Cancer risk in persons receiving prescriptions for paracetamol: A Danish cohort study
Article first published online: 3 OCT 2001
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 97, Issue 1, pages 96–101, 1 January 2002
How to Cite
Friis, S., Nielsen, G. L., Mellemkjær, L., McLaughlin, J. K., Thulstrup, A. M., Blot, W. J., Lipworth, L., Vilstrup, H. and Olsen, J. H. (2002), Cancer risk in persons receiving prescriptions for paracetamol: A Danish cohort study. Int. J. Cancer, 97: 96–101. doi: 10.1002/ijc.1581
- Issue published online: 10 DEC 2001
- Article first published online: 3 OCT 2001
- Manuscript Accepted: 30 JUL 2001
- Manuscript Revised: 24 JUL 2001
- Manuscript Received: 21 FEB 2001
- Danish Medical Research Council. Grant Number: 9700677
- Danish Cancer Society
- International Epidemiology Institute
- cohort study
The use of paracetamol has been associated with increased risks for urinary tract cancers and decreased risk for ovarian cancer, although results have been inconsistent. We conducted a population-based cohort study using data from the Prescription Database of North Jutland County and the Danish Cancer Registry. Cancer incidence among 39,946 individuals receiving prescriptions for paracetamol was compared with expected incidence based on the North Jutland population who did not receive paracetamol prescriptions, during a 9-year follow-up period. Standardized incidence ratios (SIRs) with corresponding 95% confidence intervals (95% CIs) were calculated for cancers overall and at selected sites. Overall, 2,173 cancers were observed with 1,973 expected, yielding a SIR of 1.10 (95% CI, 1.06–1.15). Significantly elevated SIRs were found for cancers of the esophagus (1.9; 95% CI, 1.3–2.8) and lung (1.6; 95% CI, 1.4–1.7). Nonsignificantly increased SIRs were observed for cancers of the liver (1.5; 95% CI, 0.96–2.2), renal parenchyma (1.3; 95% CI, 0.9–1.7) and renal pelvis/ureter (1.6; 95% CI, 0.96–2.6), whereas the SIR for cancer of the urinary bladder was close to unity (1.1; 95% CI, 0.9–1.4). For ovarian cancer, the SIR was close to expectation (0.9; 95% CI, 0.6–1.2) with no evidence of trends with duration of follow-up or number of prescriptions. A similar risk pattern was observed after exclusion of person-time experience following prescription for aspirin or other nonsteroidal antiinflammatory drugs in the study cohort and reference population. Our results do not support a major role for paracetamol in the development of cancers of the urinary tract, and we found little evidence of a protective effect of paracetamol against ovarian cancer. The elevated risks for cancers of the esophagus, lung and liver are most likely a result of confounding variables, but may warrant further investigation. © 2002 Wiley-Liss, Inc.
Concern has been raised about the carcinogenic potential of paracetamol (acetaminophen) because it is the major metabolite of phenacetin, which was classified as a human carcinogen by the International Agency for Research on Cancer (IARC) in 1987 and has been withdrawn from the market in most countries.1–3 In a recent review,4 the IARC concluded that there is currently “inadequate evidence” in both humans and experimental animals for the carcinogenicity of paracetamol. Because of the established link between phenacetin and malignant tumors of the urinary tract,1 most epidemiologic studies of paracetamol and cancer have focused on these tumors. Some of these have reported slightly elevated risks of renal cell cancer2, 5, 6 or transitional cell cancers of the renal pelvis, ureter or urinary bladder2, 7–11 with regular or long-term use of paracetamol, whereas other studies have failed to demonstrate such associations.3, 5, 7, 12–18
Few other cancer sites have been examined in relation to use of paracetamol. In a recent case-control study, Cramer et al.19 reported an inverse association between paracetamol use and risk of ovarian cancer. Based on experimental data, the authors suggested that paracetamol exerts an antigonadotropic effect that could result in a reduced risk of ovarian cancer.19 The finding by Cramer et al. was weakly supported by a prospective mortality study that showed a decreased ovarian cancer mortality rate among daily users of paracetamol at enrollment,20 whereas a large, hospital-based case-control study did not find a decreased risk of ovarian cancer associated with regular use of paracetamol.21
Paracetamol is an analgesic and antipyretic drug used for the treatment of minor noninflammatory conditions and, to a lesser extent, for more severe conditions such as chronic inflammatory arthritis and cancer. Because of the widespread use of paracetamol, any association with an increased or decreased cancer risk may have important public health implications. The nonconclusive nature of the epidemiologic evidence on its carcinogenic potential, deriving primarily from case-control studies of urinary tract cancers, prompted us to examine the incidence of cancer in a population-based cohort of persons receiving prescriptions for paracetamol.
