`
`Relationship Between Daily Dose Frequency and
`Adherence to Antihypertensive Pharmacotherapy:
`Evidence from a Meta-Analysis
`
`Michael Iskedjian, BPharm,MSc,1* Thomas R. Einarson, PhD,“
`Linda D. MacKeigan, PhD} Neil Shear, MD, FRCPC,“
`Antonio Addis, PharmD,‘ Nicole Mittmann, PhD,5 and
`A. Lane Ilersich, MSc7
`
`[Phannldeas Research & Consulting Inc, Oakville, 2Faculty of Pharmacy, ‘lGraduate
`Department of Health Policy, Management. and Evaluation, Faculty of Medicine,
`4Department of Clinical Pharmacology, University of Toronto, jDepartment of Clinical
`Pharmacology, Sunnybrook and Women ’S College Health Science Centre, Toronto,
`Ontario, Canada, 5Public Health Department, Modena, Italy, 7Roche Canada Inc,
`Mississauga, Ontario, Canada
`
`ABSTRACT
`
`Background: Rates of patient adherence (compliance) to pharmacotherapy range from
`<5% to >90%. Negative determinants include multiple daily dosing (MDD), chronic du—
`ration, and asymptomatic disease. Reports suggest that once—daily (QD) dosing may im-
`prove adherence, but their findings are inconclusive.
`Objective: The purpose of this study was to compare the rates of adherence with QD,
`twice-daily (BID), and MDD antihypertensive drug regimens.
`Methods: MEDLINE, Embase, and International Pharmaceutical Abstracts databases
`
`were searched to identify comparative trials of patient adherence to antihypertensive med—
`ication in solid, oral formulations. Data were combined using a random—effects meta—
`analytic model.
`Results: Eight studies involving a total of I ”1,485 observations were included (1830 for
`QB dosing, 4405 for BID dosing, 4147 for dosing >2 times daily [>BID], and 9655 for
`
`MDD), in which the primary objective was to assess adherence. The average adherence
`rate for QB dosing (91.4%, SD = 2.2%) was significantly higher (2 = 4.46, P < 0.001)
`than for MDD (83.2%, SD = 3.5%). This rate was also significantly higher (2 = 2.22, P =
`0.026) than for BID dosing (92.7% [SD = 2.3%] vs 87.1% [SD = 2.9%]). The difference
`in adherence rates between BID dosing (90.8%, SD = 4.7%) and >BID dosing (86.3%,
`SD = 6.7%) was not significant (Z = 1.82, P = 0.069).
`
`
`*At the time this research was performed, Michael Iskedjian was a student of the Graduate Department of Phar-
`maceutical Sciences at the University of Toronto.
`
`Accepted for publication January 28, 2002.
`Printed in the USA. Reproduction in whole or part is not permitted.
`
`302
`
`(ingestion/$19.00
`
`MYLAN - EXHIBIT 1024
`
`MYLAN - EXHIBIT 1024
`
`
`
`M. ISKEDJIAN ET AL.
`
`Conclusions: The results of this meta-
`
`analysis demonstrate that with antihyper—
`tensive medications, QD closing regimens
`are associated with higher rates of adher-
`ence than either BID or MDD regimens.
`Key words: adherence, patient compli-
`ance, dosing frequency, daily dose, hyper-
`tension, antihypertensive therapy, multiple
`daily dosing. (Clin Ther: 2002;24:302—316)
`
`INTRODUCTION
`
`Medication adherence has been defined as
`
`“the extent to which a person’s behavior
`in terms of. . .taking medications. . .coincides
`with medical advice.”I Nonadherence can
`
`lead to detrimental outcomes, including
`relapse of the disease being treated, nurs-
`ing home admission, hospitalisation,2 and
`increased morbidity (eg, increase in rela—
`tive risk of coronary heart disease3) and
`mortality. Conversely, increased adher-
`ence has the potential to improve treat-
`ment outcomes.
