Hypertension, antihypertensive treatment and cancer incidence and mortality

a pooled collaborative analysis of 12 Australian and New Zealand cohorts

Abstract

Background: 

Observational studies examining associations between hypertension and cancer are inconsistent. We explored the association of hypertension, graded hypertension and antihypertensive treatment with cancer incidence and mortality.

Method: 

Eighty-six thousand five hundred and ninety-three participants from the Australian and New Zealand Diabetes and Cancer Collaboration were linked to the National Death Index and Australian Cancer Database. Cox proportional hazards models estimated hazard ratios and 95% confidence intervals (95% CI) for the association of treated and untreated hypertension with cancer incidence and mortality.

Results: 

Over a median follow-up of 15.1 years, 12 070 incident and 4350 fatal cancers were identified. Untreated and treated hypertension, compared with normotension, were associated with an increased risk for cancer incidence [hazard ratio 1.06, 95% CI (1.00–1.11) and 1.09 (1.02–1.16) respectively], and cancer mortality (1.07, 0.98–1.18) and (1.15, 1.03–1.28), respectively. When compared with untreated hypertension, treated hypertension did not have a significantly greater risk for cancer incidence (1.03, 0.97–1.10) or mortality (1.07, 0.97–1.19). A significant dose–response relationship was observed between graded hypertension and cancer incidence and mortality; P_trend = 0.053 and P_trend = 0.001, respectively. When stratified by treatment status, these relationships remained significant in untreated, but not in treated, hypertension.

Conclusion: 

Hypertension, both treated and untreated, is associated with a modest increased risk for cancer incidence and mortality. Similar risks in treated and untreated hypertension suggest that the increased cancer risk is not explained by the use of antihypertensive treatment.

BACKGROUND

Metabolic abnormalities related to obesity, for example hyperglycemia and hypertriglyceridemia, have been linked to an increased risk for cancer ^[1,2]. Emerging evidence suggests that other metabolic abnormalities, such as hypertension, may also be associated with an increased risk for cancer ^[3–6]. Two main hypotheses have been proposed to explain this observation. First, medications used in the treatment of hypertension may cause cancer by directly promoting carcinogenesis, accelerating other carcinogens or impeding defense mechanisms ^[7]. Second, there may be a common mechanism linking blood pressure (BP) regulation and cancer development, independent of antihypertensive treatment ^[8]. In addition, it is also possible that cancer itself may lead to hypertension ^[9].

To date, observational studies on the association between hypertension and cancer show inconsistent results. A large Swedish cohort reported a 7% [95% confidence interval (CI): 4–9%)] increase in incident cancer risk per 10 mmHg increase in BP ^[10]. This association, however, did not account for antihypertensive treatment, and was observed only in men ^[10]. A second, smaller, Swedish cohort of men demonstrated a 41% increase in risk of cancer incidence when comparing the highest quintile of SBP with the lowest, even after adjustment for antihypertensive medication at baseline ^[8]. Other studies report that antihypertensive medication, in particular diuretics, may be associated with an increased cancer risk even in those who are normotensive ^[11,12]. However, a recent meta-analysis concluded that diuretics or any other single antihypertensive medication was not associated with an increased cancer risk, though it could not be ruled out that a combination of antihypertensive drugs may confer an increased risk for overall cancer ^[13].

For cancer mortality, evidence is less clear. Early work by Dyer et al.^[3] showed a 50 and 80% increased risk in cancer mortality among hypertensive men compared with normotensive men for systolic and diastolic hypertension, respectively. A more recent meta-analysis reported a 23% increased risk in cancer mortality in those with hypertension compared with those without, though significant heterogeneity among studies was noted and the included studies did not adjust for BP treatment ^[4].

Using data from a large pool of prospective studies, we examine the association between treated and untreated hypertension, graded hypertension and cancer incidence and mortality.

