eTable 1. List of Generic and Brand Names used to Define Continuous and Extended Combined Oral Contraceptives
eTable 2. List of Generic and Brand Names used to Define Cyclic Combined Oral Contraceptives
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Li J, Panucci G, Moeny D, et al. Association of Risk for Venous Thromboembolism With Use of Low-Dose Extended- and Continuous-Cycle Combined Oral Contraceptives: A Safety Study Using the Sentinel Distributed Database. JAMA Intern Med. 2018;178(11):1482–1488. doi:10.1001/jamainternmed.2018.4251
Is the risk for venous thromboembolism (VTE) higher with use of extended cyclic and continuous combined oral contraceptives (COCs) (84/7 or 365/0 days cycles) than traditional cyclic COC use (21/7 days cycle), while holding the progestogen type constant (levonorgestrel)?
This cohort study of 733007 in the US commercially insured population identified a slightly elevated VTE risk in association with continuous/extended COC use when compared with cyclic COC use. However, due to the small absolute risk difference and potential residual confounding, the findings did not show strong evidence supporting a VTE risk difference between continuous/extended and cyclic COC use.
The VTE risk difference was small, which may not translate into a clinically significant difference between continuous/extended and traditional cyclic COCs.
Continuous/extended cyclic estrogen use (84/7 or 365/0 days cycles) in combined oral contraceptives (COCs) could potentially expose women to an increased cumulative dose of estrogen, compared with traditional cyclic regimens (21/7 days cycle), and may increase the risk for venous thromboembolism (VTE).
To determine, while holding the progestogen type constant, whether the risk for VTE is higher with use of continuous/extended COCs than with cyclic COCs among women who initiated a COC containing ethinyl estradiol and levonorgestrel.
Design, Setting, and Participants
Incident user retrospective cohort study of primarily commercially insured US population identified from the Sentinel Distributed Database. Participants were women aged 18 to 50 years at the time of initiating a study COC between May 2007 and September 2015. Using a propensity score approach and Cox proportional hazards regression models, we estimated the hazard ratios of VTE overall and separately by ethinyl estradiol dose and age groups.
Initiation of continuous/extended or traditional cyclic COCs containing ethinyl estradiol or levonorgestrel of any dose.
Main Outcomes and Measures
First VTE hospitalization that occurred during the study follow-up, identified by an inpatient International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code of 415.1, 415.1x, 453, 453.x, or 453.xx.
We identified 210 691 initiators of continuous/extended COCs (mean [SD] age, 30.4 [8.6] years) and 522 316 initiators of cyclic COCs (mean [SD] age, 28.8 [8.3] years), with a mean of 0.7 person-years at risk among continuous/extended and cyclic users. Baseline cardiovascular and metabolic conditions (7.2% vs 4.7%), gynecological conditions (39.7% vs 32.3%), and health services utilization were slightly higher among continuous/extended cyclic than cyclic COC users. Propensity score matching decreased the hazard ratio estimates from 1.84 (95% CI, 1.53-2.21) to 1.32 (95% CI, 1.07-1.64) for continuous/extended use compared with cyclic COC use. The absolute risk difference (0.27 per 1000 persons) and the incidence rate difference (0.35 cases per 1000 person-years [1.44 vs 1.09 cases per 1000 person-years]) between the 2 propensity score–matched cohorts remained low, which may not translate into a clinically significant risk differences between cyclic and noncyclic estrogen use.
Conclusions and Relevance
Holding the progestogen type constant (levonorgestrel), we observed a slightly elevated VTE risk in association with continuous/extended COC use when compared with cyclic COC use. However, due to the small absolute risk difference and potential residual confounding, our findings did not show strong evidence supporting a VTE risk difference between continuous/extended and cyclic COC use.
