TY - GEN
T1 - Pharmacoepidemiologic studies on prescription drugs and cancer risk
AU - Kristensen, Kasper Bruun
PY - 2022/2/3
Y1 - 2022/2/3
N2 - Exogenous exposures such as prescription drugs may affect the risk of cancer. Pharmacoepidemiology, the study of drugs in groups of people, can be used to study these effects, however, drug-cancer studies pose a particular challenge since it may take years or even decades from the initial exposure until a cancer is diagnosed. This thesis includes five studies, of which three examined specific drug-cancer associations, one was a hypothesis-free screening study for carcinogenic effects of drugs, and the last study focused on developing a comorbidity summary score for use in pharmacoepidemiologic studies. In the first study, we summarized existing evidence on vitamin K antagonists and their association with prostate cancer risk and conducted a case-control study using Danish registries to add to the existing evidence. Taken together, the identified studies reported heterogenous results, however, we concluded that a clinically relevant preventive effect of vitamin K antagonists against prostate cancer was unlikely. In the second study, we examined whether antiepileptic drugs were associated with non-melanoma skin cancer and malignant melanoma in a series of case-control studies. Reassuringly, associations were close to unity for most antiepileptic drugsand outcomes except for a positive association between squamous cell carcinoma and lamotrigine and carbamazepine. Due to the hypothesis-generating nature of the study, we concluded that the results had no direct clinical implications and that further research was needed to qualify these signals further.In the third study, we examined whether use of calcium channel blockers was associated with increased risk of kidney cancer. We illustrated methods to identify and account for confounding by indication including the assessment of cumulative dose-response relationships, adjusting for severity of disease, using negative control exposures, and using active comparators. We concluded that the observed association between calcium channel blockers and kidney cancer was at least partly explained by confounding by indication. The fourth study, a hypothesis-free screening study, aimed to identify carcinogenic effects of drugs by examining all drugs and cancers in a series of case-control studies. We identified individuals with incident cancer of 33 sites and 85 histologicalsubtypes. Approximately 14,000 drug-cancer pairs were evaluated, and the estimates were made available online for research purposes. We identified known drug-cancer associations e.g., azathioprine and non-Hodgkin lymphoma, and highlighted anumber of drug-cancer associations that deserved further scrutiny. We concluded that hypothesis-free screening of drug-cancer associations were feasible, however, signal identification remained an issue and the hypothesis-generating nature of theresults should be stressed when communicating the findings of the study. In the fifth study, we developed a numerical score to adjust for comorbidity in pharmacoepidemiologic studies, the Nordic Comorbidity Index. We developed the index in a population-based cohort of randomly sampled Danish residents and theindex included a total of 50 diagnoses or drugs that predicted 5-year mortality. We concluded that the index was superior to the Charlson and Elixhauser comorbidity scores in predicting mortality and that it could be preferred as a summary score instudies utilizing Nordic registry data, however, it remains to be validated in specific patient populations and other Nordic countries.
AB - Exogenous exposures such as prescription drugs may affect the risk of cancer. Pharmacoepidemiology, the study of drugs in groups of people, can be used to study these effects, however, drug-cancer studies pose a particular challenge since it may take years or even decades from the initial exposure until a cancer is diagnosed. This thesis includes five studies, of which three examined specific drug-cancer associations, one was a hypothesis-free screening study for carcinogenic effects of drugs, and the last study focused on developing a comorbidity summary score for use in pharmacoepidemiologic studies. In the first study, we summarized existing evidence on vitamin K antagonists and their association with prostate cancer risk and conducted a case-control study using Danish registries to add to the existing evidence. Taken together, the identified studies reported heterogenous results, however, we concluded that a clinically relevant preventive effect of vitamin K antagonists against prostate cancer was unlikely. In the second study, we examined whether antiepileptic drugs were associated with non-melanoma skin cancer and malignant melanoma in a series of case-control studies. Reassuringly, associations were close to unity for most antiepileptic drugsand outcomes except for a positive association between squamous cell carcinoma and lamotrigine and carbamazepine. Due to the hypothesis-generating nature of the study, we concluded that the results had no direct clinical implications and that further research was needed to qualify these signals further.In the third study, we examined whether use of calcium channel blockers was associated with increased risk of kidney cancer. We illustrated methods to identify and account for confounding by indication including the assessment of cumulative dose-response relationships, adjusting for severity of disease, using negative control exposures, and using active comparators. We concluded that the observed association between calcium channel blockers and kidney cancer was at least partly explained by confounding by indication. The fourth study, a hypothesis-free screening study, aimed to identify carcinogenic effects of drugs by examining all drugs and cancers in a series of case-control studies. We identified individuals with incident cancer of 33 sites and 85 histologicalsubtypes. Approximately 14,000 drug-cancer pairs were evaluated, and the estimates were made available online for research purposes. We identified known drug-cancer associations e.g., azathioprine and non-Hodgkin lymphoma, and highlighted anumber of drug-cancer associations that deserved further scrutiny. We concluded that hypothesis-free screening of drug-cancer associations were feasible, however, signal identification remained an issue and the hypothesis-generating nature of theresults should be stressed when communicating the findings of the study. In the fifth study, we developed a numerical score to adjust for comorbidity in pharmacoepidemiologic studies, the Nordic Comorbidity Index. We developed the index in a population-based cohort of randomly sampled Danish residents and theindex included a total of 50 diagnoses or drugs that predicted 5-year mortality. We concluded that the index was superior to the Charlson and Elixhauser comorbidity scores in predicting mortality and that it could be preferred as a summary score instudies utilizing Nordic registry data, however, it remains to be validated in specific patient populations and other Nordic countries.
U2 - 10.21996/rw46-dn96
DO - 10.21996/rw46-dn96
M3 - Ph.D. thesis
PB - Syddansk Universitet. Det Sundhedsvidenskabelige Fakultet
ER -