MATERIAL AND METHODS
The population-based Pharmacoepidemiologic Prescription Database of North Jutland county in Denmark22 was used to identify a total of 375,357 paracetamol prescriptions filled for 51,935 individuals in the period January 1, 1989 (start of prescription database), to December 31, 1995. During this period, the population of the county was approximately 490,000, representing about 9% of the total Danish population. Through a computerized accounting system maintained by Danish pharmacies, the tax-supported health insurance program in Denmark refunds 50–75% of the costs associated with the purchase at pharmacies of most drugs prescribed by doctors. In the county of North Jutland, this accounting system also transfers prescription data to the Pharmacoepidemiologic Prescription Database,22, 23 including data on type of drug prescribed according to the Anatomical Therapeutical Chemical (ATC) classification system,24 date of prescription and the customer's personal identification number, a unique number assigned to all Danish residents. Use of the personal identification number ensures that a complete prescription history since 1989 can be established for each study participant, and permits valid linkage of information between registers.
Paracetamol is available over-the-counter in Denmark, but for pensioners or patients with chronic pain or disease requiring prolonged daily treatment, paracetamol can be prescribed with the insurance program reimbursing the patient for 50% of the costs. The drug was introduced in Denmark in 1977 and has been available over-the-counter since 1984. The prescriptions for paracetamol were identified in the prescription database by the ATC code N02BE01. Overall, 2,045 (3.9%) of the identified persons receiving prescriptions for paracetamol were excluded because of (i) residency outside the county of North Jutland at the date of prescription (n = 1,096); (ii) parent (of patient) registered as customer (n = 873); (iii) error in the personal identification number (n = 11); (iv) death prior to or at the date of prescription (n = 48); or (v) age younger than 16 years (n = 17); after these exclusions 49,890 (96.1%) persons were left for subsequent linkage.
Information on cancer occurrence was obtained by linkage to the Danish Cancer Registry, which has recorded incident cases of cancer on a nationwide basis since 1943.25 The Danish Cancer Registry has been shown to have accurate and virtually complete ascertainment of cancer cases.26 Cancers were classified according to a modified Danish version of the International Classification of Diseases, 7th revision.27 Kidney cancer cases classified as “kidney cancer, not otherwise specified” (ICD-7, 180.0) were regarded as situated within the renal parenchyma, as a previous validation study revealed that the vast majority of these cancers are renal cell carcinomas.28 Study subjects with a cancer diagnosis, except nonmelanoma skin cancer, prior to the date of their first recorded prescription for paracetamol (n = 6,050; 11.6%) were excluded. Further, to minimize any selection bias introduced by the inclusion of patients with recent or undiagnosed cancer receiving paracetamol for alleviation of pain, we excluded the person-time and cancer experience in the first year of follow-up after first paracetamol prescription. This led to the exclusion of 3,894 persons (7.5%) who had a cancer diagnosis (n = 782) or died (n = 3,112) within the first year of follow-up. After these exclusions, the study cohort included 39,946 (76.9%) individuals.
The follow-up period began 1 year after the date of the first recorded prescription for paracetamol and ended on the date of first primary cancer diagnosis except nonmelanoma skin cancer (n = 1,777), date of death (n = 8,492), or December 31, 1997 (n = 29,677), whichever occurred first. Information on date of death for subjects who died during follow-up was obtained through linkage to the National Mortality Files. Expected numbers of first primary cancers in the study cohort were calculated by multiplying the number of person-years of the cohort members with the sex, 5-year age group and 5-year calendar-year specific incidence rates of first primary cancers for all inhabitants of North Jutland county who had not received a prescription for paracetamol. The standardized incidence ratio (SIR), that is, the ratio of the observed to the expected number of cancers, and 95% confidence intervals (CIs) were calculated, assuming a Poisson distribution of the observed cancers.29 A χ2 test for linear trend in SIRs was used to evaluate trends with number of paracetamol prescriptions. For these dose–response analyses, each study subject was assigned to 1 of 4 categories of total prescription frequency during the registration period, and follow-up for cancer began on the date of the first prescription within the given category.