Traditional cyclic combined oral contraceptives (COCs) consist of 21 days of active hormone pills, followed by a 7-day hormone-free or low-dose hormone interval (21/7 regimen). Newer extended- or continuous-cycle COCs (referred hereafter as noncyclic COCs) with reduction or elimination of the hormone-free or low-dose hormone intervals (84/7 and 365/0 cycles) were introduced in 2003. In addition to use for contraception, noncyclic COCs are also used as a treatment for endometriosis, dysmenorrhea, and other menstrual-related symptoms.1 It was estimated that, between 2006 and 2010, 17% of US women aged 15 to 44 years used COCs, of whom 88% used traditional cyclic 21/7 regimens and 12% used extended or other cyclic (84/7, 24/4, or 23/5) regimens.2
Previous safety studies have shown that current low-dose estrogen COC use (≤35 µg ethinyl estradiol) is associated with a 2- to 3-fold higher risk for venous thromboembolism (VTE) compared with no use.3,4 However, this elevated relative risk likely translates into a small absolute excess risk in the young women who use this regimen, far less than that attributable to pregnancy or the postpartum period.3-5 In recent years, epidemiologic studies have primarily focused on examining the risks for VTE associated with different generations of progestogens.6 The literature is scant on the VTE risk difference between noncyclic and cyclic estrogen in women using COCs. However, compared with cyclic regimens, noncyclic estrogen use could expose women to an increased cumulative dose of estrogen and therefore may increase the risk for VTE.
Lybrel (Pfizer) is the first continuous-cyclic (365/0) COC approved in the United States, containing ethinyl estradiol 20 µg and levonorgestrel 90 µg. On approval, the US Food and Drug Administration requested a postmarketing commitment study to evaluate the risk of VTE associated with Lybrel use. The study was later terminated early due to declining market share of Lybrel in the US market; however, the study results were added into the labeling for Lybrel and the 2 generic products. The postmarketing commitment study shows a crude incidence rate for VTE of 1.76 per 1000 person-years among Lybrel users, compared with 0.88 per 1000 person-years for cyclic COCs containing ethinyl estradiol and a progestogen of any type, and 0.51 per 1000 person-years for cyclic COCs containing ethinyl estradiol and levonorgestrel.7 Robust confounding adjustment could not be performed due to the small sample size (n = 12 281 Lybrel users) and the rarity of outcome events in the study population. Although this observational study suggests a potentially higher risk associated with use of noncyclic COCs compared with cyclic regimens, caution is needed in the interpretation of the study results due to the relatively small sample size and unmeasured, uncontrolled confounding.7 Lybrel ceased marketing in 2012, but its 2 generic versions remain on the market. Taking the advantage of large sample sizes and patient characteristic information in the claims database, the goal of this study was to determine, while holding the progestogen type constant (levonorgestrel), whether the risk for VTE was higher with noncyclic estrogen use (extended cycle 84/7 and continuous 365/0) than cyclic estrogen use among women who initiated a COC containing ethinyl estradiol and levonorgestrel.
This was a retrospective incident-user cohort study.
This study used the Sentinel Distributed Database.8,9 The Sentinel Distributed Database comprises administrative medical and prescription drug insurance claims and demographic data, converted into a common data model.10 As of early 2017, the distributed database contained information from 17 data partners with 223 million covered lives and 425 million person-years of observation time between 2000 and 2016. All data remain under the local control of each data partner site. Only minimum necessary information, usually in aggregate form, is requested for each query. The majority of individuals in the distributed database are privately insured, with approximately two-thirds of individuals between ages 18 and 65 years. This study was conducted as part of the Sentinel surveillance activities under the auspices of the US Food and Drug Administration and therefore not under the purview of institutional review boards.11
The study included women aged 18 to 50 years at the time of initiating use of a study COC during the study period (May 22, 2007, to September 30, 2015). Eligible women had at least 6 months of continuous medical and prescription drug coverage before the exposure index date (study COC initiation date). We excluded women who had a prior medical condition that may alter patients’ regular health care utilization or baseline risk for VTE, for example, a prior VTE, HIV/AIDS, anticoagulant use, malignant cancer, chemotherapy or radiation use, pregnancy, and organ failure or organ transplantation that occurred within 6 months before the index date. We identified pregnancies using a diagnostic or procedural code–based algorithm previously established in other Sentinel work.12 We further excluded women who were in their postpartum period, defined as 42 days after a live birth delivery, during the baseline period.