METHODS

Study population

The Australian and New Zealand Diabetes and Cancer Collaboration (ANZDCC) is a pooled study composed of 18 prospective studies in Australia and New Zealand with data on 153 025 men and women. Details of sampling procedures, study designs and methods for each of the respective studies have been described elsewhere ^[14]. In brief, the chief investigators of epidemiological studies in Australia and New Zealand conducted from 1983 onwards with data on diabetes, obesity and the metabolic syndrome and with a minimum sample size of 1000 individuals were invited to participate in the ANZDCC study. For the current analysis, we included observational studies that had measured BP and information on antihypertensive treatment (12 cohorts; n = 91 781). To prevent reverse causation, we excluded participants with a cancer diagnosis prior to baseline date (n = 3022) or who were underweight at baseline (n = 843). We further excluded participants with missing data on age, sex, smoking or BP (n = 1323). A total of 86 593 participants (40 965 men; 45 628 women) with complete data were included in the final data analysis.

Data linkage

Australian participants of the ANZDCC cohort were linked to the Australian Cancer Database, a register of all primary, malignant cancers diagnosed in Australia since 1982, and the National Death Index (NDI). Linkage was performed by the Australian Institute of Health and Welfare (AIHW) and the Western Australian Data Linkage Unit (for Fremantle Diabetes Study (FDS) only), using first name, second name, last name, sex and date of birth ^[15]. New Zealand participants of the ANZDCC cohort, those from the Fletcher Challenge Study, were linked to the New Zealand Cancer Registry and Mortality Database using National Health Index numbers; unique identifiers assigned to every person who uses health and disability support services in New Zealand. Cancer status of the cohort was determined until 31 December 2008 for 10 of the 12 cohorts and until 31 August 2010 for the Melbourne Collaborative Cohort Study (MCCS) and 31 October 2012 for FDS. Mortality status was determined until 31 March 2012 for nine of the 12 cohorts, and until 31 December 2009 for the Fletcher Challenge Study, 30 April 2011 for MCCS and 31 January 2013 for FDS. We set a match link rate of 97.70% (true matches/correct links) with link accuracy of 97.92% (1.08% expected to be false-positive links). Twenty-seven percent of links underwent clerical review, performed by AIHW.

Definition of covariates and outcomes

Details of baseline BP measurements, by study, are provided in Supplementary Table 1, https://links.lww.com/HJH/A541. Untreated hypertension was defined as elevated SBP or DBP of at least 140 or 90 mmHg, respectively, with no self-reported use of antihypertensive treatment, and treated hypertension was defined by self-reported use of antihypertensive treatment; no information on specific antihypertensive drugs was available. Those who did not have elevated SBP or DBP and who did not report using antihypertensives were defined as ‘normotensive’. ‘Any hypertension’ was defined as untreated or treated hypertension. Graded hypertension was classified according to the European Society of Hypertension and the European Society of Cardiology guidelines as follows: normal: SBP <130 mmHg or DBP <85 mmHg; high normal: SBP 130–139 mmHg or DBP 85–89 mmHg; graded hypertension 1: SBP 140–159 mmHg or DBP 90–99 mmHg; graded hypertension 2: SBP 160–179 mmHg or DBP 100–109 mmHg; graded hypertension 3: SBP ≥180 mmHg or DBP≥110 mmHg ^[16].

