Jutland, Aarhus, Viborg, and Ringkoebing, which have a combined population of approximately 1.7 million inhabitants. The Danish National Health Service provides tax-supported health care to all residents of the country and refunds part of patient expenditures on a wide range of prescribed drugs, including NSAIDs and sCOX-2 inhibitors, the latter of which became available in Denmark in 1999. All health-related services are registered to individual patients by use of their civil personal registration (CPR) number, assigned to all Danish citizens since 1968. This unique CPR number facilitates linkage between population-based registries, including the Danish Cancer Registry (10
), the Danish National Registry of Patients (11
), and countywide prescription databases.
The methods we used have been described elsewhere (12
), with expansion of the study base to the four former counties for which data have been merged into a research database at Aarhus University. Briefly, we used the Cancer Registry to identify all patients (N
=403) who had a first diagnosis of HL (International Classification of Diseases, 10th
revision (ICD-10) code C81) starting on January 1, 1991, in North Jutland County; January 1, 1998, in Aarhus County; and January 1, 2000, in Viborg and Ringkoebing counties; and continuing through December 31, 2006, in all four former counties (now merged into two regions); and we used the National Registry of Patients (which has more recent data) to ascertain all patients (N
=75) who had a first diagnosis of HL within all hospitals in the two regions from January 1, 2007, through December 31, 2008. The Cancer Registry includes the CPR number and detailed individual data on all cancer diagnoses in Denmark since 1943 (10
), while the Registry of Patients includes the CPR number and detailed individual data on all non-psychiatric hospital admissions since 1977 and out-patient contacts since 1995 (11
). Until 2003, the Cancer Registry was based on mandatory notifications by Danish medical doctors (10
); since then, the Cancer Registry has been based on records from the Registry of Patients, with secondary histological confirmation from the National Pathology Registry (13
). Within the Danish Civil Registration System database (15
), we performed risk-set sampling to select 10 population controls per case among living individuals without a history of HL on the index date (i.e., the date of diagnosis for each case), for a total of 4,780 controls matched on age, sex, and county of residence.
All pharmacies in the four former counties are equipped with computerized accounting systems that record a customer’s CPR number and prescription data, including type and quantity according to the Anatomical Therapeutic Chemical (ATC) Classification System (17
), and date of dispensing at the pharmacy. This information is transferred electronically to countywide prescription databases (12
). Using these databases (established in North Jutland County in 1989, Aarhus County in 1996, and Viborg and Ringkoebing counties in 1998, thus ensuring a minimum of two years of prescription history in the present study), we identified prescriptions for low-dose aspirin (75, 100, or 150 mg per tablet; ATC codes B01AC06 and N02BA01), high-dose aspirin (500 mg per tablet; ATC codes N02BA51 and N02BA01), sCOX-2 inhibitors (ATC codes M01AH01, M01AH, M01AH03, M01AH05, M01AC05, M01AB05, and M01AC06), and other NSAIDs (remaining ATC codes within group M01A). We excluded prescriptions within 1 year of the index date to reduce any potential effect of subclinical disease on medication use.
We defined “ever users” of a medication as individuals who had >2 prescriptions and “never/rare users” as those who had ≤2 prescriptions. The average length of a prescription was 30 days. Ever users were further divided into recent users (those who had >2 prescriptions during the period 1–2 years before the index date) and former users (>2 prescriptions overall, but ≤2 during the recent period). Duration of use was classified as long-term (≥7 years) or short-term (<7 years), based on the number of days between the first and last prescriptions plus the duration of the last prescription. Intensity of use was defined as low (<25%) or medium/high (≥25%), according to the number of days of prescription coverage divided by duration of use in days (12
To obtain a nonspecific proxy for chronic NSAID use, we identified subject comorbidities before the index date using inpatient and outpatient data from the Registry of Patients. Comorbidities were summarized using Deyo’s adaptation of the Charlson Index (19
). In addition, because connective tissue disorders in particular (e.g., rheumatoid arthritis, which is associated with both greater NSAID use and higher HL risk (21
)) may confound the associations of interest, we identified connective tissue disorders before the index date.
We used conditional logistic regression to compute ORs and 95% CIs, matching on age, sex, and former county of residence, and additionally adjusting for Charlson Index (0, 1–2, or ≥3 comorbidities). Further adjustment for connective tissue disorders did not affect the results (data not shown). Because the etiology of HL varies by age (23
), we stratified the results by age (<40 years vs. ≥40 years) and performed Wald tests for interactions between medication use and age group. In all analyses, never/rare users (≤2 prescriptions) comprised the reference group. Given the risk set sampling of controls, the ORs are estimates of the incidence rate ratios in the underlying population.