We identified the exposures of interest using outpatient pharmacy dispensing records. The exposure cohort was initiators of noncyclic (84/7 and 365/0) COCs containing ethinyl estradiol/levonorgestrel of any dose (continuous and extended products combined) (see eTables 1 and 2 in the Supplement). The comparison cohort included initiators of any cyclic COCs containing ethinyl estradiol/levonorgestrel of any dose. We defined initiators as women with no use of the study COCs that were used to define each cohort during a 6-month lookback period. That is, noncyclic initiators could not use a noncyclic regimen in the lookback period, but the same women could have used cyclic products (and vice versa). To define continuous exposure, we stockpiled dispensings of the same medication by generic name, bridged any gap of 30 days or less between depletion of 1 dispensing (evidenced by days supply) and initiation of the next, and applied an episode extension period of 30 days to the last dispensing. We included only the first valid treatment episode for each woman.
The study outcome was the first inpatient diagnosis of VTE (recorded in any position) that occurred during study follow-up, identified using an algorithm of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 415.1, 415.1x, 453, 453.x, or 453.xx, which had a positive predictive value of 65% based on a prior Sentinel study.13 In a sensitivity analysis, we used a broader VTE definition that included an inpatient diagnosis of VTE or an outpatient diagnosis of VTE followed by an outpatient dispensing of anticoagulant treatment by any administration route within 4 weeks.
We followed patients from the index date until the date of the earliest occurrence of VTE hospitalization, health plan disenrollment, cessation of initiated COC, initiation of a product of the other COC regimen in comparison or a nonstudy hormone contraceptive (eg, progestogen-only contraceptives or nonoral dosage formulations), death, pregnancy or live birth delivery, or the end of study period.
We identified and included in the propensity score model the following covariates during the 6-month baseline period using relevant ICD-9-CM diagnostic or procedural codes and National Drug Codes: age, calendar year, combined comorbidity score,14 health services utilization (number of ambulatory, other ambulatory, emergency department, hospital inpatient, and nonacute institutional encounters), level of medication utilization (number of dispensing, unique generics dispensed, unique drug classes dispensed), and the following drug use and medical condition categories: use of the other COC regimen in comparison (eg, baseline cyclic COC use among initiators of noncyclic COCs), use of any nonstudy hormonal contraceptives (estrogen and progestogen combined drugs of any administration route, excluding progestogen-only products), gynecological conditions (uterine leiomyoma, inflammatory disease of the female pelvic organs, disorders of female genital tract, endometriosis, ovarian cyst or polycystic ovarian and related symptoms, premenstrual tension syndromes, menorrhagia, and migraine), hypercoagulable states and coagulation defects, cardiovascular and metabolic conditions (metabolic syndrome, hyperlipidemia, diabetes mellitus, hypertension), cardiac conditions, venous catheterization, renal conditions, inflammatory conditions, obesity and overweight, tobacco use, immobility, and surgery. We did not include cerebral palsy, cystic fibrosis, sickle cell anemia, thoracic outlet syndrome, and infectious disease (sepsis and osteomyelitis) owing to their low prevalence. The diagnosis and procedure codes used to identify these covariates can be found on the Sentinel Initiative website.15
The primary analysis was an on-treatment analysis comparing noncyclic with cyclic COC initiators on the risk for VTE. We adjusted for potential confounders using propensity score methods. Specifically, we included all the aforementioned covariates in a logistic regression model to predict the probability of receiving the exposure of interest. We matched each noncyclic COC user with a cyclic COC user within the same data partner site based on their estimated propensity score using the nearest-neighbor approach (caliper = 0.01). We used Cox proportional hazards regression models to estimate hazard ratios (HRs) and corresponding 95% confidence intervals in the unmatched population (adjusting for study site only), and separately in the matched population. The analytical process, including propensity score calculation and effect estimation, was completed using pretested, parameterizable modular programs in Sentinel, including the Cohort Identification and Descriptive Analysis Tool and the Propensity Score Analysis tool, version 188.8.131.52,17
We conducted a number of subgroup and sensitivity analyses, including HR estimates by age group (18-24, 25-34, and 35-50 years), ethinyl estradiol dose (20 and 30 µg), duration of follow-up time (1-90 and 1-183 days), and HR estimates for continuous regimens and extended regimens separately. As a sensitivity analysis, we used the aforementioned broader VTE definition described. Last, we restricted the analysis to new users with no prior use of combined hormonal contraceptive of any type (naive new users) during the 6-month baseline period.