In an additional analysis, using the prescription history for each study subject, we excluded the person-time experience after first recorded prescription for aspirin or other nonsteroidal antiinflammatory drugs (NSAIDs), if any, during the registration period (1989–1995). This led to exclusion of 21,832 persons (54.7% of full cohort) who received aspirin or nonaspirin NSAIDs before receiving paracetamol, and 4,632 persons (11.6% of full cohort) who received aspirin or nonaspirin NSAIDs within 1 year after first prescription for paracetamol, leaving 13,482 “exclusive” users of paracetamol (33.8% of full cohort) for this analysis (Table I). Censoring of follow-up time was performed for 4,696 persons (34.8% of subcohort) who had a prescription for aspirin or nonaspirin NSAIDs during the follow-up period. Similarly, we excluded the person-time experience after prescription for NSAIDs in the North Jutland reference population. The additional analysis was performed in an effort to disentangle the effect, if any, of paracetamol from that associated with aspirin or nonaspirin NSAIDs.
|Characteristic||Total cohort||Paracetamol only1|
|Age at entry2 (years)|
|Year of entry2|
|Number of prescriptions|
Characteristics of the study cohort of 13,674 men (34%) and 26,272 women (66%) are presented in Table I. The average age at cohort entry, that is, time of first recorded prescription for paracetamol, was 63 years, and the average follow-up time since first prescription was 4.7 years (range 1–9 years), based on 148,850 accrued person-years. Sixty-five percent of study subjects received 2 or more prescriptions for paracetamol during the registration period. The subcohort of “exclusive” users of paracetamol (as defined above) had a similar age distribution (Table I), and they accrued 38,888 person-years of follow-up, with a mean follow-up period since first prescription of 3.9 years (range 1–9 years).
Overall, 2,173 cancers were diagnosed in the full cohort with 1,973 expected, yielding a SIR of 1.10 (95% CI, 1.06–1.15) (Table II). Significantly increased SIRs were found for cancers of the esophagus (SIR, 1.9; 95% CI, 1.3–2.8) and lung (SIR, 1.6; 95% CI, 1.4–1.7). Nonsignificantly increased SIRs were seen for cancer of the liver (SIR, 1.5; 95% CI, 0.96–2.2), cancer of the renal parenchyma (SIR, 1.3; 95% CI, 0.9–1.7) and cancer of the renal pelvis or ureter (SIR, 1.6; 95% CI, 0.96–2.6), whereas for urinary bladder cancer the SIR was close to unity (SIR, 1.1; 95% CI, 0.9–1.4). A slight risk reduction was observed for colon cancer (SIR, 0.8; 95% CI, 0.7–1.0), whereas the SIR for ovarian cancer was close to expectation (SIR, 0.9; 95% CI, 0.6–1.2).
|Cancer site (modified ICD-7 code)||Total cohort (n = 39,946)||Paracetamol only1 (n = 13,482)|
|Obs2||SIR||95% CI||Obs2||SIR||95% CI|
|All malignant neoplasms (140–205)||2173||1.10||1.06–1.15||530||1.10||1.01–1.20|
|Buccal cavity and pharynx (140–148)||37||1.2||0.8–1.6||12||1.3||0.7–2.3|
|Cervix uteri (171)||21||0.8||0.5–1.2||4||0.5||0.1–1.1|
|Corpus uteri (172)||58||1.1||0.8–1.4||15||1.2||0.7–2.0|
|Urinary tract (180–181)||171||1.2||1.0–1.4||34||1.0||0.7–1.4|
|Renal parenchyma3 (180.0/180.3)||38||1.3||0.9–1.7||7||1.0||0.4–2.1|
|Urinary bladder (181)||115||1.1||0.9–1.4||25||1.0||0.7–1.5|
|Nonmelanoma skin cancer (191)||375||1.0||0.9–1.1||114||1.3||1.1–1.6|
|Lymphatic and haematopoietic tissue (200–205)||126||1.1||0.9–1.3||33||1.2||0.8–1.6|
|Non-Hodgkin lymphoma (200, 202)||47||1.1||0.8–1.5||14||1.2||0.7–2.0|
|Hodgkin's disease (201)||6||2.0||0.7–4.4||1||1.4||0.0–8.0|
|Multiple myeloma (203)||25||1.1||0.7–1.7||7||1.6||0.6–3.2|
Limiting the analysis to the period 5 or more years after first recorded paracetamol prescription yielded further elevations in the SIRs for cancers of the esophagus (SIR, 2.5; 95% CI, 1.1–5.0), lung (SIR, 1.8; 95% CI, 1.4–2.2), liver (SIR, 2.0; 95% CI, 0.8–4.1) and renal pelvis/ureter (SIR, 3.0; 95% CI, 1.2–6.2), but not for cancer of the renal parenchyma (SIR, 1.0; 95% CI, 0.4–2.1) (data not shown). For ovarian cancer, the SIR was slightly increased (SIR, 1.3; 95% CI, 0.7–2.2) in the 5- to 9-year follow-up period, whereas the ratio for colon cancer was nonsignificantly decreased in this period (SIR, 0.7; 95% CI, 0.5–1.0).