`
`Haynes and coworkers4 compiled a list
`of >250 factors that may affect patient ad-
`herence and classified these factors as
`
`modifiable or nonmodifiable. One non-
`
`modifiable factor is the asymptomatic na—
`ture of a disease (eg, hypertension). Lack
`of symptoms is an insidious factor asso-
`ciated with patients’ forgetting about or
`ignoring their disease condition. Drug reg-
`imen complexity, on the other hand, is a
`modifiable factor. It consists of 3 major
`components—the number of medications
`prescribed, daily dosing frequency, and
`complexity of administration (eg, par-
`enteral vs oral). Hence, it may be possible
`to simplify the medication profile or re—
`duce the dosing frequency to a minimum
`to enhance medication adherence.
`
`The association between adherence to
`
`treatment and patient outcomes has been
`
`extensively investigated in the hyperten-
`sive population. Hershey and coworkers5
`demonstrated a positive correlation be-
`tween adherence and blood pressure con-
`trol, and Eisen et alf’ established adher-
`
`ence as a good predictor of blood pressure
`control. Sackett and colleagues7 deter—
`mined that an adherence level of 280%
`
`was necessary to decrease diastolic blood
`pressure in a systematic manner. Although
`the relationship between adherence and
`clinical outcome (eg, mortality) has not
`been directly established, the relationship
`between blood pressure control and mor—
`tality has been studied. Horwitz and Hor—
`witz8 reported a mortality rate of 1.4% for
`patients who were prescribed propranolol
`and took at least 75% of their medication,
`versus a rate of 4.2% for those who took
`
`<75% of their medication.
`
`An initial survey of literature reviews
`of adherence to drug therapy failed to
`clearly identify the association between
`simplified dosing regimen and increased
`rate of adherence. Blackwell9 cited 2 stud-
`
`ies reporting negative effects of multiple
`daily dosing (MDD) on adherence, 2 stud-
`ies reporting positive effects, and 2 stud-
`ies reporting mixed effects. Haynes10
`reviewed several studies reporting a neg-
`ative association between frequency of
`dosing and adherence, and 3 studies re—
`porting no association. Reid11 and Berg
`and colleagues12 could not reach a defin-
`itive conclusion, based on reviews of pub-
`lished studies, that the simplification of a
`treatment regimen could improve adher—
`ence. Overall, reviews of the literature
`have failed to reach consensus on the as-
`
`sociation between adherence and daily
`
`dose frequency.
`The present study used meta-analysis
`to examine the relationship between daily
`dosing frequency and patient adherence
`
`303
`
`
`
`to antihypertensive drug therapy, and to as-
`sess whether a lower daily dose frequency
`is associated with higher adherence to anti-
`
`hypertensive pharmacotherapy. The spe-
`cific study questions addressed were (Q1)
`
`whether once-daily, or QD, dosing is as-
`sociated with higher adherence rates than
`MDD; (Q2) whether QD dosing is asso—
`ciated with higher adherence rates than
`BID dosing; and (Q3) whether BID dos-
`
`ing is associated with higher adherence
`rates than dosing >2 times daily (>BID).
`
`METHODS
`
`We searched the MEDLTNE, Embase, and
`International Pharmaceutical Abstracts
`
`(IPA) databases for articles published in
`
`English or French between 1980 and 1998
`
`using the key words compliance, non-
`compliance, adherence, nonadherence,
`drug, drug therapy, drug treatment, hy—
`
`pertension, blood pressure, and study or
`trial. A manual search was also performed
`on all references from retrieved articles
`
`and from review articles identified in the
`
`initial literature search, as well as text-
`
`books on the topic.
`We identified all primary studies that
`compared rates of adherence between dif-
`ferent dosing frequencies of a drug regi-
`men. We included any type of research
`design that involved a comparison, in—
`cluding prospective trials (cg, randomized
`
`controlled trials or cohort studies), retro-
`
`spective chart reviews, and database analy—
`ses. Blinding/masking was not mandatory,
`but was noted. Any published study using
`an instrument to measure patient adher-
`
`ence was considered acceptable. However,
`studies must have used the same instru-
`
`ment to measure adherence in each com-
`
`parison group and also have reported rates
`of adherence to chronically administered
`
`304
`
`CLINICAL THERAPEUTICS‘D
`
`medications (ie, 210 weeks’ duration) in
`
`solid, oral formulations (ie, tablets or cap-
`sules) to treat essential hypertension in
`adults 218 years of age.
