Objective: To determine the effect of switching drugs on thecompliance and persistence of new statin users.
Study Design: Retrospective database analysis of pharmacyclaims provided by a large pharmacy benefit manager. The studysample consisted of 38 866 new statin users 18 to 65 years oldbeginning treatment with atorvastatin calcium, fluvastatin sodium,lovastatin, pravastatin sodium, or simvastatin.
Methods: Compliance was measured by the "medication possessionratio," and persistence was measured by the time to discontinuation.Switching rates were derived from the proportions ofpatients filling a prescription other than the initial statin.
Results: Patients who switched statins were less compliant by18.9% (odds ratio, 0.81; < .001), as defined by the probability ofhaving a medication possession ratio of 0.8 or higher, and wereless persistent by 20.9% to 48.3% (< .001) depending on the gaplength used to define discontinuation.
Conclusions: Switching statins substantially reduces the likelihoodthat patients will be compliant and remain on treatment longenough to obtain the full benefit of statin treatment. To ensure bettercompliance, special care should be given to patients whochange drugs.
(Am J Manag Care. 2005;11:670-674)
Coronary heart disease and stroke impose a largeburden on individual patients and society. In2001, coronary heart disease and stroke hadprevalences of 6.4% and 2.0%, respectively,1 and theircombined projected cost in 2005 is $142.1 billion.2Since the National Cholesterol Education Program(NCEP) issued in 2001 its third report of the ExpertPanel on Detection, Evaluation, and Treatment of HighBlood Cholesterol in Adults (Adult Treatment PanelIII),3 other investigations have suggested that a moreaggressive approach should be taken in cholesterolmanagement, targeting lower levels of low-densitylipoprotein (LDL) cholesterol.4 These 2004 guidelinesrecommend a low-density lipoprotein cholesterol levelof less than 100 mg/dL (<2.59 mmol/L). Regardless oftheir initial LDL cholesterol levels, patients should initiatetherapeutic lifestyle changes first, followed by drugtherapy if additional reduction is needed. Statins arerecommended as the first-line drug therapy because oftheir effectiveness and good tolerability,3 but nonstatinoptions such as bile acid sequestrant or nicotinic acidare also available.
Statins have been shown to reduce the number of vascularevents when used in primary and secondary preventionof coronary heart disase.5-8 They also contributeto a lower incidence of stroke.1,2 Despite the weight ofevidence supporting sustained treatment, nonadherenceand suboptimal adherence remain common problemsamong patients treated with statins.9-14 Poor compliancenot only deprives patients of the benefit of treatment butalso can lead to serious adverse events.15-17
Compliance and persistence depend partially on howsuccessful the treatment is at lowering cholesterol to atarget concentration and on what the patient perceivesto be the health benefit.18,19 Treatment adverse effectsare also a major source of low compliance, as 41% ofpatients discontinue therapy because of poor efficacy oradverse events.20
Prescribing suboptimal initial drugs and maintainingpatients on low dosages for too long, among other inappropriatetherapies, reduce the chance that treatmentwill succeed in reaching cholesterol targets.21 Once drugtreatment has started, LDL cholesterol control mayrequire higher dosages of the initial drug, additionallipid-lowering agents, or a switch to another statin, atherapeutic step justified by variation in the tolerabilityand efficacy of statins.22-24
In this context, it is critical to determine what effectswitching statins has on compliance and whether a drugchange compromises treatment goals. A better knowledgeof statin utilization patterns in clinical practicewould clarify the options available to physicians forimproving lipid management and would allow them tobetter understand the consequences of altering drugtreatment. To our knowledge, no published researchhas focused on this issue. This research seeks to fill thisvoid by estimating how switching statins affects complianceand persistence among new statin users in a managedcare environment.
Study Sample and Selection Criteria
We conducted a retrospective cohort study based ondata provided by a West Coast pharmacy benefit manager.Patients were between the ages of 18 and 65years. All patients were new statin users as defined bya 1-year wash-out period before filling the first prescriptionfor a statin. Each patient had been continuouslyeligible for benefits for 2 years between April2001 and June 2004, specifically for 1 year before and1 year after filling the first statin prescription. Thebeginning date for continuous eligibility varied amongpatients. The date of first prescription fill defined theindex date for the analysis.