Biomedical tests collected data on cholesterol (mmol/l); diabetes was defined by self-report, fasting plasma glucose of at least 126 mg/dl (7.0 mmol/l) or use of glucose-lowering medication; BMI was computed as weight (kg) divided by the square of height (m). Information on education, smoking status and alcohol status was collected by questionnaires. These risk factors were harmonized across studies to reflect common categories as follows: smoking [current, ex-smoker and never-smoker (ex and current smokers were combined into a single category of ‘ever’ smokers)]; education (high school or lower, above high school); alcohol (meeting guidelines of ≤ two drinks on any occasion or not meeting guidelines) ^[17]. Total physical activity time was calculated as the sum of time spent walking (if continuous and for >10 min) or performing moderate-intensity activity, and double the time spent in vigorous-intensity activity. This double weighting has been used because of the need to reflect that participation in vigorous intensity physical activity confers greater health benefits than participation in moderate activity ^[18]. Participants were then categorized as meeting Australian physical activity guidelines (≥150 min/week) or not meeting guidelines (≥0 min/week and <150 min/week) ^[19]. Cancer was defined according to the International Classification of Disease 10th Revision (ICD-10); C00-C97, D45-D46 and D47.1-D47.3 with cancers coded according to ICD-9 recoded as appropriate. Incidence of cancer was defined as the first occurrence of cancer or death from cancer if that was the first time the cancer had been reported. Fatal cancer includes all those who died from cancer as reported by the NDI.

Statistical analysis

For outcomes of cancer incidence (fatal and nonfatal cancer), individuals were followed from baseline date to date of cancer diagnosis, date of death or date of linkage to cancer registries, whichever occurred first. For outcomes of cancer mortality, individuals were followed from baseline date to date of death or date of linkage to mortality registries, whichever came first.

Differences in baseline characteristics, by cancer outcome, were assessed using Pearson's chi-square test for proportions and Student's t-test for means as appropriate. Heterogeneity of studies was explored by conducting a meta-analysis using a random effects model and statistical heterogeneity was estimated by the I² statistic ^[20]. Cox proportional hazards models (with age as the time scale) were used to compute hazard rate ratios and 95% confidence intervals (95% CI) of cancer incidence and mortality associated with untreated and treated hypertension, any hypertension and graded hypertension, as defined above. Analyses of graded hypertension were performed on the total population, and then stratified by antihypertensive use. Proportional hazards assumptions were satisfied, as assessed with graphs of log–log plots of the relative hazards by time for discrete variables and by scaled Schoenfeld residuals. All models used age as the time scale and were adjusted for sex, smoking and study cohort whereby baseline hazards for each included study were adjusted for. Additional models were adjusted for education, diabetes status, BMI, physical activity, cholesterol and alcohol intake where data were available.

All analyses were done using STATA version 12.1 (StataCorp, College Station, Texas, USA). This study was approved by the Alfred Health Human Research and Ethics Committee (HREC), the AIHW HREC and the Western Australian Department of Health HREC.

RESULTS

Over a median follow-up of 15.1 years, 12 070 incident and 4350 fatal cancers were identified over 1 128 742 and 1 317 274 person-years, respectively. Baseline characteristics of the study population, by cancer outcome, are shown in Table 1. In brief, those who developed cancer were more likely to be men, older, less educated, ever smokers, hypertensive, diabetic, and have higher mean BMI, cholesterol, SBP and DBP values compared with those who did not develop cancer.

TABLE 1: Baseline characteristics among the Australian and New Zealand Diabetes and Cancer Collaboration cohort which developed an incident cancer during follow-up compared with those that did not

Baseline characteristics, by study, are shown in Supplementary Tables 2A and 2B, https://links.lww.com/HJH/A541. The proportion of participants with hypertension (treated or untreated) in each study cohort ranged from 27.4% in the Fletcher Challenge Study to 75.9% in the Dubbo Study of the Elderly. There was no significant heterogeneity across studies for the relationship between hypertension and cancer incidence: I² = 24.1 and 41.2% for untreated hypertension and treated hypertension, respectively.

Those with untreated and treated hypertension at baseline were more likely to develop cancer during follow-up compared with those who were normotensive in fully adjusted models: hazard ratio: 1.09, 95% CI (1.02–1.16) and 1.06 (1.00–1.11), respectively (Table 2). There was no difference in risk for cancer among those with treated hypertension compared with untreated hypertension in fully adjusted models (1.03, 0.97–1.10). Compared with normotensive, any hypertension had a 1.06 (1.02–1.12) increased risk for cancer incidence (Table 2).