`
`Published abstracts or posters from
`symposia or colloquia were excluded.
`Also excluded were studies that dealt ex—
`
`clusively with very old patients (>74 years
`of age) since factors unrelated to closing
`frequency (eg, memory loss or confusion
`
`experienced by many of these individu-
`als”) could have. influenced the findings.
`The inclusion criteria were kept stringent
`
`enough to capture comparative studies in
`the same therapeutic area and to avoid the
`possible introduction of bias from non—
`comparative trials or from trials compar-
`ing different therapeutic areas.
`One investigator (Ml) screened poten—
`tial articles from the original search. Ti-
`tles and abstracts were screened to deter—
`
`mine eligibility. Potential articles were
`
`then masked by differential photocopying
`and by removing all identifiers such as
`names of authors, institutions, sponsors,
`and journals, as well as publication date.
`After training and practice to ensure in—
`terrater reliability, each paper was re-
`
`viewed by 2 experienced judges (AA. and
`N.M.), with disagreements settled by a
`third reviewer (A.L.I.). Evaluations of ac-
`
`ceptability criteria were recorded on a
`checklist. Data were extracted from each
`
`selected article by 2 reviewers, who en-
`tered the data onto a collection form. Dis-
`
`crepancies were again arbitrated by the
`third reviewer.
`
`For each eligible study, the effect size
`was calculated as the difference between
`
`adherence rates (Pl — P2), where P1 was
`the proportion of adherent patients taking
`medication on 1 closing regimen (eg, QD)
`and P2 was the proportion using another
`regimen (eg, BID or MDD). Data were
`
`
`
`M. ISKEDJIAN ET AL.
`
`combined using a random-effects model
`as originally described by Cochran.l4
`Differences in rates of adherence were
`
`calculated between (1) Q1) dosing and
`MDD regimens, (2) QD and BID dosing
`regimens, and (3) BID dosing and >BID
`dosing regimens. In the primary analyses,
`adherence was defined as the proportion
`of patients who had taken 280% of doses.
`If this outcome measure was not avail-
`
`able, the main adherence outcome as re—
`
`ported by the authors was used in the pri-
`mary analysis.
`
`All articles included in the meta-analysis
`were reviewed for characteristics such as
`
`publication year, study design, drug class,
`study duration, and adherence definition
`and measurement method. This examina—
`
`tion was performed for further catego—
`rization of studies for subgroup analyses
`according to common characteristics.
`Subgroup analyses were performed, with
`subgroups identified a priori according to
`the following variables: method of mea-
`
`suring adherence, definition of adherence
`(ie, using 90% and 80% as minimum ac—
`cepted rates"), study design (ie, prospec-
`tive vs retrospective), medication class
`(cg, calcium channel blockers), and dura-
`
`tion of treatment (ie, 3—6 months vs 12—24
`
`months). Sensitivity analyses included re—
`analysis that excluded apparent outliers.
`Homogeneity of effects was examined
`using a chi—square test. In addition, rates
`were plotted against each other to identify
`obvious outliers, as suggested by L’Abbé
`et al,15 and regression analysis was used
`to confirm those observations, according
`to the method described by Tiku et al."’
`The quality of the accepted articles was
`evaluated using a quality checklist adapted
`from Haynes et al.4 The checklist exam-
`ined 6 aspects of the article, including study
`design, selection and specification of the
`
`study sample, specification of the illness or
`condition, adherence measure used, de-
`
`scription of the therapeutic regimen, and
`definition of adherence. The total possible
`score was 17 points; articles rated 28.5
`(50%) were considered to be acceptable.
`Quality ratings were determined as for data
`extraction by 2 reviewers, with discrepan-
`cies arbitrated by the third reviewer.
`
`RESULTS
`
`An initial literature search yielded 871
`potential articles. The investigators screened
`these articles by reading through their ti—
`tles and abstracts to eliminate those that
`
`were obviously inappropriate for this re-
`search, and to compile a shorter list to be
`assessed for inclusion. This screening re—
`sulted in a list of 34 articles possibly con—
`taining pertinent information for the meta-
`analysis. Of these, a total of 8 articles”’24
`were selected in the review and selection
`
`process described previously.