Compliance and persistence were measured for thefollowing 5 statins: atorvastatin calcium, fluvastatinsodium, lovastatin, pravastatin sodium, and simvastatin.(Rosuvastatin calcium was introduced too recentlyto allow for evaluation of compliance within the timeframe of the analysis.)
Compliance was measured by the number of daysthe medication was available to the patient, called the"medication possession ratio"(MPR). The MPR is thesum of the days of supply for 1 year after the first prescriptionwas filled, divided by 360. Any prescriptionsfilled later than 360 days after the index date wereexcluded, as were any days of supply that occurred afterthe end of follow-up. The MPR calculation included allstatins after the index date, even if different from theinitial drug.
Persistence represents the length of time patientsremained supplied with statins. It was measured by thetime to discontinuation, defined as the number of daysbefore the first gap in medication possession. A gap isthe number of days between the time patients ran out oftheir current prescription and the time they filled a newprescription. Before reaching this threshold, patientswere assumed to be fully compliant. The supply of overlappingstatin prescriptions was added to the total daysof continuous supply. Results are provided herein forgap lengths of 15, 30, and 60 days.
Finally, we defined indicators of switching statinsand of increasing the dosage. We used binary indicatorsfor switching to another statin during the 1-year followupand for increasing the dosage from the strength ofthe initial statin.
All predictors were used as covariates in the multivariateanalyses. They were divided into 4 categories.First, we defined predictors related to patients, includingage, sex, geographic region, and health status asdetermined by RxRisk, a risk assessment system thatclassifies patients by their medication use during theyear before their statin index date.25 Second, we definedpredictors related to health insurance, including formularytypes, health plan sizes, and mean copayments.There are 3 formulary types, ranked from the most tothe least restrictive, namely, closed, incentive, and opentypes. A closed formulary does not cover nonformularydrugs, an incentive formulary covers nonformularydrugs but at a higher copayment, and an open formularycovers nonformulary drugs at the same copayment.The health plan size variables were divided into the followingcategories based on the total number ofenrollees: fewer than 50 000, 50 000 to 200 000, andmore than 200 000. The mean copayments were calculatedby averaging out-of-pocket costs across all prescriptionsfilled before switching statins if there was aswitch and across all prescriptions otherwise. Third, tocontrol for the effect of complex drug treatment andpossible drug interactions on adherence, we definedvariables related to drugs used after the initial statin.These variables included the utilization of nonstatinantilipidemic agents, the use of antihypertensive drugsfor cardiovascular disease (antiarrythmics, vasodilators,and cardiac stimulants), and the number of prescriptionsper month before the switch for switchers andduring the whole year for nonswitchers. We alsoaccounted for the use of nonstatin antilipidemic agentsbefore treatment initiation. Fourth, the year of treatmentinitiation was included in our estimates to removethe effect of treatment trends and stricter treatmentguidelines.
One of the challenges of observational studies is todistinguish the effect of treatment from the effect offactors that influence treatment selection. Without randomassignment, treatment groups may differ in waysthat influence treatment choice and treatment outcomes,thereby introducing bias in the estimation oftreatment effect. In this study, it was important toaccount for the differences that may affect outcomesbetween patients who do and do not switch drugs. Tocontrol for this problem, the propensity score (PS)method was used to adjust for pretreatment differencesbetween treatment groups.26 Controlling forpotential bias and adjusting for pretreatment differencesrequire that variables determining selection,such as alternative treatments, be included in the estimationof the PS as controls. The PS is a patient's estimatedprobability of switching drugs as calculated bylogistic regression of all pretreatment variables (age,sex, health status, mean copayment, use of nonstatinantilipidemic agents, etc) on a binary variable equal to1 if the patient switched drugs. The inverse of the PSis then used to weigh each observation in other logisticregressions. Propensity score-weighted logisticregressions were used to evaluate the effect of treatmenton the probability of compliance (ie, having anMPR of ≥ 0.8) and different degrees of low compliance(ie, having MPRs of 0.4-0.8 and < 0.4). Other PSweightedlogistic regressions were used to estimate theprobability of the time to discontinuation (ie, havingfirst gaps in treatment of ≥ 180 and ≥ 360 days).Analyses were performed with SAS 9.1 (SAS InstituteInc, Cary, NC).