TABLE 2: Hazard ratios and 95% confidence intervals for the association between hypertension and cancer incidence and mortality

Treated hypertension and untreated hypertension had an increased risk for cancer mortality compared with those who were normotensive in fully adjusted models (1.07, 0.98–1.18) and (1.15, 1.03–1.28), respectively, though this reached statistical significance only in those with treated hypertension (Table 2). When compared with untreated hypertension, treated hypertension had similar risks for cancer mortality (1.07, 0.97–1.19) in fully adjusted models. Compared with normotensive, any hypertension had a 1.10 (1.01–1.20) increased risk for cancer mortality.

A dose–response relationship was observed between graded hypertension and cancer incidence in fully adjusted models, P_trend = 0.053 (Table 3). The highest risk estimates were seen in those with SBP 160–179 mmHg or DBP 100–109 mmHg: 1.08 (1.00–1.17). When stratified by antihypertensive status, this relationship remained significant in those with untreated hypertension, P_trend = 0.015, but was attenuated in treated hypertension, P_trend = 0.258. A dose–response relationship was also observed between graded hypertension and cancer mortality, P_trend = 0.001 with highest risk estimates seen in those with SBP 160–179 mmHg or DBP 100–109 mmHg, 1.36 (1.16–1.59). When stratified by treatment status, this dose–response relationship remained significant in untreated hypertension (P_trend = 0.002), but not in treated hypertension (P_trend = 0.543).

TABLE 3: Hazard ratios (95% confidence interval) for the association between graded hypertension and cancer incidence and mortality

DISCUSSION

In this large prospective pooled cohort, we found a significant but modest increased risk for cancer incidence among those with treated and untreated hypertension compared with those who were normotensive. For cancer mortality, untreated and treated hypertension had similar risk estimates, but this reached statistical significance in those with treated hypertension only. We found no difference in risk for cancer incidence or mortality in treated hypertension as compared with untreated hypertension. For graded hypertension, we found that those with SBP 160–179 mmHg or DBP 100–10 mmHg had an 8 and 36% increased risk for cancer incidence and mortality, respectively, compared with the lowest grade of hypertension. This association was independent of treatment status with elevated risks observed in both untreated and treated hypertension, though these were significant only in untreated hypertension.

Comparison to literature

Our findings are consistent with other observational data that have found a positive association between hypertension and cancer incidence. Stock et al.^[10] found a 29% increased risk for cancer among men and women in the highest quartile of BP compared with the lowest quartile, though this study did not account for antihypertensive treatment which may explain the higher effect size relative to our study. Our analyses of graded hypertension showed that those with SBP 160–179 mmHg or DBP 100–109 mmHg had an 8% increased risk for cancer relative to those in the lowest category. These estimates are, however, lower than those reported by a Swedish study of over 7000 men which found a 41% increased risk for cancer incidence among men in the highest quintile of BP compared with those in the lower quintile, even after adjustment for antihypertensive treatment ^[8]. However, this study recruited men referred to a specialist hypertension clinic and therefore it is expected that this population would have a higher risk estimate than a general population sample. For cancer mortality, our finding of a 7–15% increased risk for untreated and treated hypertension is similar to that by Goldbourt et al.^[21] of a 10% increased risk for men with than without hypertension. We also noted a 23% increased risk for mortality in those with the highest grade of hypertension as compared with the lowest, similar to the overall estimate of 23% found in the 2002 meta-analysis by Grossman et al.^[4]. This meta-analysis, however, was unable to account for the role of antihypertensive medication and the authors could not rule out publication bias.