`Seven articles with 4669 observations
`
`(number of patients, doses, or other mea-
`
`sure, as reported by authors) were used in
`the analysis of QD dosing versus MDD; 5
`studies with 2152 observations were in-
`
`cluded in the analysis of QD versus BID
`dosing; and 4 articles with 7926 observa-
`tions were used for the analysis of BID
`dosing versus >BID dosing. The respec-
`tive numbers of observations were 1830
`
`for QD dosing, 4405 for BID dosing, 4147
`for >BID dosing, and 9655 for MDD, for
`an overall total of 11,485 observations.
`
`Tables I and Il summarize the major
`characteristics of the 8 selected articles,
`
`including sample sizes, reported adher—
`ence rates, definitions used for adherence,
`
`patient characteristics, study design, drug
`class, type of therapy, and adherence mea-
`surement methods.
`
`305
`
`
`
`
`
`
`
`Exama5:551Eastman882.6£3,
`
`CLINICAL "[‘l-IEIUKPEUTICSm
`
`
`
`3%2E:93%63%£5.56Egan—Ea
`
`23%?can«ww—
`
`8%
`
`
`
`
`
`HEBcountumui035$3£39mbeowaE6233%N3306552$50uEGQQShoM£32.35555$
`
`
`
`mafia—2Savages—05
`
`
`
`33525Egan038$3.EaumG0mm:85Bucca—m@336582$50uéuommofi339.EBBmflom
`
`
`
`Em:53:EE3»omown:muEionvemf3303NH:362$50353%on$me.1“EESQ
`
`
`
`
`
`
`
`EsouEm.ufiE§mm.maumcmown:35BEESmafiufiocoi£0035385:89.38ENEQEI
`
`
`
`
`
`
`
`
`
`.mEBmmmgig—SE:555352vanmEmE”BEBE5:5582.38HmeU
`
`
`
`
`
`
`
`
`
`HEP.countomoiomAQua,3qummownan:2308NHmcfifi83$8Esmmfiom03688.:qugammaAm6583:
`
`
`
`
`
`
`
`5:55
`
`munEEEOQM35
`
`
`
`
`
`REESEE239Emucusmmv330532:magnum“
`
`can»_A
`
`5+350%$556
`
`
`
`
`
`2:95.85$3.960»Eown5.6222:88ouBmoEEEZmmaw25E»?03838.528$.35cuflm
`
`
`
`
`
`
`
`
`
`was:58Em23»cmemu33235065“oz3&5“amen3055ofiuuommohmEmma.Eom93E3
`
`
`
`
`
`
`
`
`
`
`
`:58:5d—mEflown£89mmm0mm:323095:—Ccmufioficco—Zmuuxuofigfiumu>fioummoiSwaganEBaum
`
`
`
`
`
`
`
`
`
`
`
`@0505muflmtuuumEF—UEdwin—om:BESQREEFmo09?AmuvmmEUmam=mfimflmvammoauhumom
`
`
`
`
`
`Eon—2:302ESE
`
`
`
`
`
`.moumunaomaao33mafibgméuufi05Evuufiofi86::weSEE—am.H03$
`
`306
`
`
`
`M. ISKEDJIAN ET AL.
`
`
`
`
`
`
`
`8:56mg3:38:088¢maunamuu5:.»+3de855:?Btomom”mdwbumwépofi2.:Enova—ES8:5?hobafifinm.=2a.?
