Considerations for Pharmacoepidemiological Studies of Drug-Cancer Associations

ACE (Angiotensin-Converting Enzyme) Inhibitors/Angiotensin Receptor Blockers Are Associated With Lower Colorectal Cancer Risk: A Territory-Wide Study With Propensity Score Analysis

Whether ACE (angiotensin-converting enzyme) inhibitors and angiotensin receptor blockers modify colorectal cancer risk remains controversial. We aimed to determine association between their use and colorectal cancer risk after a negative baseline colonoscopy. This is a territory-wide retrospective cohort study recruiting patients aged ≥40 who had undergone colonoscopy between 2005 and 2013. Exclusion criteria included colorectal cancer detected <6 months of index colonoscopy, prior colorectal cancer, inflammatory bowel disease, and prior colectomy. The primary outcome was colorectal cancer diagnosed between 6 and 36 months after index colonoscopy. Sites of colorectal cancer were categorized as proximal (proximal to splenic flexure) and distal cancer. The adjusted hazard ratio of colorectal cancer with ACE inhibitor/angiotensin receptor blocker use (≥180-day use within 5 years before index colonoscopy) was derived by propensity score regression adjustment of 23 covariates (including patient's factors, concurrent medication use, and endoscopy center's performance). Of 187 897 eligible patients, 30 856 (16.4%) were ACE inhibitors/angiotensin receptor blocker users. Eight hundred fifty-four (0.45%) developed colorectal cancer between 6 and 36 months after index colonoscopy (proximal cancer: 147 [17.2%]). These drugs were associated with lower risk of colorectal cancer that developed <3 years after index colonoscopy (adjusted hazard ratio, 0.78 [95% CI, 0.64-0.96]), but not colorectal cancer that developed >3years (adjusted hazard ratio, 1.18 [95% CI, 0.88-1.57]); every single year increase in the drug use was associated with 5% reduction in adjusted hazard ratio risk. ACE inhibitors/angiotensin receptor blocker were associated with a lower colorectal cancer risk in a duration-response manner.

Use of proton pump inhibitors and histamine-2 receptor antagonists and risk of gastric cancer in two population-based studies

BACKGROUND

Studies have shown increased gastric cancer risk in users of proton pump inhibitors (PPI) and histamine-2 receptor antagonists, questioning the safety of gastric acid suppression. Therefore, we conducted a case-control study within the Scottish Primary Care Clinical Informatics Unit (PCCIU) database and a cohort study in the UK Biobank.

METHODS

In PCCIU, five controls were matched to cases diagnosed in 1999-2011, and medications were determined from GP records. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using conditional logistic regression. In the UK Biobank, medications were self-reported at cohort entry 2006-2010, and gastric cancer ascertained from cancer registries until 2014. Hazard ratios (HR) were calculated using Cox regression.

RESULTS

PCCIU contained 1119 cases and 5394 controls. UK Biobank contained 250 cases in 471,779 participants. PPI users had a higher gastric cancer risk in PCCIU and UK Biobank when applying a 1-year lag (adjusted OR = 1.49, 95% CI 1.24, 1.80; adjusted HR = 1.28, 95% CI 0.86, 1.90, respectively), but these associations were attenuated when using a 2-year lag (adjusted OR = 1.13, 95% CI 0.91, 1.40; adjusted HR = 1.15, 95% CI 0.73, 1.82, respectively).

CONCLUSIONS

Overall, we observed little consistent evidence of an increased risk of gastric cancer with PPI use.

Incretin-based drugs and risk of lung cancer among individuals with type 2 diabetes

AIM

To assess whether dipeptidyl peptidase-4 inhibitors and glucagon-like peptide-1 receptor agonists are associated with an increased lung cancer risk among individuals with type 2 diabetes.