There were 38 866 patients included in the studysample. Among these, 8.4% switched drugs during theirfirst year of treatment, and the others continued takingtheir initial drug throughout follow-up or stopped statintreatment completely (Table 1). Patients who switchedstatins were slightly older than nonswitchers (< .001,test) and were more likely to be female (< .05, χ2test). They were also more likely to receive drug treatmentsfor comorbidities, but the differences in theirhealth status were generally small (< .05, χ2 test).Their geographic distribution was also similar (<.05, χ2 test), with the largest numbers of patients livingin the eastern United States.
Statin switchers, 56.5% of whom switched within 6months of the initial prescription, were less likely to becompliant than nonswitchers (Table 2). They weremore likely to experience a gap in therapy of at least 30days within the first year or first 6 months of initialtherapy. Switchers were less likely to increase thedosage of their initial prescription. Finally, by switching,patients reduced their out-of-pocket cost per prescriptionby $1.33 on average.
Multivariate analysis confirmed that switchingstatins can hurt compliance. Switching reduced theodds of high compliance by 18.9% (odds ratio [OR],0.81; <.001). Relative to nonswitchers, the probabilitiesof an effect on compliance (adjusted with PSweights) were as follows for the 3 MPR levels (≥0.8, 0.4-0.8, and <0.4): OR, 0.81 (95% confidence interval [CI],0.79-0.84); OR, 1.82 (95% CI, 1.76-1.88); and OR, 0.66(95% CI, 0.64-0.68); respectively.
before 360 days,
before 180 days,
Persistence was similarly affected by switching, asdemonstrated by the probabilities that patients will discontinuetreatment or experience a gap in treatmentbefore 360 or 180 days. Depending on the gap lengthused to define discontinuation, the probability of interruptingtreatment before the end of the first year was20.9% (for a 60-day gap) to 48.3% (for a 15-day gap)greater for switchers than nonswitchers. Relative tononswitchers, the probabilities of an effect on persistence(adjusted with PS weights) were as follows for the3 gap lengths (15 day, 30 day, and 60 day): OR, 1.48 (95% CI, 1.43-1.54); OR, 1.44 (95% CI, 1.39-1.49); and OR, 1.21 (95%CI, 1.17-1.25); respectively; and OR, 1.36 (95% CI, 1.31-1.40); OR, 1.35(95% CI, 1.30-1.39); and OR, 1.16 (95% CI, 1.12-1.19);respectively.
Evidence showed that switching was the cause of thelow compliance and persistence observed among switchers.First, the observed difference in persistence betweenswitchers and nonswitchers increased substantially overtime, from 34.6% at 6 months to 44.2% at 1 year for a 30-day gap in treatment. This difference resulted in partbecause the proportion of switchers increased with thelength of time from the initial prescription (at 6months, only 56.5% of the switchers had changed theirstatin treatment) (Table 2). Second, most switchersfilled 4 or more prescriptions (based on an MPR level of≥ 0.4). This level of compliance is to be expectedbecause switchers need to be sufficiently motivated tostay on treatment long enough to undergo a change intherapy. Those who lack this minimum level of motivationand consider stopping treatment after 1 or 2 prescriptionsare unlikely to try a different statin. Despitethis, the mean MPR and the likelihood of complianceand persistence were lower for switchers than nonswitchers,lending support to the idea that compliancedrops after switching drugs.
The decision to switch to a new drug for the treatmentof hypercholesterolemia is motivated by severalfactors. A physician may observe that a course of statintreatment has failed to lower cholesterol sufficiently, ora patient may experience intolerable adverse effects andask for another drug. Some patients may also try tolower their out-of-pocket cost byswitching to a drug with a lowercopayment. For these patients,switching translates into low complianceand persistence. Patientswho switched were 18.9% lesslikely than nonswitchers to maintainsufficient compliance toobtain the full benefit of statintreatment, as measured by theMPR. Patients who switched werealso 20.9% to 48.3% more likely todiscontinue treatment (< .001),depending on the gap length usedto define discontinuation and theduration of follow-up.