`
`
`
`
`
`
`
`35m035mm”.33“8:80:04
`
`
`3:20:1qumo.92
`
`
`
`Eng3%3E.23:2:Ea5:33%82:NA351:8
`
`
`
`
`
`mon:EmnEmQ0SEER.2253030ooauéom
`
`mz3+3“no36335am.mm:88a»3«Sammags2:23a.333V
`
`m23+3”m0womenwe
`
`33v3v33%3V35%3V33.32:figEwoman3mzE:033.3E33.83“3%E9.35.8332m$53353%.33+392ExamEaif
`
`85v3+3”mo38SE:a5383338+852593:3£83pm.33
`
`
`
`m23+3“3game3v33%33$0.:3333%338353awas“.3EEE9
`
`36v3+3:033.8339%3$333$3.?33:28358258cE33Easacam
`
`mo53:5:2“Sum—u.—@3052
`
`
`
`Enigmamemo—u
`
`
`
` 5828—:dfihe2::8353a$8“.3ESE
`
`32:335m3.3539::2
`
`Sodv“a38.?EE32w9:22BE?38c23%.333$2in5Emma:
`
`
`
`86v”a3%.;33:m4233.89”£22BEER8+8.33,“.sawisaqueue:
`
`35:33338mmags9:23.3“Na3:
`
`women.mo
`
`
`
`
`
`adv333%.;333%?53+83£3E:0<233%3$8333mmmaze€qu33E9:.:12“.Bum
`
`
`
`
`
`
`
`3%.”a03.:33m.33$3.?03.393:2833%“3333an»«Ea;g585m
`
`86v“mo3:0Rm
`
`
`
`modv”do3::d5.33333vaBanana8833535::
`
`Atacama“:
`
`
`
`307
`
`
`
`CLINICAL THERAPEUTICS"
`
`mEEmmv2mmuuaohonv<
`
`
`
`
`oonouunfiax.E.0ZEmma
`
`53vas$0.a3w«3.53V$8.33vBataanmesa.8Hosea.
`
`SodvE“SE3«09%3$3M“Q36th£85%momm.fie
`
`EodvA3“$3.?3$3.9“g“at:A3a3.“6H38233:52
`
`
`
`Sodv3V99%.355:5:.322“SEER
`
`
`
`SodvA3mEEBE335::m0E39:
`
`
`
`BodvA3.35,052386USEfinn92638
`
`m23”No838863.95%5E853..5%E“No3V
`
`
`
`m23:0$133<29%.a3$3323922ESE:was”Eea39%as3Eva
`
`Sodv3v6&8«an:3v«3%EVmesa{Ho
`
`adv3“mo4293.3E9333<2$3Mwigwannawe§32%isaimam
`
`SEa“.8me33$gem33325:3Us§3v
`
`womenmo
`
`itsBumuommmE9:83005593..na
`-mouQB5:33de3:226.“Saw:—EtsumufluommmE95%Q0.6523,um0”Q92.553%..85.6%““Emma
`
`
`
`
`
`33Q?8%em:33HzBEL
`
`,
`
`
`
`
`
`m992Em?amno5935:ofitufioSaving
`
`
`
`
`
`602.959.:2g
`
`308
`
`
`
`.mEmouQHm—A:2:3:33:83?HERE5::dufimocmwaEmEmEU25.5521:nm0”mi
`
`
`
`
`
`
`
`
`
`
`
`.QEEEQBo?mnoufimow23:2:355:3A3BEwas:355:?m0mnoficmuu2managesP5533guano“mm3%3a..32:05Emacs.may-Ea2E:
`
`
`
`
`
`
`
`
`
`M. ISKEDJIAN ET AL.
`
`Results are first presented for primary
`analyses, then for subgroup analyses ac-
`
`cording to specific characteristics, and ti—
`nally for sensitivity analyses with regard
`to heterogeneity of reported data. Results
`are further categorized according to the 3
`comparisons of daily dose regimens.
`No outliers were detected, indicating the
`absence of heterogeneity, even when the Q
`
`test (which is known to be a weak test)
`
`results.
`yielded statistically significant
`However, the graphical representation of
`trend lines identified 1 study, by Fujii and
`Selti,l8 that consistently produced results
`that lay farthest from the trend line. Al-
`though the other tests did not identify the
`study by Fujii and Seki18 as being an out—
`lier, it was considered an apparent outlier
`
`since the chi-square test was significant.
`
`Sensitivity analyses were then performed
`whenever the chi—square test was statisti—
`cally significant, discarding the study by
`Fujii and Selti18 from the previous analysis.
`All 8 articles were given a quality rat-
`ing >90 (range, 9.5—16) and were hence
`considered to have acceptable quality. The
`correlation coefficient was small and non-
`
`significant between quality scores and
`year of study (Spearman’s rho = +0.10, df =
`6, P = 0.81), as well as between quality
`scores and difference in adherence rates
`
`(Spearman’s rho = 0.23, df= 6, P = 0.59).