The SIR estimates did not vary markedly when stratified by age at entry (<70 years; ≥70 years). Similarly, the results were comparable in men and women, except for a slight difference in risk of digestive system cancers between men (SIR, 1.1; 95% CI, 0.8–1.5) and women (SIR, 0.8; 95% CI, 0.6–1.0) with 5 or more years of follow-up (data not shown). This difference mainly pertained to cancers of the esophagus [SIR, 4.9 (95% CI, 2.0–10.1) in men vs. 0.6 (95% CI, 0.0–3.2) in women] and liver [SIR, 2.8 (95% CI, 0.9–6.5) in men vs. 1.2 (95% CI, 0.1–4.3) in women].
In the subcohort of “exclusive” users of paracetamol, the SIRs were similar to those in the full cohort for overall risk of cancer (SIR, 1.10; 95% CI, 1.01–1.20), and for most specific cancer sites including urinary bladder cancer (SIR, 1.0; 95% CI, 0.7–1.5), ovarian cancer (SIR, 0.8; 95% CI, 0.4–1.4) and colon cancer (SIR, 0.9; 95% CI, 0.7–1.3) (Table II). However, there was no excess of upper urinary tract cancers, that is, cancers of the renal parenchyma (SIR, 1.0; 95% CI, 0.4–2.1) and renal pelvis/ureter (SIR, 0.9; 0.1–3.2) in the subcohort.
Table III presents results, stratified by number of prescriptions, for cancer overall and for the sites of a priori interest, that is, cancers of the urinary tract and ovary. For cancer overall, the SIR estimates were similar in the 4 categories of prescription frequency. For cancer of the renal parenchyma, there was a significant trend of higher SIRs with increasing number of prescriptions (p = 0.04). For cancer of the urinary bladder and ovarian cancer, no consistent trends in SIRs with number of prescriptions were observed. Similar results were found for cancer of the urinary bladder and cancer overall in the subcohort of “exclusive” users of paracetamol, whereas the data for upper urinary tract cancers and ovarian cancer were too sparse to be evaluated by number of prescriptions.
|Cancer site/No. of prescriptions (1, 2–4, 5–9, ≥10)||Total cohort||Paracetamol only1|
|Obs2||SIR||95% CI||Obs2||SIR||95% CI|
With follow-up beginning 1 year after the first recorded prescription, the observed cancer incidence among nearly 40,000 persons receiving prescriptions for paracetamol was close to that expected in the general population who did not receive prescriptions for paracetamol. We found no evidence of an association between use of paracetamol and risk of urinary bladder cancer, but some evidence of an association with upper urinary tract cancers, including cancers of the renal parenchyma, renal pelvis and ureter. The tendency toward a dose–response relationship between number of prescriptions and increase in risk for cancer of the renal parenchyma, and the tendency toward increasing risk with increasing length of follow-up for cancers of the renal pelvis and ureter, support the possibility of a causal relation. However, none of the risk estimates was statistically significant. Further, although based on small numbers, the absence of an overall increase in risk of upper urinary tract cancers in the subcohort of “exclusive” users of paracetamol may indicate that the associations observed in the full cohort were related to former or concomitant use of NSAIDs rather than to paracetamol.
