Incidence of New-Onset Hypertension in Cancer Patients: A Retrospective Cohort Study

This retrospective cohort study was conducted to estimate incidence rates of new-onset hypertension in adult cancer patients identified from the Varian Medical Oncology outpatient database. Incidence rates of increasing levels of hypertension severity were calculated overall and for periods of chemotherapy exposure and nonexposure. Cox models sought predictors of new-onset hypertension severity among baseline and chemotherapy exposure variables. New-onset hypertension was observed in about one-third of 25,090 patients with various cancer types. The incidence rates (IR) of severe and crisis-level hypertension, respectively, were the highest in patients with gastric (18.5 cases per 100 person-years (PY), 5.6 per 100 PY) and ovarian cancer (20.2 per 100 PY, 4.8 per 100 PY). The highest IR of moderate hypertension was observed in patients with renal cancer (46.7 per 100 PY). Across all cancers, chemotherapy exposure was associated with a 2–3.5-fold increase in risk of any degree of hypertension compared to periods of no chemotherapy; higher hypertension levels showed greater variability in relative risks by type and line of therapy but indicated an overall increase associated with chemotherapy exposure. These results help to elucidate the factors influencing HTN among cancer patients and the incidence of HTN relative to chemotherapy exposure.

New-onset HTN has been reported as an adverse event for numerous cancer therapies, with risk estimates regularly approaching 10% and in certain instances as high as 36% [14][15][16][17]. These new-onset HTN rates vary greatly by cancer treatment type and dose, and increased risk of developing high-grade HTN at severe or crisis levels has been observed with more recent cancer therapies, such as those that target vascular endothelial growth factor (VEGF), known as anti-VEGF agents [18,19]. The mechanism(s) by which different chemotherapy agents induce HTN is not fully understood. Potential mechanisms include vascular rarefaction (decrease in microvessel density), decreased sex hormones leading to impaired vasodilator and potentiated vasoconstrictor effects, and endothelial dysfunction causing an interference with nitric oxide (NO) signaling and thereby an increase in oxidative stress [15,16,20].
As HTN is a risk factor for coronary heart disease, stroke, heart failure, and end-stage renal disease [20], an improved understanding of HTN prior to starting chemotherapy, newonset HTN during and after chemotherapy, and factors influencing HTN among cancer patients is paramount. Prior studies have focused on the prevalence of HTN in cancer patients 2 International Journal of Hypertension rather than incidence [4,[8][9][10][11][12][13], but the incidence of newonset HTN is an important clinical outcome that requires investigation. Using an electronic medical record (EMR) database of outpatient oncology practices, the present study uses sequential blood pressure (BP) measurements to identify new-onset HTN of different severity levels, allowing an estimation of incidence of HTN in cancer patients with solid tumors that are highly vascularized and/or are dependent on VEGF stimulation for growth.

Data
Source. This retrospective cohort study was conducted using the Varian Medical Oncology EMR database, which contained data on more than 185,000 cancer patients at the time of data extraction for this study. The database included information from 18 outpatient oncology practices in 15 states across the United States. During each patient visit to the clinic, the clinic staff enters data into the database about the visit, including diagnoses, treatments, and other relevant information. Treatment data include orders or prescriptions for antineoplastic agents, with specific information such as dose and route, as well as duration of supply of oral medications and amount and timing of drugs administered in the clinic. Blood pressure (BP) measurement data are entered separately for systolic Blood pressure (SBP) and diastolic blood pressure (DBP) and include the date of the test and measurement result in mm Hg. The data used for the present study were deidentified, as required by the Health Insurance Portability and Accountability Act (HIPAA).

Study
Population. The study cohort included adult patients (age ≥18 years) with any of 11 cancer types of interest-breast, lung, colorectal, head and neck, gastric, ovarian, cervical, renal, melanoma, prostate, and connective and other soft tissue-diagnosed between January 1, 2000, and April 30, 2008. The International Classification of Diseases, 9th Revision (ICD-9) diagnosis codes for these cancer types are given in Table 1. Patients with multiple primary cancer diagnoses were excluded from the study as were patients with unknown age or gender.
The index date was defined as the first date of entry of a qualifying ICD-9 cancer diagnosis code. The period from 30 days prior to the index date through the index date was defined as the baseline period, and the period from the day after the index date to the last recorded BP measurement on or before the study end date of July 31, 2008, was defined as the follow-up period. To be eligible for inclusion in the study, patients were required to be normotensive or prehypertensive at baseline (i.e., SBP ≤ 150 mm Hg and DBP ≤ 100 mm Hg) and to have at least 1 BP measurement during the follow-up period.