METHODS

We conducted a population-based cohort study using the UK Clinical Practice Research Datalink. We identified 130 340 individuals newly treated with antidiabetes drugs between January 2007 and March 2017, with follow-up until March 2018. We used a time-varying approach to model use of dipeptidyl peptidase-4 inhibitors and glucagon-like peptide-1 receptor agonists compared with use of other second- or third-line antidiabetes drugs. We used Cox proportional hazards models to estimate the adjusted hazard ratios, with 95% CIs, of incident lung cancer associated with use of dipeptidyl peptidase-4 inhibitors and glucagon-like peptide-1 receptor agonists, separately, by cumulative duration of use, and by time since initiation.

RESULTS

A total of 790 individuals were newly diagnosed with lung cancer (median follow-up 4.6 years, incidence rate 1.5/1000 person-years, 95% CI 1.4-1.6). Compared with use of second-/third-line drugs, use of dipeptidyl peptidase-4 inhibitors and glucagon-like peptide-1 receptor agonists was not associated with an increased lung cancer risk (hazard ratio 1.07, 95% CI 0.87-1.32, and hazard ratio 1.02, 95% CI 0.68-1.54, respectively). There was no evidence of duration-response relationships.

CONCLUSIONS

In individuals with type 2 diabetes, use of incretin-based drugs was not associated with increased lung cancer risk.

Medication Use and the Risk of Newly Diagnosed Diabetes in Patients with Epilepsy

Epilepsy is a common neurological disorder that affects millions of people worldwide. Patients with epilepsy generally require long-term antiepileptic therapy and many of them receive polypharmacy. Certain medications, including older-generation antiepileptic drugs, have been known to predispose patients to developing diabetes. Although data mining techniques have become widely used in healthcare, they have seldom been applied in this clinical problem. Here, the authors used association rule mining to discover drugs or drug combinations that may be associated with newly diagnosed diabetes. Their findings indicate in addition to the most common culprits such as phenytoin and valproic acid, prescriptions containing carbamazepine, oxcarbazepine, or lamotrigine may be related to the development of newly diagnosed diabetes. These mined rules are useful as guidance to both clinical practice and future research.

Use of Antidepressants and Risk of Cutaneous Melanoma: A Prospective Registry-Based Case-Control Study

Purpose

Melanoma is the cancer with the most rapidly rising incidence rate in Norway. Although exposure to ultraviolet radiation (UVR) is the major environmental risk factor, other factors may also contribute. Antidepressants have cancer inhibiting and promoting side effects, and their prescription rates have increased in parallel with melanoma incidence. Thus, we aimed to prospectively examine the association between use of antidepressants and melanoma by using nation-wide data from the Cancer Registry of Norway, the National Registry, the Norwegian Prescription Database and the Medical Birth Registry of Norway.

Patient and Methods

All cases aged 18–85 with a primary cutaneous invasive melanoma diagnosed during 2007–2015 (n=12,099) were matched to population controls 1:10 (n=118,467) by sex and year of birth using risk-set sampling. We obtained information on prescribed antidepressants and other potentially confounding drug use (2004–2015). Conditional logistic regression was used to estimate adjusted rate ratios (RRs) and 95% confidence intervals (CIs) for the association between overall and class-specific use of antidepressants and incident melanoma.

Results

Compared with ≤1 prescription, ≥8 prescriptions of antidepressants overall were negatively associated with melanoma (RR 0.81 CI 0.75–0.87). Class-specific analyses showed decreased RRs for selective serotonin reuptake inhibitors (RR 0.82 CI 0.73–0.93) and mixed antidepressants (RR 0.77 CI 0.69–0.86). The negative association was found for both sexes, age ≥50 years, residential regions with medium and highest ambient UVR exposure, all histological subtypes, trunk, upper and lower limb sites and local disease.

Conclusion

Use of antidepressants was associated with decreased risk of melanoma. There are at least two possible explanations for our results; cancer-inhibiting actions induced by the drug and less UVR exposure among the most frequent users of antidepressants.