There is evidence that lowercompliance and persistence amongpatients who switch statins can be mostly attributed tothe effect of switching drugs. As time from the initialprescription passes, an increasing proportion of patientstaking statins switch drugs. If the switch is detrimentalto persistence, we should observe a widening gapbetween switchers and nonswitchers. This is preciselywhat happens, with the odds of discontinuation within ayear being at least 25% greater than within the first 6months. Moreover, patients need to maintain a level ofcompliance before switching to a new drug. Patients whoseriously try to follow their treatment plan are far morelikely to fill a prescription for another statin thanpatients who stop their treatment altogether. This patternwas clear in the analysis of compliance herein, withfewer switchers than nonswitchers filling less than 4 prescriptionsduring follow-up.
The decision to switch drugs does not appear to bemotivated by cost considerations. Patients who switchedto a new drug saved $1.33 per prescription on average,which is too small an amount to act as a significantincentive for most patients.
Our results suggest that switching statins, while animportant tool of drug therapy for lipid management,should be done with caution. Physicians should beaware of the compliance risks that accompany switchingand give special attention to patients undergoing achange in statin therapy.
Regarding future research, more extensive datawould provide additional insight into drug utilizationbehaviors. Limitations of our data did not allow us toevaluate the effect of nondrug treatments such as programsto encourage lifestyle changes. Interventions topromote lifestyle changes are important first steps inlipid therapy and may be partial substitutes for drugtreatments. Accounting for treatment options wouldallow us to better assess the effect of switching drugs oncompliance.
Comprehensive information about patients' socioeconomicbackground, unavailable to us, would alsoimprove our understanding of the effect of switchingdrugs on utilization and adherence patterns. For example,low-income patients may reduce their out-of-pocketcosts by using free samples or by filling prescriptionsfor higher dosages and splitting these pills. Free samplesmay also be used to try out new medications, alteringthe observed drug switching rates and compliancelevels.
Further research is also needed to evaluate the longtermeffect of drug switching on compliance. Our studyincluded a short follow-up, restricting our analysis tothe evaluation of patients who switched drugs within afew months of treatment initiation.
Finally, the prescription utilization patterns of populationswho are not reflected in our pharmacy claimsdatabase are also of interest. Because our study sampleconsisted of continuously eligible commercially insuredpatients, the results may not be generalizable to otherpopulations such as retirees or patients transitioningbetween employer-sponsored insurance plans. Therefore,caution should be used when applying theseresults to other populations.
From the Department of Pharmaceutical Economics and Policy, School of Pharmacy,University of Southern California, Los Angeles (PT, MBN); and MedImpact HealthcareSystems, Inc, San Diego, Calif (BVP, DMB).
This research was conducted as part of postdoctoral work at the University of SouthernCalifornia (PT) and was made possible by a grant from Pfizer Inc, New York, NY. Pfizer Inchad no control over the collection, analysis, or interpretation of data or the article resultingfrom this research. The authors have unrestricted publication rights.
Address correspondence to: Patrick Thiebaud, PhD, Department of PharmaceuticalEconomics and Policy, School of Pharmacy, University of Southern California, 1540 EastAlcazar Street, CHP 140, Los Angeles, CA 90033. E-mail: firstname.lastname@example.org.
Morbidity and Mortality: 2004
Chartbook on Cardiovascular, Lung, and Blood Diseases.
1. National Heart, Lung, and Blood Institute. Bethesda, Md: NationalHeart, Lung, and Blood Institute. May 2004.
2. American Heart Association. Heart disease and stroke statistics: 2005 update.Available at: http://www.americanheart.org/downloadable/heart/1105390918119HDSStats2005Update.pdf. Accessed January 10, 2005.
3. Expert Panel on Detection, Evaluation, and Treatment of High BloodCholesterol in Adults. Executive Summary of the Third Report of the NationalCholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, andTreatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). 2001;285:2486-2497.
J Am Coll Cardiol.
4. Grundy SM, Cleeman JI, Bairey Merz CN, et al; Coordinating Committee ofthe National Cholesterol Education Program. Implications of recent clinical trialsfor the National Cholesterol Education Program Adult Treatment Panel IIIGuidelines. 2004;44:720-732.
5. Scandinavian Simvastatin Survival Study (4S) Group. Randomized trial of cholesterollowering in 4444 patients with coronary heart disease. 1994;344:1383-1389.
N Engl J Med.