Outcome Definition and Measures.
The study outcome was new-onset hypertension separated into three mutually exclusive categories of moderate, severe, and crisis-level hypertension. The BP ranges for each hypertension outcome category were based on definitions provided by the Seventh Report of the Joint National Committee of Prevention, Detection, Evaluation, and Treatment of High Blood Pressure [21].
(iii) Crisis-level hypertension was defined as either of the following: (1) follow-up SBP increased to >180 mm Hg, or (2) follow-up DBP increased to >120 mm Hg.
The time to first occurrence of each level of hypertension was calculated as number of days from the index date to the date of the first hypertensive BP measurement at that level. The same patient was eligible for each of the three hypertension outcomes.

Chemotherapy Exposure Definition and Measures.
For each eligible patient, all person-time of followup after the index date was divided into periods of time with and without chemotherapy exposure. Chemotherapy exposure comprised the periods of one or more chemotherapy regimens consisting of combinations of cytotoxic and/or targeted chemotherapy agents administered orally or through IV/injection. The time periods outside chemotherapy regimens were considered the periods with no chemotherapy exposure.
All chemotherapy agents administered through IV or injection were assigned an effective duration of 30 days. The duration of oral chemotherapy agents was determined by a combination of available variables for prescribed days' supply, dispensed quantity, administration frequency, and number of refills.
The specific agents in a chemotherapy regimen were determined by all cytotoxic or targeted chemotherapy agents, International Journal of Hypertension 3 either prescribed orally or administered through IV or injection at the clinic, during the first 8 days of the regimen. The chemotherapy regimen continued while only these agents were administered. If a new drug was given on day 9 or later within a regimen, the first regimen was considered to have ended and a new regimen begun. This new regimen consisted of all chemotherapy agents administered from the start date of the first drug through the next 8 days. This cycle of assigning chemotherapy agents to regimens continued until all chemotherapy days were assigned to a regimen. Patients were allowed to stop receiving a chemotherapy agent without ending the regimen, as long as at least 1 agent in the regimen continued without the addition of any new chemotherapy agents. Substitution of a chemotherapy agent with another in the same class but with generally higher tolerability (substituting carboplatin for cisplatin or epirubicin for doxorubicin) did not start a new chemotherapy regimen. The reverse of switching to the less tolerable agent was considered a new regimen.
Gaps of up to 14 days between the end of a chemotherapy administration and the start of the next administration within a regimen were ignored. If a gap between chemotherapy administrations was 15 to 89 days, then the gap was considered to be an off-chemotherapy break in the regimen, but the same line of therapy resumed when the chemotherapy administrations resumed. If a gap lasted for 90 or more days, then restarting the same chemotherapy agents was considered to be a new line of therapy.

Covariate Definition and Measures.
Covariates included patient demographics (age and gender), year of diagnosis, cancer type, geographic region, clinic type, health insurance type, line of chemotherapy, and type of chemotherapy regimen. Patient age was calculated as the number of years from the year of birth to index date and categorized into groups of 18-<30, 30-<50, 50-<65, and ≥65 years. The year of the first qualifying cancer diagnosis was categorized into intervals of 2000-2002, 2003-2005, and 2006-2008. Cancer type was determined by ICD-9 codes and was categorized into breast, lung, colorectal, head and neck, gastric, ovarian, cervical, renal, melanoma, prostate, and soft tissue sarcoma.

Data Analysis.
All data management and analyses were conducted using SAS version 9.2. Incidence rates of each level of hypertension severity were calculated both overall and separated into the periods of chemotherapy exposure and no chemotherapy exposure. The overall incidence rates were computed as the number of patients with each level of new-onset hypertension divided by the sum of all follow-up person-time through the latest available BP measurement, ending at the date of event for the patients who developed each level of hypertension severity.
Time to first occurrence of each level of hypertension severity was calculated for each patient and summarized with the median, mean, and standard deviation (SD) among patients who developed hypertension. The BP measurement at the first occurrence of each level of hypertension severity was summarized with the median and mean (SD). The incidence rates of hypertension severity during chemotherapy were calculated as the number of patients with each newonset hypertension level divided by the sum of all followup intervals of time on chemotherapy in which there were BP measurements. Follow-up time for each on-chemotherapy interval was truncated at the last BP measurement during the interval. For patients with each new-onset hypertension level during followup, follow-up time was truncated as of the date of the event. Parallel rules for each level of hypertension severity were used to calculate the incidence rates during the intervals without chemotherapy exposure.
Cox proportional hazards regression analyses were performed to identify predictors of each level of new-onset hypertension among the baseline characteristics, including patient age and gender, year of index date, cancer type, geographic region, clinic type, health insurance type, and baseline BP measurement. For the moderate and severe hypertension models, to avoid censoring patients whose first elevated BP fell into one of the more severe categories, the outcomes were modeled as nonmutually exclusive outcomes of (1) moderate, severe, or crisis-level hypertension and (2) severe or crisis-level hypertension. Chemotherapy exposure, including both line and type of chemotherapy, was entered into the models as time-varying covariates, with no chemotherapy exposure used as the reference level.