Urinary tract infections and risk of squamous cell carcinoma bladder cancer: A Danish nationwide case-control study

Schistosoma haematobium infection can lead to squamous cell carcinomas (SCC) of the bladder. Whether this also applies to more common urinary tract infections (UTIs) is unclear. We therefore aimed to investigate the association between UTIs, reflected by the use of specific antibiotics and risk of SCC of the bladder. We conducted a Danish nationwide case-control study and identified histologically verified bladder cancer cases (2000-2015; n = 12,271) and age- and sex-matched cancer-free controls. We computed odds ratios (ORs) with 95% confidence intervals (CI) associating the use of UTI-specific antibiotics with SCC bladder cancer, using conditional logistic regression. We applied a 2-year lag-time to minimize reverse causation. To aid interpretation, similar analyses were performed for other bladder cancer types and other antibiotics. We identified 333 SCC cases (2.7% of all bladder cancers). Compared to no use (0-1 prescription), high-use (≥10 prescriptions) of UTI-specific antibiotics was associated with SCC with an OR of 11.4 (CI 7.6-17.2) and a clear dose-response pattern (p_trend < 0.001). Use of phenoxymethylpenicillin, an antibiotic not used against UTIs, was not associated with SCC after adjustment for use of UTI-specific antibiotics (OR 0.5). Furthermore, UTI-specific antibiotic use was not associated with urothelial carcinomas (n = 11,029; OR 1.13; CI 0.97-1.32). Excluding patients with known urogenital disease did not influence the SCC estimates (overall OR 10.8; CI 6.2-18.9). Data on smoking were lacking, however, a quantitative bias analysis suggested this to be of limited importance. In conclusion, common UTIs are strong, dose-dependent and specifically associated with risk of SCC of the bladder.

Study design choices for evaluating the comparative safety of diabetes medications: An evaluation of pioglitazone use and risk of bladder cancer in older US adults with type-2 diabetes

AIM

The aim of the study was to empirically demonstrate the effect of varying study designs when evaluating the safety of pioglitazone in treating bladder cancer.

METHODS

We identified Medicare beneficiaries above 65 years of age with diabetes between 2008 and 2015 and with classified exposure (at least two claims within 180 days) to glucose-lowering drugs (GLD), pioglitazone or another drug. The effects of varying the following study design parameters on bladder cancer risk were assessed: use of a new vs existing drug, choice of referent (all non-users and users of GLDs, non-insulin GLDs and DPP-4s) and whether or not censoring accounted for treatment change. We used the Cox proportional hazards model to obtain adjusted HRs and 95% CIs.

RESULTS

We included 1,510,212 patients classified as pioglitazone users (N = 135,188) or non-users (N = 1,375,024). Users had more diabetic complications than non-users, but fewer than insulin users. The HR ranged from 1.10 (1.01-1.20) to 1.13 (0.99-1.29) when censoring ignored treatment change, suggesting a weak association or none between pioglitazone and bladder cancer, probably under-estimating risk. However, the HR was 1.20 (1.01-1.42) when cohorts were restricted to new users, censored upon treatment change, and when DPP-4 was used as the referent, suggesting an increased risk of bladder cancer associated with pioglitazone.

CONCLUSIONS

The continued demand for new GLDs indicates the need for more robust observational methods to improve the value of generating real-world evidence in equipping clinicians to make informed prescribing decisions. Although there is no one-size-fits-all approach, we recommend active comparator new user study designs that compare therapeutically equivalent drugs and account for treatment changes during follow-up to present the least biased comparative safety estimates.