We identified 210 691 initiators of noncyclic COCs, including 11 504 continuous COC users. There were 522 316 initiators of cyclic COCs. The mean person-years at risk were 0.7 for both noncyclic and cyclic users. The women in the noncyclic COC cohort were slightly older than those in the cyclic COC cohort (mean [SD] age, 30.4 [8.6] vs 28.8 [8.3]). A total of 3.0% of the noncyclic COC users had used a cyclic COC product within the 6 months prior to the index date and 35.0% had used a nonstudy combined hormonal contraceptive. In the cyclic COC cohort, 0.9% had used a noncyclic COC product and 26.9% a nonstudy contraceptive. A total of 5.7% of women were excluded from the noncyclic COC cohort owing to pregnancy or postpartum status in the baseline period, and 0.7% of eligible women were censored during study follow-up owing to pregnancy or live birth delivery. The corresponding proportions were 7.9% and 1.5%, respectively, in the cyclic COC cohort.
As reported in Table 1, baseline comorbidities were generally rare, except for gynecological conditions (39.7% vs 32.3% in the noncyclic and cyclic COC cohorts, respectively), cardiovascular and metabolic conditions (7.2% vs 4.7%), and inflammatory diseases (2.7% vs 1.8%). Baseline health services utilization was slightly higher among noncyclic than cyclic COC users. The patient characteristics were generally balanced between the 2 study cohorts after 1:1 propensity score matching, which resulted in 203 402 matched pairs.
There were 228 incident VTE cases identified in noncyclic COC users and 297 in cyclic COC users. As reported in Table 2, before propensity score matching, the incidence rate estimate was 1.54 (95% CI, 1.34-1.74) cases per 1000 person-years among noncyclic COC users, compared with 0.83 (95% CI, 0.74-0.93) cases per 1000 person-years among cyclic COC users, yielding a crude HR estimate of 1.84 (95% CI, 1.53-2.21). After propensity score matching, the incidence rate estimates decreased to 1.44 (95% CI, 1.24-1.64) and 1.09 (95% CI, 0.92-1.27) per 1000 person-years in the 2 cohorts, and the adjusted hazard ratio (aHR) decreased to 1.32 (95% CI, 1.07-1.64). The absolute incidence rate difference between the propensity score–matched cohorts was 0.35 cases per 1000 person-years, and the absolute risk difference was 0.27 per 1000 persons.
Subgroup analyses showed that, among users of ethinyl estradiol 20 µg, the aHR was 1.60 (95% CI, 0.94-2.71) for noncyclic COC use compared with cyclic COC use; and among users of ethinyl estradiol 30 µg, the aHR was 1.23 (95% CI, 0.88-1.73). In the age group 18 to 24 years, the aHR was 1.66 (95% CI, 0.95-2.90); in the age group 25 to 34 years, the aHR was 1.19 (95% CI, 0.81-1.74); and in the age group 35 to 50 years, the aHR was 1.38 (1.03-1.85). The aHR comparing noncyclic with cyclic COC use in the first 3 months was 1.37 (95% CI, 0.98-1.93), and 1.47 (95% CI, 1.11-1.94) in the first 6 months of use. The use of continuous noncyclic COCs alone (n, matched pair = 11 504) was associated with an aHR of 1.45 (95% CI, 0.70-2.99), and the use of extended COCs alone (n, matched pair = 195 637) with an aHR of 1.34 (95% CI, 1.08-1.66). A broader VTE definition, which additionally allowed VTE outpatient diagnoses coupled with evidence of anticoagulant use, did not alter the HR estimate generated from the narrow VTE definition (inpatient VTE diagnosis codes only). Naive new users (n, matched pair = 127 256) who did not use combined hormonal contraceptives of any type during the 6-month baseline period showed an aHR of 1.49 (95% CI, 1.17-1.92).