The underlying mechanisms between hypertension and increased cancer risk are not clear. In animal models, dysregulation of apoptosis, induced by high BP, has been shown to promote the growth of cancer cells ^[22]. In addition, hormones, which play a role in the development of hypertension, possess mitogenic effects ^[23,24] and it is also possible that the association between hypertension and cancer is because of shared risk factors such as genetics, obesity, smoking and poor diet ^[8]. In our study, we adjusted for potential confounding effects of BMI, smoking, physical activity, cholesterol, alcohol and diabetes, but we were unable to explore further the roles of poor diet and/or genetics. Alternatively, antihypertensive medication may increase cancer risk, though findings are conflicting. A 2001 meta-analysis found an independent association between thiazide diuretics and an increased risk for cancer (hazard ratio 2.00, 95% CI 1.55–2.59), though this finding was not supported in a second meta-analysis which concluded that no single antihypertensive class has sufficient or consistent evidence for a significant increase in malignancy risk, including thiazide diuretics ^[13,25]. Our results are consistent with the latter finding insofar as cancer risk was similar in both treated and untreated hypertension groups.

Strengths and weaknesses

Our study combined data from 12 large population-based studies from Australia and New Zealand, with sufficient power to investigate the association between hypertension and risk of cancer incidence and mortality, by hypertension treatment status. There are several limitations to this study that should also be acknowledged. First, it is possible that the relationship between hypertension and cancer differs by cancer site with studies showing increased risks specifically for renal cancers among those with hypertension ^[26]. Examining site-specific cancers was beyond the scope of the pooled cohort. Second, we did not have information about the type or dose of antihypertensive medications that participants were taking and therefore could not explore the role of specific antihypertensive treatments on the development of cancer and subsequent mortality. Further, we only had data on hypertension and antihypertensive use at baseline. It is possible that the results in untreated patients might be because of unmeasured confounding by antihypertensive drug intake during follow-up and lack of compliance to therapy. Third, it is possible that our results of graded hypertension, when stratified by antihypertensive status, may be the result of a type II error whereby we were underpowered to detect true associations among those with treated hypertension in the highest categories of graded hypertension. Last, the advantage of pooled cohort studies is the increase in study power which allows meaningful analyses of outcomes such as cancer. However, limitations of pooled cohorts include the heterogeneity of included studies and the loss of discrimination of covariates during harmonization across studies. In the current study, covariates of physical activity and education were each collapsed into binary categories to provide consistency across studies. All other covariates were objectively measured (e.g. cholesterol BP, height and weight) and therefore more easily combined across studies. In addition, the level of heterogeneity was low (I² = 24.1–41.2%) and the magnitude of associations between hypertension and cancer was similar between studies. Therefore, we believe that it is unlikely that this harmonization process had a significant impact on our results.

In conclusion, this large pooled collaborative study suggests that treated and untreated hypertension are modestly associated with a higher risk for the development of cancer and subsequent cancer mortality in men and women in Australia and New Zealand. For men and women in the highest grade of hypertension, we additionally show that there is an increased cancer risk that is not explained by the use of antihypertensive treatment. Our findings are of public health importance insofar as both hypertension and cancer are common and potentially preventable conditions.

ACKNOWLEDGEMENTS

We thank all the study participants and the personnel who collected and/or handled the data in the cohorts. The authors also wish to thank the staff at the Australian Institute of Health and Welfare (AIHW) and the Western Australian Data Linkage Branch.

Authors’ contributions: J.L.H. (Baker IDI Heart and Diabetes Institute, Monash University) wrote the article, had full access to all the data and conducted the analyses. M.S. (Baker IDI Heart and Diabetes Institute, Monash University) contributed to data analysis and reviewed/edited the article. J.E.S. (Baker IDI Heart and Diabetes Institute, Monash University) and D.J.M. (Baker IDI Heart and Diabetes Institute, Monash University) contributed to conceptualization, discussion and reviewed/edited article. D.J.M. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the collaboration and reviewed the article and approved the final version.

Funding: This work is funded by a National Health and Medical Research Council (NHMRC grant APP1002663) and Australian Government Department of Health. The work is supported, in part, by the Victorian Operational Infrastructure Scheme.

Conflicts of interest

There are no conflicts of interest.