`All available data were pooled for the
`primary analyses (ie, Q1, Q2, and Q3), al—
`
`though all studies were not of the same
`duration and did not use the same defini—
`
`tion of adherence or measurement method.
`
`However, the treatment and control groups
`had similar basic characteristics within
`
`each study; for example, results reported
`for the QD regimens versus the MDD reg-
`imens within each study were based on
`the same design, duration, adherence def-
`inition, and measurement method.
`
`Table III summarizes the results of all
`
`analyses. The meta-analytic estimate of
`the difference between adherence rates for
`
`QD dosing and MDD regimens was sta-
`tistically significant (P < 0.001), that is,
`QD dosing was associated with a higher
`adherence
`rate
`(91.4%)
`than MDD
`
`(83.2%). The meta—analytic difference be—
`
`tween adherence rates for QD dosing
`(92.7%) and BID dosing (87.1%) was also
`statistically significant (P = 0.026), al-
`though the difference in this analysis was
`
`smaller
`
`than in the QD—versus»MDD
`
`analysis (5.7% vs 8.2%).
`The difference in adherence rates be—
`
`tween BID (90.8%) and >BID dosing
`(86.3%) was not statistically significant
`(P = 0.069). However, a subgroup analy—
`sis using a stricter definition of adherence
`(290% intake) did reveal a statistically
`significant difference between BID and
`>BID dosing (respective adherence rates
`of 76.1% and 67.0%, P < 0.001). Fur—
`thermore, statistical significance would
`be achieved by adding 1 hypothetical
`study of average size to the existing sum-
`mary results. Hence, the lack of signifi-
`cance may be due to a lack of statistical
`
`power.
`
`All subgroup analyses between QD
`dosing and MDD confirmed the statisti-
`cally significant difference found in the
`
`primary analysis. The greatest difference
`(15.4%) was found in the analysis using
`stricter definitions of adherence, with
`
`82.9% of patients on QD regimens versus
`67.6% of patients on MDD regimens hav-
`ing an acceptable level of adherence. In
`the analysis using 80% adherence as the
`minimum acceptable level, the difference
`between QD dosing and MDD was 10.0%,
`
`with respective proportions of adherent
`patients of 95.6% and 85.7% (Table III).
`
`The same difference (10.0%) was ob—
`
`309
`
`
`
`CLINICAL THERAPEUTICS‘“
`
`
`
`83vm:23mmme?nn52m>no:0Ewarm:gag55E
`
`
`
`as:Z:9toENewmmm.a,no505H“5.25SEE:bani;
`
`
`
`$339me.am8%qan9»EmSoé$3235523$53
`
`
`
`
`
`SodvQ:93ms8%hon:.0.»no335“Eva855%.“us8:52,35:5
`
`
`
`Sodwe2omv.2Mam.wEmEno355@8&855:2ESusana533m
`
`
`
`so?393.Wm82wEm."E,EmESE958$8:23?EEEEE55ch
`
`so?mi239.23amwan:m»sowas”.EQ.2.3.59855:3
`
`83vmg2mm3:a2mon:a,no£35gowm5528.3%msfix.a3:3855:3
`
`
`
`
`
`adv932m:gcm:WQB;9’no@5393basin5mmanefiab8may“03338“;
`
`
`
`53vWm3md382mBanSEmb822:25anon
`
`53v2:2Eq:EamQ02E00b825E£30350
`
`
`
`m8:0.:eno33%.wan:9,no:5was89.52
`
`23w:9SE82qn52a,nor5Bane:58Eamew:
`
`
`
`
`
`wood3:9mm2:m:NQ522,no3:8aufiuualfifi28%on
`
`53v”.323.2.$21992a,mo832%M582w9m3“RE.