Previous epidemiologic investigations of the association between use of paracetamol and cancer of the renal parenchyma (ie, renal cell cancer) have shown diverging results,3, 5–7, 13, 15, 17 and the weak associations identified in some of the studies may be explained by bias or uncontrolled confounding.30 In a recent study of 1,204 matched case-control pairs, Gago-Dominguez et al.6 reported a significant odds ratio (OR) of 1.7 (95% CI, 1.3–2.1) for renal cell cancer associated with regular use of paracetamol, whereas McCredie et al.,3 in the largest study of this cancer to date, found no increase in risk of renal cell cancer among regular users of paracetamol (1,732 cases and 2,309 controls). The epidemiologic results for transitional cell cancers of the urinary bladder, ureter and renal pelvis, all derived from case-control studies, are also inconsistent.2, 5, 7–14 Most studies of cancer of the renal pelvis and ureter have been small,2, 7, 10 because of the rarity of these tumors, and the early studies reporting positive associations7, 8–10 may not have been able to discriminate between risks associated with earlier phenacetin use and more recent use of paracetamol. In the largest study of renal pelvis and ureter cancer to date,12 no association with paracetamol use was seen. Of 6 case-control studies that examined use of paracetamol in relation to cancer of the urinary bladder, only the study by Steineck et al.11 found an increased risk. These authors reported a significantly increased risk for transitional cell cancer among all users of paracetamol (OR, 1.6; 95% CI, 1.1–2.3), with the highest risk among users of paracetamol not exposed to aspirin (OR, 2.1; 95% CI, 0.7–6.6). However, analyses by duration of use and quantity of paracetamol did not support these associations. In the largest case-control study of analgesic agents and bladder cancer, Castelao et al.14 reported an OR of 1.0 (95% CI, 0.7–1.6) among regular users of paracetamol.
Our study provides only weak support for the hypothesis that use of paracetamol is associated with a decreased risk of ovarian cancer.19 Although we did observe a slight decrease in risk in both the full cohort (SIR, 0.9) and the subcohort of “exclusive” users of paracetamol (SIR, 0.8), we found no evidence of a dose–response relation or temporal trend when the analysis was stratified according to number of prescriptions or duration of follow-up. Cramer et al.19 reported a reduced OR of 0.5 (95% CI, 0.3–0.9) for epithelial ovarian cancer associated with regular use of paracetamol, based on 26 users among 563 cases and 46 users among 523 population controls. The lowest risks were found for women who had used the drug daily and for more than 10 years. However, low response rates among both cases and controls, leading to possible biased selection of study subjects, limit the interpretation of these results. To test the hypothesis, Rodriguez et al.20 examined the association between use of paracetamol and death rates from ovarian cancer in a prospective study of more than half a million American women. After 12 years of follow-up, women who reported daily use of 1 brand of paracetamol (Tylenol) during the month before enrollment had a 45% lower death rate from ovarian cancer than women reporting no use at baseline. However, when examining all current use of paracetamol (<30 times during the month before study entry), no relation was observed and there was no association with duration of use. A major limitation of this study was the collection of data on medication intake only at enrollment. Most recently, Rosenberg et al.21 examined the relation between paracetamol use and ovarian cancer in a large, hospital-based case control study, comparing 780 women with epithelial ovarian cancer to 2,053 cancer controls and 2,570 noncancer controls. For paracetamol use 1 day or more per week for at least 6 months, the ORs were 0.9 (95% CI, 0.6–1.4) and 1.0 (95% CI, 0.6–1.5) derived with cancer and noncancer controls, respectively.