Statins and pancreatic cancer risk in patients with chronic pancreatitis: A Danish nationwide population-based cohort study

Statins (HMG-CoA reductase inhibitors) have antiinflammatory and possibly anticancer properties. We hypothesized that statin use is associated with lower risk of pancreatic cancer in patients with chronic pancreatitis. This nationwide population-based cohort study included all Danish patients diagnosed with incident chronic pancreatitis from 1 January 1996 to 31 December 2012. We used the Danish National Prescription Registry to ascertain information on statin prescriptions for members of the study population before and after their pancreatitis diagnosis. We computed crude incidence rates, incidence rate ratios (IRRs) and adjusted hazard ratios (HRs) with associated 95% confidence intervals (CIs) for pancreatic cancer, comparing statin users with nonusers. We computed HRs using Cox proportional hazards regression with statins treated as a time-varying exposure lagged by 1 year, adjusting for age, sex, socioeconomic status and individual comorbidities. The study included 8,311 chronic pancreatitis patients with a median age of 54 years. We observed 153 pancreatic cancers during 60,365 person-years of follow-up. The unadjusted IRR comparing statin users with nonusers was 1.00 (95% CI: 0.60-1.60). Adjustment for potential confounders only had a small impact on the estimate (adjusted HR: 0.90; 95% CI: 0.56-1.44). Our findings suggest that statin use is not associated with pancreatic cancer risk in patients with chronic pancreatitis.

Hypothyroidism and the risk of breast cancer recurrence and all-cause mortality - a Danish population-based study

Background

Hypothyroidism may occur as a late effect of breast cancer-directed treatment, particularly after radiotherapy, but little is known whether hypothyroidism affects the prognosis after breast cancer. We investigated the association between hypothyroidism and breast cancer recurrence, and all-cause mortality.

Methods

In this population-based cohort study, we used national medical registries to identify all Danish women 35 years or older diagnosed with stage I–III, operable breast cancer between 1996 and 2009. Hypothyroidism was defined as hospital diagnoses ascertained via diagnostic codes, or as prescriptions for levothyroxine. Two analytic models were used: (i) hypothyroidism present at the time of the breast cancer diagnosis (prevalent) and (ii) hypothyroidism diagnosed during follow-up as a time-varying exposure lagged by 1 year (incident). Breast cancer recurrence was defined as any local, regional, or distant recurrence or contralateral breast cancer. All-cause mortality included death from any cause in any setting. We used Cox regression models accounting for competing risks to compute adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) of breast cancer recurrence and all-cause mortality.

Results

The study cohort included 35,463 women with breast cancer with 212,641 person-years of follow-up. At diagnosis, 1272 women had hypothyroidism and 859 women developed hypothyroidism during follow-up. In total, 5810 patients developed recurrent breast cancer. Neither prevalent nor incident hypothyroidism was associated with breast cancer recurrence (adjusted HR_prevalent 1.01, 95% CI 0.87–1.19; adjusted HR_incident 0.93, 95% CI 0.75–1.16, respectively). Furthermore, no differences were seen for all-cause mortality for prevalent or incident hypothyroidism (adjusted HR_prevalent 1.02, 95% CI 0.92–1.14, and HR_incident 1.08, 95% CI 0.95–1.23, respectively). Stratification by menopausal status, oestrogen receptor status, chemotherapy, or radiotherapy did not alter the estimates.

Conclusions

Hypothyroidism present at diagnosis or during follow-up was not associated with breast cancer recurrence or all-cause mortality in women with breast cancer. Our findings provide reassurance to patients and their physicians that hypothyroidism is unlikely to impact on the clinical course of breast cancer or survival.

Electronic supplementary material

The online version of this article (10.1186/s13058-019-1122-3) contains supplementary material, which is available to authorized users.

Bias in pharmacoepidemiologic studies using secondary health care databases: a scoping review

BACKGROUND

The availability of clinical and therapeutic data drawn from medical records and administrative databases has entailed new opportunities for clinical and epidemiologic research. However, these databases present inherent limitations which may render them prone to new biases. We aimed to conduct a structured review of biases specific to observational clinical studies based on secondary databases, and to propose strategies for the mitigation of those biases.

METHODS

Scoping review of the scientific literature published during the period 2000-2018 through an automated search of MEDLINE, EMBASE and Web of Science, supplemented with manually cross-checking of reference lists. We included opinion essays, methodological reviews, analyses or simulation studies, as well as letters to the editor or retractions, the principal objective of which was to highlight the existence of some type of bias in pharmacoepidemiologic studies using secondary databases.