6. Sacks FM, Pfeffer MA, Moye LA, et al; Cholesterol and Recurrent Events TrialInvestigators. The effect of pravastatin on coronary events after myocardial infarctionin patients with average cholesterol levels. 1996;336:1001-1009.
N Engl J
7. Long-term Intervention With Pravastatin in Ischaemic Disease (LIPID) StudyGroup. Prevention of cardiovascular events and death with pravastatin in patientswith coronary heart disease and a broad range of initial cholesterol levels. 1998;339:1349-1357.
8. Sever PS, Dahlof B, Poulter NR, et al. Prevention of coronary and stroke eventswith atorvastatin in hypertensive patients who have average or lower-than-averagecholesterol concentrations, in the Anglo-Scandinavian Cardiac OutcomesTrial-Lipid Lowering Arm (ASCOT-LLA): a multicentre randomised controlled trial.2003;361:1149-1158.
J Gen Intern
9. Ellis JJ, Erickson SR, Stevenson JG, Bernstein SJ, Stiles RA, Fendrick AM.Suboptimal statin adherence and discontinuation in primary and secondary preventionpopulations: should we target patients with the most to gain? 2004;19:638-645.
10. Avorn J, Monette J, Lacour A, et al. Persistence of use of lipid-lowering medications:a cross-national study. 1998;279:1458-1462.
Am J Manag Care.
11. Hiatt JG, Shamsie S, Schectman G. Discontinuation rates of cholesterol-loweringmedications: implications for primary care. 1999;5:437-444.
12. Jackevicius CA, Mamdani M, Tu JV. Adherence with statin therapy in elderlypatients with and without acute coronary syndromes. J2002;288:462-467.
13. Joshua SB, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Longtermpersistence in use of statin therapy in elderly patients. 2002;288:455-461.
J Manag Care Pharm.
14. Abughosh S, Kogut SJ, Andrade SE, Larrat P, Gurwitz JH. Persistence withlipid-lowering therapy: influence of the type of lipid-lowering agent and drug benefitplan option in elderly patients. 2004;10:404-411.
15. Heeshen C, Hamm CW, Laufs U, Snapinn S, Bohm M, White HD. Withdrawalof statins increases event rates in patients with acute coronary syndrome.2002;105:1446-1452.
Arch Intern Med.
16. McDermott MM, Schmitt B, Wallner E. Impact of medication nonadherenceon coronary heart disease outcomes: a critical review. 1997;157:1921-1929.
17. Wei L, Wang J, Thompson P, Wong S, Struthers AD, McDonald TM.Adherence to statin treatment and readmission of patients after myocardial infarction:a six year follow up study. 2002;88:229-233.
J Clin Epidemiol.
18. Gregoire JP, Moisan J, Guibert R, et al. Determinants of discontinuation ofnew courses of antihypertensive medications. 2002;55:728-735.
19. Catalan VS, LeLorier J. Predictors of long-term persistence on statins in a subsidizedclinical population. 2000;3:417-426.
Med J Aust.
20. Simon LA, Levis G, Simon J. Apparent discontinuation rates in patients prescribedlipid-lowering drugs. 1996;164:208-211.
Curr Med Res Opin.
21. Riesen WF, Darioli R, Noll G. Lipid-lowering therapy: strategies for improvingcompliance. 2004;20:165-173.
Am J Cardiol.
22. Jones PH, Davidson MH, Stein EA, et al; STELLAR Study Group. Comparisonof the efficacy and safety of rosuvastatin versus atorvastatin, simvastatin, andpravastatin across doses (STELLAR Trial). 2003;92:152-160.
Am J Cardiol.
Am J Cardiol.
23. Jones P, Kafonek S, Laurora I, Hunninghake D. Comparative efficacy study ofatorvastatin versus simvastatin, pravastatin, lovastatin, and fluvastatin in patientswith hypercholesterolemia [published correction appears in 1998;82:128]. 1998;81:582-587.
24. Schuster H. Improving lipid management: to titrate, combine or switch? 2004;58:689-694.
25. Fishman PA, Goodman MJ, Hornbrook MC, et al. Risk adjustment using automatedambulatory pharmacy data: the RxRisk model. 2003;41:84-99.26. Imbens GW. The role of the propensity score in estimating dose response functions.2000;87:706-710.