Since the introduction of the first combined hormonal oral contraceptive in the US in 1960, the estrogen dose in COCs has been substantially decreased from approximately 50 µg to the recent 20 to 35 µg ethinyl estradiol, which corresponded to a reduced incidence of VTE and fewer estrogen-related adverse effects among COC users.18 In addition, the cycle regimens have been evolving, from the initial 21/7 monthly cycle to an extended or continuous cycle (84/7 and 365/0 days). Yet, few epidemiology studies have been conducted to examine the VTE risk associated with the altered cyclic regimens. To our knowledge, the present study is one of the first epidemiologic studies to examine the VTE risk with noncyclic estrogen use, compared with cyclic estrogen use, among women using COCs in the United State.
Prior safety studies primarily focus on examining the VTE risk associated with different types of progestogen in COCs (such as third-generation pills compared with first- or second-generation pills). Study results have been inconsistent largely as a result of the variations in study designs and potential residual confounding. In a meta-analysis published in 2018, Dragoman et al19 reported that the use of COCs containing cyproterone acetate, desogestrel, dienogest, drospirenone, or gestodene was associated with a small increase in risk for VTE (pooled risk ratios, 1.5-2.0), compared with the use of levonorgestrel-containing COCs. However, Bateson et al argued that the difference in VTE risk based on the choice of progestin in COCs is likely small in absolute terms.3 The present study held the type of progestogen constant between the study cohorts to effectively control for the potential risk difference attributable to progestogen types.
Some epidemiology studies compared the risk for VTE between various generations of COC regimens without sufficient control for potential confounders (such as body mass index or smoking). For example, an observational study reported that the rate ratio of confirmed VTE for users of oral contraceptives with desogestrel was 2.2 (95% CI, 1.7-3.0), with gestodene was 2.1 (95% CI, 1.6-2.8), and with drospirenone was 2.1 (95% CI, 1.6-2.8), compared with levonorgestrel-containing COCs, adjusting for age, calendar year, level of education, and duration of COC use.20 Other baseline covariates may have contributed to the observed, elevated risks with these progestogen uses. Furthermore, some COC safety studies included primarily healthy COC users, which may limit the generalizability of the study results.21-23 The present study included all initiators of COCs containing ethinyl estradiol/levonorgestrel of any dose, and measured and adjusted for (to the extent possible) not only the known baseline comorbidities associated with VTE and COC use, but also additional potential confounders, such as gynecological conditions.
The present study revealed that, in this population-based cohort study of a primarily commercially insured population, noncyclic COC users were slightly older than cyclic COC users. Baseline cardiovascular and metabolic conditions, as well as gynecological conditions and baseline health services utilization, were higher among noncyclic than cyclic COC users. This imbalance suggested potential preferential prescribing by physicians (ie, generally healthier patients were more likely to receive cyclic COCs than noncyclic COCs), which could bias the relative risk estimate away from null. Propensity score matching decreased the HR estimate from 1.84 (95% CI, 1.53-2.21) to 1.32 (95% CI, 1.07-1.64) for noncyclic use compared with cyclic COC use. Yet, the absolute risk difference (0.27 per 1000 persons) and the incidence rate difference (0.35 cases per 1000 person-years) between the propensity score–matched noncyclic and cyclic estrogen cohorts remained low, which may not translate into a clinically significant risk difference between cyclic and noncyclic estrogen use. Therefore, we did not see strong evidence supporting a VTE risk difference between noncyclic and cyclic estrogen use. It is generally believed that the excess risk for VTE is highest during the first year a woman ever uses a COC and declines afterward, compared with nonuse.7 Yet, in the present study, the HRs for VTE for noncyclic vs cyclic COC use seemed relatively stable in the first 6 months of study follow-up, 1.37 (95% CI, 0.98-1.93) for the first 90 days and 1.47 (95% CI, 1.11-1.94) for the first 183 days.