RESULTS

A total of 117 articles were included. An increasing trend in the number of publications concerning the potential limitations of secondary databases was observed over time and across medical research disciplines. Confounding was the most reported category of bias (63.2% of articles), followed by selection and measurement biases (47.0% and 46.2% respectively). Confounding by indication (32.5%), unmeasured/residual confounding (28.2%), outcome misclassification (28.2%) and "immortal time" bias (25.6%) were the subcategories most frequently mentioned.

CONCLUSIONS

Suboptimal use of secondary databases in pharmacoepidemiologic studies has introduced biases in the studies, which may have led to erroneous conclusions. Methods to mitigate biases are available and must be considered in the design, analysis and interpretation phases of studies using these data sources.

Cardiovascular, antidepressant and immunosuppressive drug use in relation to risk of cutaneous melanoma: a protocol for a prospective case-control study

INTRODUCTION

The incidence of cutaneous melanoma (hereafter melanoma) has increased dramatically among fair-skinned populations worldwide. In Norway, melanoma is the most rapidly growing type of cancer, with a 47% increase among women and 57% among men in 2000-2016. Intermittent ultraviolet exposure early in life and phenotypic characteristics like a fair complexion, freckles and nevi are established risk factors, yet the aetiology of melanoma is multifactorial. Certain prescription drugs may have carcinogenic side effects on the risk of melanoma. Some cardiovascular, antidepressant and immunosuppressive drugs can influence certain biological processes that modulate photosensitivity and immunoregulation. We aim to study whether these drugs are related to melanoma risk.

METHODS AND ANALYSIS

A population-based matched case-control study will be conducted using nation-wide registry data. Cases will consist of all first primary, histologically verified melanoma cases diagnosed between 2007 and 2015 identified in the Cancer Registry of Norway (14 000 cases). Ten melanoma-free controls per case (on date of case melanoma diagnosis) will be matched based on sex and year of birth from the National Registry of Norway. For the period 2004-2015, and by using the unique personal identification numbers assigned to all Norwegian citizens, the case-control data set will be linked to the Norwegian Prescription Database for information on drugs dispensed prior to the melanoma diagnosis, and to the Medical Birth Registry of Norway for data regarding the number of child births. Conditional logistic regression will be used to estimate associations between drug use and melanoma risk, taking potential confounding factors into account.

ETHICS AND DISSEMINATION

The project is approved by the Regional Committee for Medical Research Ethics in Norway and by the Norwegian Data Protection Authority. The study is funded by the Southeastern Norway Regional Health Authority. Results will be published in peer-reviewed journals and disseminated further through scientific conferences, news media and relevant patient interest groups.

Bisphosphonate use and risk of renal cell carcinoma: A population-based case-control study

The purpose of this study was to evaluate the association between the use of bisphosphonates and the risk of developing renal cell carcinoma (RCC). We conducted a case-control study in Denmark, using data linked from population-based health and administrative registries. We identified all cases of RCC from 1996 to 2013 and sampled population controls in a 10:1 ratio from the underlying population free of RCC, while matching on sex, birth year and calendar time. Bisphosphonate use before RCC diagnosis, excluding the year leading up to the diagnosis, was measured using outpatient prescription dispensations. We used conditional logistic regression to compute crude and adjusted odds ratios (ORs) comparing ever vs. never bisphosphonate use in doses indicated for treatment of osteoporosis, overall and stratified by sex, with the OR estimating the incidence rate ratio. We also examined the effects by cumulative dose and specific agent. There were 2748 RCC cases and 27 480 controls. The adjusted ORs for ever vs never bisphosphonate use were 1.07 (95% confidence interval: 0.94-1.22) overall; 1.15 (1.00-1.32) for women; and 0.78 (0.54-1.12) for men. Smoking could not be directly controlled for in the analysis. We found a weak association between use of oral bisphosphonates and risk of renal cell carcinoma in females. The observed association could be due to confounding by cigarette smoking, and future studies are required to assess this association further.

