Varenicline and Nicotine Replacement Use Associated With US Food and Drug Administration Drug Safety Communications

Key Points Question Did use of varenicline change after early US Food and Drug Administration drug safety communications regarding neuropsychiatric adverse events? Findings This cross-sectional study observed a reduction in varenicline prescribing following the release of US Food and Drug Administration drug safety communications on the medication. Interrupted time series analysis showed a 68.7% decrease in Veterans Health Administration outpatient prescriptions and a 38.0% decrease in Medicaid prescriptions. Meaning In the wake of US Food and Drug Administration and Veterans Health Administration communications, prescriptions for varenicline decreased significantly, which may have been associated with negative public health consequences.


Introduction
The US Food and Drug Administration (FDA) drug safety communications and public health advisories are intended to inform the public about emerging drug-related safety issues. 1 These communications are frequently based on early safety signals that require ongoing assessment for confirmation but nonetheless can substantially affect drug use with potentially unintended consequences. 2,3 The FDA recognizes this concern and recently commissioned a study to evaluate postmarket FDA safety labeling changes. 4 That report developed a framework for future studies to inform the FDA on the outcome of safety labeling changes.
Varenicline was approved by the FDA in May 2006. On November 20, 2007, the FDA released the first safety communication regarding suicidal thoughts and aggressive and erratic behavior, which was followed by a public health advisory on February 1, 2008, and the addition of a boxed warning on July 1, 2009. 5 In June 2016, the Evaluating Adverse Events in a Global Smoking Cessation Study (EAGLES) was published, showing no significant increase in psychiatric/behavioral effects with varenicline relative to nicotine replacement therapy (NRT) or placebo. 6 Varenicline's boxed warning was removed on December 16, 2016.
The Veterans Health Administration (VHA) has recognized the importance of smoking in healthrelated outcomes and actively promotes efforts to identify smokers and offer treatment. By 1998, more than 90% of smokers in the VHA were actively offered treatment; by 2007, the VHA adopted smoking cessation performance measures that encouraged offering smoking cessation drugs and referral for intensive smoking cessation programs. 7 As one of the treatments available to meet such performance metrics, varenicline's availability was accepted widely within the VHA.
In this study, we aimed to evaluate the association between FDA drug communications and the use of varenicline within the VHA and evaluate the generalizability of our findings using Medicaid data. We also simulated the potential consequences of decreased varenicline use on lost opportunities to assist patients to quit smoking and their downstream health outcomes.

Use of Varenicline and NRT in the VHA
We extracted data on outpatient prescriptions between October 1, 2001, and December 31, 2018, from the VHA Pharmacy Benefits Management Services. We identified the number of unique patients receiving varenicline and/or NRT every quarter. Patients who received medications in more than 1 quarter were counted in each period. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. 8 This study was deemed exempt by the VHA Pittsburgh Institutional Review Board and informed consent was not required owing to the use of aggregate, unidentifiable data.

Use of Varenicline in Medicaid
We extracted all fee-for-service and managed care records for varenicline from July 1, 2006, the second quarter (Q2) of 2006 (varenicline approval), to September 30, 2018, the Q3 of 2018 (most recent data at the time of analysis), which contained the number of units, number of prescriptions, and total amount reimbursed in each state and quarter for a given product. We calculated the number of prescriptions filled for varenicline nationwide every quarter.

Estimated Health Outcomes
To assess the potential outcome of lost opportunities for successful smoking cessation with the use of varenicline, we used data from several observational studies. To estimate the additional use of varenicline vs NRT for long-term smoking cessation in a real-world primary care setting, we used a rate of 4.99 per 100 successful long-term smoking cessation attempts using varenicline from a large study of more than 220 000 patients. 9 To estimate the outcome of sustained smoking cessation on mortality, we used data from a study of successful quitters in a smoking cessation program that examined mortality at a mean of 14.5 years after intervention. 10 We assumed that without safety warnings for varenicline, its use in the VHA would have remained stable in 2009-2016 at the 2008 rate. This estimation is conservative since drug adoption often takes 2 to 3 years; thus, it is likely that use trend would have continued to increase. 11 We calculated the difference between actual varenicline use through 2016, when the warnings were lifted. In addition, we estimated deaths due to lost smoking cessation opportunities over the 8 years of varenicline warnings.