The present study is subject to several limitations. First, although we adjusted for a large number of covariates, there might still be residual or unmeasured confounding. For example, we did not control for concurrent use of medications, although the baseline comorbidities and health care utilization results in Table 1 suggested a possible low prevalence of concurrent medication use among the study population. In this study propensity score matching appeared to be generally effective in confounding adjustment because the baseline patient characteristics were well balanced in the propensity score–matched cohorts and the HR estimate attenuated from a crude estimate of 1.84 (95% CI, 1.53-2.21) to 1.32 (95% CI, 1.07-1.64) after matching. However, there might still be residual confounding due to unmeasured or incompletely measured covariates. Second, we did not examine and account for potential noncyclic utilization of cyclic COC products (ie, skipping inactive pills to achieve extended contraceptive effects) because such a medication use pattern is generally discouraged by physicians and/or health insurers and may be uncommon due to available low-cost generics for the extended and continuous COC products. Nevertheless, the study results could be biased due to potential misuse or nonuse of prescribed COCs in the study populations. Third, the use of noncyclic COCs, compared with cyclic COC use, does not necessarily lead to a higher level of cumulative estrogen exposure because the noncyclic and cyclic users were not matched by duration of COC use. However, the mean duration of COC use was balanced between the 2 study cohorts, suggesting a likely higher cumulative exposure among noncyclic estrogen users. Fourth, we did not limit the primary analysis to naive new users (women with no use of combined hormonal contraceptive of any kind in the 6-month washout period) due to the concern of its smaller sample size (n, matched pair = 127 256). The inclusion of nonnaive users may lead to underestimation of the VTE risk with noncyclic COC use. Last, the VTE events were not adjudicated or confirmed by medical record review; however, we used in the primary analysis a strict claims-based VTE definition by limiting to inpatient diagnosis codes only, which likely increased the accuracy of VTE case identification.
Because of the small absolute risk difference and potential residual confounding, these findings did not show strong evidence supporting a VTE risk difference between noncyclic and cyclic estrogen use. Accordingly, we do not recommend selective prescribing of COCs based on the cyclic and continuous/extended type. Clinicians should prescribe COCs based on patients’ individual risk factors and preferences.
Accepted for Publication: June 28, 2018.
Corresponding Author: Jie Li, PhD, Division of Epidemiology, Office of Surveillance and Epidemiology, CDER, FDA, 10903 New Hampshire Ave, Building 22, Office 2483, Silver Spring, MD 20903 (firstname.lastname@example.org).
Published Online: October 1, 2018. doi:10.1001/jamainternmed.2018.4251
Author Contributions: Dr Huang had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Li, Panucci, Moeny, Liu, Toh, Huang.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Li, Moeny, Toh, Huang.
Critical revision of the manuscript for important intellectual content: Li, Panucci, Liu, Maro, Toh, Huang.
Statistical analysis: Panucci, Liu, Toh.
Administrative, technical, or material support: Li, Moeny, Liu, Maro, Huang.
Study supervision: Li, Moeny, Maro, Toh, Huang.
Conflict of Interest Disclosures: None reported.
Funding/Support: The Sentinel Initiative is funded by the US Food and Drug Administration through the Department of Health and Human Services contract HHSF223200910006I.
Role of the Funder/Sponsor: The US Food and Drug Administration had a lead role in the design and conduct of the study; interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views expressed in this article are those of the authors and are not intended to convey official US Food and Drug Administration policy or guidance.
Additional Contributions: The authors thank the Sentinel Data Partners who provided data used in the analysis: Aetna, Blue Cross Blue Shield of Massachusetts, Harvard Pilgrim Health Care, HealthCore, HealthPartners Institute, Humana, Kaiser Permanente Colorado, Kaiser Permanente Hawaii, Kaiser Permanente Northern California, Kaiser Permanente Northwest, Kaiser Permanente Washington, Marshfield Clinic Research Institute, Meyers Primary Care Institute, OptumInsight, and Vanderbilt University Medical Center (who received data from the Division of TennCare of the Tennessee Department of Finance and Administration). The authors thank April Duddy, MS, Andrew Petrone, MPH, and Anita Wagner, PhD, at the Sentinel Operations Center for their programming and clinical review assistance. The Sentinel Data Partners and personnel at the Sentinel Operations Center received funding support from the US Food and Drug Administration as part of Sentinel Initiative for such contribution. The authors thank Christine Nguyen, MD, and Shelley Slaughter, MD, PhD, from Office of New Drugs, and Dr Rima Izem, PhD, and Dr Rongmei Zhang, PhD, from the Division of Biostatistics 7 at Center for Drug Evaluation and Research, US Food and Drug Administration, for their input on clinical review and statistical analyses. The FDA personnel are employees of FDA and receive no additional compensation for such contributions.