With respect to other cancer sites, our results should be interpreted cautiously given the multiple comparisons performed and the absence of a priori hypotheses. Confounding by tobacco smoking and alcohol consumption remains a possible explanation for some of our results. Several studies have demonstrated that smoking and alcohol consumption are positively associated with the use of analgesic agents.31, 32 We observed a significantly increased SIR for cancer of the esophagus, which is strongly associated with both tobacco smoking33 and alcohol consumption,34 and we observed elevated SIRs for several other cancers known to be associated with smoking or alcohol, including cancers of the buccal cavity and pharynx, liver, lung and urinary tract. Individuals with a high intake of alcohol may be overrepresented in our study cohort, because physicians are likely to have preferentially prescribed paracetamol for these high morbidity patients because of the gastrointestinal and hematologic side effects of aspirin and other NSAIDs. However, our finding of an increased risk of liver cancer may warrant further research, as paracetamol can be hepatotoxic and has been shown to induce liver tumors in some strains of rodents.4
In this cohort study, we had the advantage of being able to collect information on prescription for paracetamol and other analgesic agents from a pharmacoepidemiologic prescription database covering all pharmacies in a Danish county. This method enabled us to establish a large study population with virtually complete follow-up for cancer obtained by computerized linkage to the Danish Cancer Registry covering the entire population of Denmark. The ability to exclude follow-up experience after prescription for aspirin or nonaspirin NSAIDs during the registration period, and thus separate the effect of paracetamol from that of other analgesic agents, was another advantage of our study. For example, the slightly reduced risk of colon cancer observed in the full cohort may be related to concomitant use of NSAIDs, shown in a number of epidemiologic studies to be associated with lower rates of this cancer.35 Limitations of this approach were the lack of data on use of analgesic agents prior to the start of the prescription database and use of over-the-counter analgesic agents during the study period; it is possible that former or concurrent use of aspirin and other NSAIDs may have confounded the risk estimates for cancers associated with use of these drugs even in the subcohort of “exclusive” users. However, confounding by over-the-counter analgesic agents is unlikely to be significant for several reasons. First, the use of over-the-counter analgesic agents, including paracetamol, is likely to be minimal among persons accepted for reimbursement of paracetamol, because aspirin and nonaspirin NSAIDs are also reimbursable if prescribed by a physician. Second, a substantial proportion of the study population consisted of elderly persons who can obtain over-the-counter drugs more cheaply through prescription. Third, of all preparations of nonaspirin NSAIDs on the market during the study period, only a low-dose formulation of ibuprofen was available over-the-counter. Indeed, the use of over-the-counter analgesic agents may have been less common among persons accepted for reimbursement of paracetamol than in the general population not receiving prescriptions for paracetamol and NSAIDs (subcohort analysis), so possible exposure misclassification (paracetamol) and confounding by over-the-counter analgesic agents may have led to underestimation of any true associations.
In our register-based approach, we had to rely on prescription data, with no information on dosages and dose schedules, or degree of compliance. The fact that drug exposure was based on prescriptions actually dispensed at pharmacies and paid for in part by the patient is likely to have improved compliance. Nonetheless, some misclassification due to noncompliance is possible, leading to attenuation of risk estimates.
We had no information on the indications for use of paracetamol in our study population. Confounding by indication could arise if some of the patients receiving paracetamol by prescription had medical conditions associated with cancer risk, either directly or indirectly through other medications used to treat the conditions. One such condition might be rheumatoid arthritis. However, we found no apparent excess of cancers associated with rheumatoid arthritis, notably non-Hodgkin's lymphoma.36
Finally, our follow-up period of up to 9 years may be too short to observe a carcinogenic effect, although a substantial proportion of the persons included in our study were likely to have been prevalent users at the start of the prescription database. Interestingly, when the study cohort was restricted to persons enrolled during the first 6 months of the registration period—presumably including a considerable proportion of prevalent users—the results were similar to those for the full cohort (data not shown).
In summary, we found that total cancer incidence among individuals receiving prescriptions for paracetamol was close to that expected in the general population not prescribed paracetamol or NSAIDs. Although use of paracetamol is not likely to play a major role in the development of cancers of the urinary tract, a small increase in risk of upper urinary tract cancers cannot be ruled out. We found limited evidence to support the previously hypothesized preventive effect of paracetamol on the development of ovarian cancer. The elevated risks of lung, liver and esophageal cancers may deserve further investigation in studies with information on potential confounders, notably tobacco and alcohol consumption.
We thank Mr. L. Thomassen and Mr. S. Bang for help with the data management.
The Pharmaco-Epidemiologic Prescription Database has been supported by The Danish Medical Research Council (grant 9700677). The activities at the Institute of Cancer Epidemiology were supported by The Danish Cancer Society and The International Epidemiology Institute.
- 1IARC Monographs on the evaluation of carcinogenic risks to humans, suppl. 7. Overall evaluations of carcinogenicity: an updating of IARC monographs 1 to 42. Lyon: IARC, 1987. 310–2.
- 4IARC Monographs on the evaluation of carcinogenic risks to humans, Vol. 73. Some chemicals that cause tumours of the kidney or urinary bladder in rodents and some other substances. Lyon: IARC, 1999: 401–49.
- 22The Pharmaco-epidemiologic Prescription Database of North Jutland—a valid tool in pharmacoepidemiological research. Int J Saf Med 1997;10: 203–5., , , et al.
- 24Descriptive tools and analysis. In: DukesMNG, ed. Drug utilization studies: methods and uses. WHO regional publications (European series No. 45). Geneva: World Health Organization, 1993: 55–78..
- 27Cancer incidence in Denmark, 1993. Copenhagen: Danish Cancer Society, 1996., , .