`
`
`
`advv.22ms2:mmmmN82a»no:58E2Suafimoalzmfi35%23
`
`36mi2S.3.5%mODEz,oo8:53E32:33fl,822¢
`
`
`
`
`
`mu6fimmHR$333930mommamEofiwmumEwh_m=<
`
`
`
`
`
`033.3930“:
`
`3:22:5‘EBahama
`
`iMo.02DE.cZUuEmEoU
`
`
`
`
`
`.wombmcmgame;6:3Sfitm5m3:300:20:?:35Emuofizmfimuuubmaméoamo$885.52an
`
`
`
`
`
`
`
`310
`
`
`
`
`
`.553RENE£3;3338?.m.“mime—uQ0BEERH#0uBfioE.uucmauE323umUUSEE8E:NAmime—VHOH?"mEmovbanEmEseMDDS
`
`
`
`
`
`
`
`
`
`Sr:@828anEwEmouDE5E2?um0memavhamE2:3::355:?EsmE55@2283.»2wEwo—uno55053MNO“DOE.35mazecoco
`
`
`
`
`
`
`
`
`
`.mEmowmum—A:2:£558555552mm:
`
`
`
`M. ISKEDJIAN ET AL.
`
`served in the analysis based on adherence
`as a percentage of the dispensed doses.
`The smallest difference was observed in
`
`the subgroup analysis based on all studies
`except those in which the patients took
`calcium channel blockers (6.1% differ-
`
`ence, adherence rates of 92.4% for MDD
`
`vs 86.1% for QB dosing, P = 0.032). The
`mean adherence rates decreased from
`
`93.8% to 86.7% for MDD and 85.7% to
`
`80.0% with duration of therapy (3—6
`
`months vs 12724 months).
`
`One subgroup analysis of QD versus
`BID dosing, which was based on prospec-
`tive trials using a stricter definition of ad—
`herence, yielded a P value of 0.07, for a
`difference in adherence rates of 6.0%.
`
`Three additional sensitivity analyses of
`QD versus BID dosing, performed after
`
`discarding an apparent outlier, yielded sta-
`tistically significant differences ranging
`between 5.2% and 8.6% (P < 0.0l).
`
`DISCUSSION
`
`At the time of completion of this study, 1
`review study in the literature focused ex-
`
`clusively on the relationship between daily
`dosing frequency and adherence.25 The
`author performed a meta—analysis com-
`bining data from single arms of different
`studies and concluded that QD and BID
`
`regimens were associated with signifi-
`cantly better adherence than TlD regi-
`
`mens; however, his methods were ques-
`
`tionable. Furthermore, from a total of 57
`
`studies evaluated, only 36 (63%) were
`
`limited to a single agent. Studies included
`in the review combined acute and chronic
`
`treatments, pediatric and adult popula-
`tions, and symptomatic and asymptomatic
`diseases. Moreover, the definition of ad-
`
`herence varied from study to study, as did
`measurement methods. Studies included
`
`in that meta—analysis did not necessarily
`
`compare the adherence rates between dif—
`ferent daily dose frequencies. Most re—
`ported rates were extracted from studies
`assessing the rate of adherence for only 1
`dosage frequency without comparing it to
`adherence rates associated with other
`
`dosage frequencies.
`In a more recent publication, Claxton
`et al26 reported on adherence rates as mea—
`sured by electronic monitoring devices in
`
`76 studies. Adherence rates ranged from
`51% to 79%, on average, across various
`therapeutic areas, and the investigators
`
`found that adherence was inversely re—
`lated to daily dose frequency, consistent
`
`with the results of the present study. How—
`ever, the adherence rates reported by Clax-
`ton et al were lower than those calculated
`
`in the present meta—analysis. Moreover,
`they reported statistically significant dif-
`ferences only in the comparisons between
`QD and TID regimens, between QD and
`QID regimens, and between BID and QID
`regimens. Their report was an analysis of
`a heterogeneous data set consisting of ad-
`
`herence rates across various therapeutic
`
`classes of drugs that also included data
`from noncomparative trials.
`The results of the present meta-analysis
`suggest that QD regimens are associated
`with greater adherence to antihyperten-
`
`sive pharmacotherapy. However, the ad-
`
`herence rate associated with QD regimens
`
`ranged from 82.9% to 95.6%, indicating
`that a simple QD regimen alone may not
`result in adequate compliance (ie, 100%
`medication intake or 100% of patients tak—
`ing a sufficient amount of medication [eg,
`8096]). Moreover,
`the medical conse-
`
`quences may be more grave for those pa*
`tients failing to adhere to QD regimens,
`
`since missing a dose results in missing
`the total daily dose.
