Associations between antihypertensive drugs and patterns of blood pressure changes: a strategy to reduce the burden of anti-VEGF induced hypertension

The primary endpoint for aim 1 is systolic blood pressure. Systolic blood pressure has been selected as primary end-point because 1) the available evidence from anti-VEGF studies indicates that increases in the prevalence of hypertension is primarily mediated by increases in systolic blood pressure; and 2) there is good evidence from randomized controlled trials and observational studies that systolic blood pressure, and pulse pressure, better predicts cardiovascular disease risk than diastolic blood pressure. Secondary endpoints are alternative blood pressure components (diastolic blood pressure, pulse pressure), heart rate, and variability in blood pressure.The primary endpoint for aim 2 is defined as a rise in systolic blood pressure of >10 mmHg in the first 60 days of anti-VEGF drug therapy (compared to baseline systolic blood pressure). An additional criterion is that the blood pressure must be >140/90 while on VEGF inhibitor therapy. The secondary endpoints are hypertension of any grade and grade 3 or 4 hypertension.The primary endpoint for aim 3 is proteinuria of any grade. Secondary endpoint is grade 3 or 4 proteinuria.The primary endpoint for aim 4 is overall survival. The secondary endpoint is progression free survival.Aim 1 will include all individuals using an anti-VEGF drug (pazopanib or sunitinib) that are treated (new anti-hypertensive drug or increased dose of existing anti-hypertensive drug) for hypertension while on treatment with an anti-VEGF drug. Multivariate linear regression will be used to model the first systolic blood pressure measurement recorded between 4 to 8 weeks following initiation of treatment for VEGF inhibitor induced hypertension. Systolic blood pressure at the time of initiation of treatment for hypertension will be included as a covariate. The effect of time of the blood pressure measurement will be assessed and adjusted for, if necessary. Causal effects of BP treatment for hypertension will be assessed by using propensity score analysis techniques with individuals' propensity for BP treatment being included as a covariate in the SBP outcome model which will also include the exposure of interest (BP treatment) and other potential confounders of SBP. Factors that may influence propensity for anti-hypertensive treatment selection include age, sex, ethnicity, history of hypertension, existing anti-hypertensive treatment, whether the intervention was a dose increase or addition of a new drug, pre-existing cardiovascular disease, chronic kidney disease and geographical region. Aim 2 will include all individuals using an anti-VEGF drug (pazopanib or sunitinib). The odds of a hypertension adverse event will be modelled using multivariate binary logistic regression with the primary covariate of interest being use of drugs with salutary effects on endothelial function (ACEI, ARB and CCB-based treatment regimens). Potential confounders of anti-VEGF drug induced hypertension will be included in the model. Specifically, baseline blood pressure, age, gender, ethnicity, pre-existing hypertension and cardiovascular disease, diabetes, chronic kidney disease, concomitant drugs such as current anti-hypertensive treatments [other than ACEI/ARB/CCB], NSAIDs, steroids, and other immunosuppressants, and geographical region. Whether preventative effects are dose dependent will be assessed in a secondary analysis along with alternative definitions of anti-VEGF drug induced hypertension.Aim 3 will include all individuals using an anti-VEGF drug (pazopanib or sunitinib). The odds ratio for proteinuria for individuals using an ACEI or ARB at baseline will be evaluated using multivariate binary logistic regression. The model will adjust for other baseline covariates that are associated with ACEI/ARB use and that may also be predictors of anti-VEGF drug mediated proteinuria (e.g. age, gender, ethnicity, pre-existing hypertension and cardiovascular disease, diabetes, chronic kidney disease, baseline blood pressure, geographical region, and concomitant drugs such as antihypertensive agents [other than ACEI/ARB], NSAIDs, steroids, and other immunosuppressants).Aim 4 will include all individuals using an anti-VEGF drug (pazopanib or sunitinib). The hazard ratio for overall survival for individuals using an ACEI or ARB at baseline will be evaluated using Cox proportional hazards regression. The model will adjust for potential confounders such as age, performance status, gender, ethnicity, pre-existing hypertension and cardiovascular disease, diabetes, chronic kidney disease, and geographical region.For all aims the handling of missing data will depend on the extent of missing data and the risk that it is missing not at random. If it can be assumed that the data is missing at random and less than 10% of the data is missing then a sensitivity analysis using complete cases will be performed. Where more than 10% of data is missing, a multiple imputation approach will be undertaken based on a predictive model for missingness utilising the other covariates available. All standard modelling assumptions will be tested for each analysis and data will be transformed and sensitivity to outliers assessed where necessary.PowerStudies available with individuals using pazopanib or sunitinib: COMPARZ (n=1110) and VEG105192 (n=290)Aim 1: Are ACEI/ARB/CCB drugs more effective at reducing BP for individuals with hypertension due to an anti-VEGF drug?Assuming that 20% x 1400 patients (n=280) on pazopanib/sunitinib develop hypertension requiring (new or more intensive) antihypertensive treatment. Assuming a 50%/50% split between antihypertensive drug groups, systolic blood pressure standard deviation of 15mmHg and a correlation between baseline and post-treatment systolic blood pressure of 0.6, we will have 94% power to detect a difference of 5mmHg.Aim 2: Does ACEI/ARB/CCB use influence risk of hypertension with use of an anti-VEGF drug?Assuming there is a 29%/71% split [J Clin Oncol 32, 2014 (suppl 4; abstr 437)] between the groups using an ACEI/ARB/CCB (n=406) and not using an ACEI/ARB/CCB (n=994), and a baseline risk of developing hypertension of 45% [The Oncologist 2013; 18:273-280] in the group not using an ACEI/ARB/CCB, we would have 93% power to detect a significant decrease in risk to 35% (i.e. relative risk of 0.78) for the group using an ACEI/ARB/CCB.Aim 3: Does ACEI/ARB use influence risk of proteinuria with use of an anti-VEGF drug?Assuming there is a 29%/71% split between the groups using an ACEI/ARB (n=406) and not using an ACEI/ARB/CCB (n=994), and a baseline risk of developing proteinuria (any grade) of 20% [N Engl J Med 2013;369:8] in the group not using an ACEI/ARB, we would have 87% power to detect a significant decrease in risk to 13% for the group using an ACEI/ARB/CCB (i.e. relative risk of 0.65).Aim 4: Is ACE/ARB use prognostic of survival for individuals using an anti-VEGF drug?Assuming there is a 29%/71% split between the groups using an ACEI/ARB (n=406) and not using an ACEI/ARB (n=994), and an event rate of 50% [N Engl J Med 2013;369:8] over follow-up for the non ACEI/ARB group, we would have 88% power to detect a hazard ratio of 0.75 for individuals using an ACEI/ARB [HR of 0.74 observed in previous study]. Power will be greater if updated survival data with a greater number of events is now available.