Antihistamine use and risk of ovarian cancer: A population-based case-control study

OBJECTIVE

Histamine is suggested to play a role in ovarian carcinogenesis. We examined the association between antihistamine use and ovarian cancer risk in a nationwide case-control study.

STUDY DESIGN

Cases (n = 5 556) comprised all women in Denmark aged 30-84 years with a histologically verified first diagnosis of epithelial ovarian cancer during 2000-2015. Age-matched population controls (n = 83 340) were selected using risk-set sampling. Data on prescription use, patient and demographic characteristics were retrieved from nationwide registries.

MAIN OUTCOME MEASURES

We used conditional logistic regression to estimate odds ratios (ORs) with 95% confidence intervals (CIs) for epithelial ovarian cancer associated with antihistamine use (≥2 prescriptions). The association was evaluated according to patterns of antihistamine use, menopausal status, and histological subtype of ovarian cancer.

RESULTS

Ever use of antihistamines was not associated with ovarian cancer overall (OR = 0.97, 95% CI = 0.90-1.05). The lack of association remained in subanalyses for patterns of antihistamine use. We observed an inverse association between antihistamine use and ovarian cancer among pre-menopausal women (<50 year: OR = 0.72, 95% CI = 0.57-0.90), but not post-menopausal women (≥50 year: OR = 1.02, 95%CI = 0.93-1.11). In analyses of histological subtypes, an inverse association emerged for mucinous ovarian cancer (OR = 0.74, 95% CI = 0.57-0.96), but not for other subtypes.

CONCLUSION

Antihistamine use was not associated with overall ovarian cancer risk. Additional research is needed to confirm inverse associations between antihistamine use and mucinous ovarian cancer, and overall ovarian cancer among pre-menopausal women.

Use of N-nitrosodimethylamine (NDMA) contaminated valsartan products and risk of cancer: Danish nationwide cohort study

OBJECTIVE

To perform an expedited assessment of cancer risk associated with exposure to N-nitrosodimethylamine (NDMA) through contaminated valsartan products.

DESIGN

Nationwide cohort study.

SETTING

Danish health registries on individual level prescription drug use, cancer occurrence, and hospital diagnoses.

PARTICIPANTS

5150 Danish patients with no history of cancer, aged 40 years or older, and using valsartan at 1 January 2012 or initiating use between 1 January 2012 and 30 June 2017. Participants were followed from one year after cohort entry (lag time period) until experiencing a cancer outcome, death, migration, or end of study period (30 June 2018). Each participant's exposure to NDMA (ever exposure and predefined categories of cumulative valsartan exposure) was mapped out as a time varying variable while also applying a one year lag.

MAIN OUTCOME MEASURES

Association between NDMA exposure and a primary composite endpoint comprising all cancers except non-melanoma skin cancer, estimated using Cox regression. In supplementary analyses, the risk of individual cancers was determined.

RESULTS

The final cohort comprised 5150 people followed for a median of 4.6 years. In total, 3625 cohort participants contributed 7344 person years classified as unexposed to NDMA, and 3450 participants contributed 11 920 person years classified as ever exposed to NDMA. With 104 cancer outcomes among NDMA unexposed participants and 198 among exposed participants, the adjusted hazard ratio for overall cancer was 1.09 (95% confidence interval 0.85 to 1.41), with no evidence of a dose-response relation (P=0.70). For single cancer outcomes, increases in risk were observed for colorectal cancer (hazard ratio 1.46, 95% confidence interval 0.79 to 2.73) and for uterine cancer (1.81, 0.55 to 5.90), although with wide confidence intervals that included the null.

CONCLUSIONS

The results do not imply a markedly increased short term overall risk of cancer in users of valsartan contaminated with NDMA. However, uncertainty persists about single cancer outcomes, and studies with longer follow-up are needed to assess long term cancer risk.