Statistical Analysis
To formally test the association between safety communications and varenicline and NRT use in the

Trends in Varenicline Use in Medicaid
Varenicline use in Medicaid experienced a similar increase during the first year after its approval,

Discussion
Using varenicline as a case study, we found that FDA drug safety communications were followed by significant reductions in drug use, with what may be resultant lost opportunity for successful smoking cessation. Because varenicline has been shown to be more effective than other pharmacotherapy options in assisting quit attempts in both clinical trials and real-world situations, 12 it is likely that the FDA warnings regarding varenicline safety and resultant actions of health care organizations and prescribing clinicians led to a net health care loss.
Our estimates of a potential of 454 lost lives within the VHA might be viewed as speculative owing to multiple assumptions used in our analysis. We are unaware of any prospective data comparing mortality benefit of successful smoking cessation between varenicline and NRT. One short-term observational study in the United Kingdom with 2 years of follow-up did not find any difference in mortality or hospitalization between varenicline and NRT users. 13 Conversely, another study using national claims databases in the United States reported fewer smoking-attributable outcomes and lower health care costs in patients receiving varenicline compared with NRT, although mortality was not reported. 14 We believe that, while the magnitude of mortality estimates are uncertain, the overall effect on public health is probably conservative, as we did not assume an describe the occurrence of less common outcomes that were not identified in clinical trials owing to small sample sizes or limited follow-up. The VHA uses these FDA reports to examine potential adverse outcomes in its covered population and in some cases may develop communications to send to VHA clinicians advising them of potential drug safety concerns.
The VHA response to FDA warnings was likely more intense than in most other health care This hearing, which was stimulated by a series of reports in the Washington Times, focused on a clinical trial funded by the VHA (CSP 519) that included varenicline as part of a smoking cessation intervention for veterans with posttraumatic stress disorder. 20 The VHA was accused of a lack of response to the FDA communications on safety concerns with varenicline, and concern was raised that the VHA was experimenting on veterans by offering varenicline as part of a larger smoking cessation intervention. Subsequent internal VHA concerns regarding the safety of varenicline in patients with mental illness, combined with FDA safety warnings, media reports, and congressional concern, triggered a series of reiterated safety warnings by the VHA and changes in their varenicline prescribing criteria. Specifically, the VHA established requirements to formally assess veterans for suicide risk prior to starting varenicline therapy, as well as to obtain ongoing risk assessments if the medication was prescribed. 21 The VHA was particularly concerned about the safety of varenicline treatment for patients with underlying mental illness. In addition to changes in prescribing criteria, the VHA developed an ongoing varenicline safety surveillance program and reviewed observational data from veterans who received varenicline, including those with and without serious mental illnesses. These additional VHA warnings, prescribing guidance changes, and ongoing safety initiatives likely explain the observation that the decrease in use of varenicline was more substantial in the VHA population than in Medicaid.
In the era of accelerated drug approvals, it is particularly important to reexamine the evidence necessary for approval as well as the evidence base for the release of postmarket safety concerns or for the identification of lack of effectiveness. It is estimated that, prior to approval, a drug is studied in a range from a few hundred to 3000 individuals. 22,23 In the case of varenicline, FDA approval was based on a total of 4690 patients enrolled in clinical trials; however, most of these individuals were enrolled in short-term studies of 12 weeks. 24 With no signal for suicide or serious adverse neuropsychiatric events identified in clinical trials, the FDA faced a dilemma when early reports of serious neuropsychiatric events in patients exposed to varenicline emerged. This difficult situation led to several drug safety communications and, ultimately, a boxed warning, which remained until subsequent clinical trials and studies-including a varenicline safety study performed by the VHA Center for Medication Safety 25 -did not confirm these safety signals. Although we acknowledge the importance of drug safety communications and understand the need to respond with label changes with a goal to protect the public as drug safety signals are being evaluated, the FDA and health care organizations, such as the VHA, need to balance safety concerns with the possibility that even preliminary warnings can have significant implications for prescribing. While alerting the public about potential drug-related harms, these communications can also have unintended consequences and may lead to poorer health outcomes by preventing patients from benefiting from effective medications.
Given the uncertain nature of early safety signals, it will be important to reexamine the evidence base necessary to release safety communications and ponder it against the strength of the evidence collected from preapproval clinical trials. Conversely, although our report and previous reports have

Limitations
Our study is subject to several limitations. Our findings of significant changes in varenicline prescribing using an observational study design may not be generalizable to other patient populations. However, we observed a significant decrease of 38.0% in varenicline use in Medicaid, which is consistent with prior reports from Pfizer that documented a decrease in sales of approximately one-third in varenicline use from 2008 to 2014. 26 Our projections of lives lost were derived from non-VHA patient populations and thus might not be applicable to our patient population. The greatest limitation is that, for VHA varenicline slopes, we were unable to differentiate prescribing changes associated with FDA warnings and labeling changes from those associated with VHA drug safety communications and changes in prescribing guidance and national temporal trends.

Conclusions
This study found that early communications from the FDA followed by a labeling change were associated with additional safety measures within the VHA and a subsequent decrease in prescribing of varenicline for smoking cessation, which may have been associated with negative public health consequences. Analyses from other health care systems should evaluate the association between early drug safety communications and medication prescribing and potential benefits or harms that might result from those changes. In responding to these early communications, health care systems should consider the magnitude of potential harm against known benefit from use of that medication.