Study Population.
A total of 38,940 adult cancer patients had an initial diagnosis of 1 of the 11 cancer types of interest between January 1, 2000, and April 30, 2008. Of these patients, 30,682 had a BP measurement at baseline; 25,090 had only normal BP measurements during the baseline period and at least 1 BP measurement during the follow-up period, hence qualifying for the study.
Of the 25,090 patients included in the analyses (Table 2), 64.5% were female, and the mean age was 61 years. The most common cancer type was breast (36.1%), and most of the patients (88.6%) had their first cancer diagnosis between 2003 and 2008. The majority of patients (70.7%) were classified as having systolic BP of 120-150 mm Hg and/or diastolic BP of 80-100 mm Hg.

Time to First
Occurrence of Hypertension. The times to first occurrence of a hypertensive BP measurement for each level of hypertension severity are given in Table 3. The median number of days to the first moderate hypertension event was 96 days, increasing to 122 days for severe hypertension and 183 days for crisis-level hypertension. Elevated SBP was more common and occurred earlier than elevations in DBP for each level of hypertension severity. The mean SBP level at first occurrence of moderate hypertension was 155 mm Hg (SD: 3.1), and the mean DBP level at the first occurrence of moderate hypertension was 104 mm Hg (SD: 2.9). These levels increased to 168 mm Hg (SD: 5.5) SBP and 114 mm Hg (SD: 3.0) DBP for severe hypertension and 188 mm Hg (SD: 7.6) SBP and 131 mm Hg (SD: 10.8) DBP for crisis-level hypertension.  For all levels of severity, the incidence of hypertension was considerably higher during periods of chemotherapy exposure than while patients were not exposed to chemotherapy. The overall incidence rates of moderate hypertension during chemotherapy and during periods of no chemotherapy were 90.07 (CI: 72.43-110.70) and 20.89 (12.92-31.97) cases per 100 PY, respectively. A similar pattern of higher rates during chemotherapy exposure was seen for severe and crisis-level hypertension, although the rates were progressively lower in the latter 2 groups, with rates in the periods during and without chemotherapy exposure, respectively, of 40.21 and 9.64 cases per 100 PY for severe hypertension and 8.98 and 2.09 cases per 100 person-years for crisis-level hypertension.
One caveat to these results is that the monitoring of the patients' BP occurred more frequently during the periods when they were being treated with chemotherapy, with means of 34 measurements per person-year on chemotherapy and 6 per person-year off chemotherapy. This difference in the number of available measurements during the different periods of chemotherapy exposure allows greater opportunity to find an elevation during chemotherapy exposure.
Rates of each level of hypertension varied considerably across tumor types (Table 4). Patients with melanoma, breast cancer, and connective and other softtissue cancers had the lowest incidence of hypertension at each of the severity levels. Moderate hypertension was most commonly seen in renal and lung cancers; severe in ovarian, gastric, prostate, and lung cancers; and crisis-level hypertension in gastric, ovarian, lung, and colorectal cancers. Within each tumor type, the finding of higher rates of each level of hypertension associated with periods of chemotherapy use compared to nonuse held constant. For the most severe (crisis-level) hypertension outcome, the highest rate observed was 19.54 (11.86-30.33) cases per 100 PY among renal cancer patients during chemotherapy exposure, and the lowest rate of this outcome was 1.26 (0.06-6.01) cases per 100 PY in connective and other softtissue cancers during periods of no chemotherapy use.

Predictors of Hypertension.
Results of separate Cox proportional hazards models of moderate or higher, severe or higher, and crisis-level hypertension revealed an increased risk in all models associated with increasing age and with baseline prehypertension (Table 5). Many differences were seen by cancer type, with the highest risk of moderate or higher hypertension in patients with renal, cervical, and prostate cancers and the lowest risk in breast cancer and malignant melanoma. For severe or crisis-level hypertension and for crisis-level hypertension only, the highest risk was found in patients with gastric cancer and the lowest risk in patients with connective and other soft tissue and breast cancers.
As seen in the incidence rate data, the risk of hypertension was considerably higher among patients on chemotherapy in all models. These results were found with any type of chemotherapy (cytotoxic, targeted, or combination) and in all lines of chemotherapy, with hazard ratios (HR) ranging from 2.2 to 3.5 for moderate or higher hypertension, 1.6-8.0 for severe or higher hypertension, and 1.2-6.7 for crisis-level hypertension, although these risks were not statistically significant for the severe and crisis-level hypertension models.

Discussion
The present analysis of outpatient oncology EMR data examined a cohort of patients with any of 11 solid tumors who were not hypertensive at the time of their first cancer diagnosis. We found incident hypertension to have a median time to onset of 96 days after the first cancer diagnosis and an overall incidence of 27 cases per 100 PY for moderate hypertension, 122 days to onset and 12 cases per 100 PY for severe hypertension, and 183 days to onset and 3 cases per 100 PY for crisislevel hypertension. Although the incidence rates of severe and crisis-level hypertension are lower than the rate of moderate hypertension for all cancer types, these lower rates are of more clinical concern because of the associated increased risks of coronary heart disease, stroke, heart failure, and end-stage renal disease.
All levels of hypertension severity appeared to be far more frequent during periods of chemotherapy use than nonuse for all cancer types. However, blood pressure was monitored far more often during chemotherapy, and the extent to which the more frequent monitoring may have biased results in favor of finding higher rates of hypertension with chemotherapy treatment is difficult to ascertain. The relationship between cancer treatment and hypertension as an adverse event has been well documented for a variety of agents [15,16]. It is therefore likely that the increased rates seen here during chemotherapy are not merely a statistical artifact resulting from the more frequent monitoring, although the magnitude of the difference may be overestimated in the present study. Conversely, patients exhibiting increasing BP during chemotherapy may have been started on an antihypertensive medication or, for those already on an antihypertensive, had the dose increased. Antihypertensive use is not reliably recorded in this oncology database, and so changing patterns of use during chemotherapy exposure and after the end of a course of chemotherapy cannot be determined. Continued use of antihypertensives after the end of chemotherapy could lead to decreased BP relative to baseline, whereas discontinuing 6 International Journal of Hypertension    the antihypertensive may be associated with a reduced need for frequent BP monitoring. Patients with breast cancer and malignant melanoma had the lowest incidence of hypertension at all severity levels compared to patients with other cancer types. For the more severe levels of hypertension, gastric and ovarian cancers were associated with the highest rates. These variations by cancer type might be related to the different cancer types and their progression, the specific treatments used for each cancer, and/or other risk factors that are associated with both the specific tumor type and hypertension.
There are several limitations to the present study that should be noted. As discussed above, the monitoring of blood pressure was not consistent and has in our analyses been found to be biased toward more monitoring during chemotherapy exposure, which may lead to underestimation of hypertension during the periods without chemotherapy exposure. Any blood pressure measurements that were taken outside the oncology clinic would not appear in the EMR database, although the treating oncologist may have been aware of those outside results, and they may have impacted patient care. For patients who had controlled hypertension at baseline through the use of antihypertensives, the incidence of hypertension during followup reflects breakthrough or worsening hypertension rather than new-onset hypertension.
In this study, we used a single BP measurement to capture a hypertension event for each level of hypertension severity. Patients who had a single high BP due to the "whitecoat effect" would be misclassified as hypertensive patients. However, we used relatively high BP thresholds to define even the lowest level of BP severity. Patients with a single BP measurement higher than 150 mm Hg SBP or 100 mm Hg DBP are likely to be genuinely hypertensive and not experiencing a brief increase in BP. Even if a few cases of white-coat hypertension were included in the analyses, it is unlikely that this would introduce any systematic bias.
This study looked at the natural history of cancer and the associated incidence of hypertension in various cancer types during periods of cancer treatment. The risk of hypertension associated with specific cancer treatments or supportive therapies should be examined in future research using established drug safety study methods. The majority of the clinics contributing to this EMR database are located in the southern US, limiting the geographic representativeness of the source population.
Despite the above limitations of the database, this data source is also an important strength of the present analysis. This study is based on data from outpatient oncology practices as opposed to the artificially constrained data obtained in clinical trials. Any patient with 1 of the 11 cancer types who did not have hypertension at baseline and who did have at least 1 BP measurement after the first cancer diagnosis was included in the study. No exclusions were made based on noncancer-related health status, age, or any of the other exclusions typically made in clinical trials. These data reflect realworld practice in oncology and the patterns of chemotherapy use and new-onset hypertension that occur.

Conclusions
HTN is a risk factor for coronary heart disease, stroke, heart failure, and end-stage renal disease, and the results of this study provide real-world data on the natural history of hypertension in cancer patients from US outpatient oncology practices. New-onset hypertension, regardless of severity, was observed in about one-third of cancer patients with various types of solid tumors. The incidence of hypertension at different severity levels varied with tumor type, and patients with gastric and ovarian cancer experienced the highest incidences of severe and crisis-level hypertension. Chemotherapy use appeared to be associated with an elevated risk of hypertension at each severity level, but blood pressure monitoring was more frequent during chemotherapy exposure.

Conflict of Interests
Kathy H. Fraeman and Beth L. Nordstrom are employees of Evidera, which was contracted by GlaxoSmithKline to work in collaboration on this study. Weixiu Luo is a former employee of Evidera. Sarah H. Landis and Sumitra Shantakumar are employees of GlaxoSmithKline.