`
`311
`
`
`
`CLINICAL TI-HERAPEUTICS”
`
`regimen to a QD regimen to produce 1
`additional adherent patient. However, for
`this type of analysis, the use of adherence
`
`rates as a percentage of patients rather
`than doses would be more appropriate.
`Therefore, the NNS would estimate, for
`
`the number of patients with
`example,
`<80% adherence needed to switch. This
`
`estimate would be 110.10, or 10 patients
`to switch from MDD to QD dosing to
`achieve acceptable adherence in 1 addi-
`tional patient. In the primary analysis of
`QD versus BID dosing, the NNS would
`then be 1/0057 or 18 patients, and for
`
`BID versus >BID dosing, the NNS would
`be 1/0032 or 31 patients.
`Although the NNS values obtained for
`QD versus MDD or BID dosing are
`smaller and seem more attractive com-
`
`pared with the higher NNT numbers for
`
`the prevention of cardiovascular failure
`and mortality with hypertensive pharma-
`cotherapy, it is important to note that ad-
`
`herence is only an intermediate outcome
`and may need to be improved in larger
`numbers of patients to achieve blood pres-
`sure control and prevent cardiovascular
`and cerebrovascular events.
`
`Assumptions and limitations
`
`A meta-analysis has inherent biases be-
`cause it combines data from different stud-
`
`ies that may have different sample sizes,
`study designs, outcome definitions, and
`other study parameters.
`In this meta-
`analysis, all variables that could affect ad-
`herence, other than daily dDSe frequency,
`
`were assumed to be equal between com—
`parators, a situation that may not hold true
`in the real world. In an attempt to control
`confounding factors, this meta—analysis
`targeted hypertension, a chronic asymptom—
`atic disease.
`
`One way of establishing the clinical rel—
`evance of an intervention is by calculating
`the “number needed to treat” (NNT). The
`
`NNT is the number of patients required in
`the active treatment group to have 1 addi-
`tional patient with a successful outcome
`at the end of the trial or procedure com-
`pared with the number of patients given
`placebo or a comparator drug.27 The NNT
`is often used in studies of preventive ther-
`apies.”29 For example, authors report an
`NNT of 5 for the use of ondansetron to
`
`prevent postoperative nausea and vomit-
`ing,28 whereas a much higher NNT of 48
`was calculated for the use of chlorthali-
`
`done and atenolol for prevention of heart
`failure.29 A lower NNT is preferable; how—
`ever, a larger NNT may be acceptable
`when the goal is to prevent more serious
`disease conditions. The NNT may also
`vary within subgroups of patients receiv—
`ing the same type of therapy. For exam-
`ple, in a study of US women 230 years of
`age receiving antihypertensive pharma-
`cotherapy for 5 years, an NNT of 282 was
`estimated for the reduction of cardiovas—
`
`cular mortality, whereas an NNT of 32
`
`was estimated for the prevention of mor—
`tality among African-American women.30
`To quantify the clinical implications of
`an improvement in adherence by switch-
`ing from a higher to a lower daily dose
`frequency, one could calculate the “num-
`ber needed to switch" (NNS): for exam-
`
`ple,
`the number of patients needed to
`switch from an MDD regimen to a QB
`regimen to avoid 1 nonadhercnt patient
`(or to achieve adherence in 1 additional
`
`patient).
`In the primary analysis of QD dosing
`versus MDD, the difference in adherence
`
`was 8.2%; hence,
`
`the NNS would be
`
`ll0.082 or 12 patients. Thus, one would
`need to switch 12 patients from an MDD
`
`312
`
`
`
`M. ISKEDIIAN ET AL.
`
`The primary analyses combined data
`from studies that used various definitions
`
`of adherence. Subgroup analyses pooled
`data from those studies that used similar
`
`or comparable definitions of adherence
`(eg, 280% of medication intake). How-
`ever, biases from other sources, such as
`
`research design or duration of therapy,
`could not be controlled in those subgroup
`analyses; that is, due to the limited num-
`ber of studies available, we could not per-
`form subgroup analyses that would con-
`trol more than 1 or 2 types of bias. Despite
`this limitation, all subgroup analyses con-
`firmed the statistically significant differ-
`
`ences between QD dosing and MDD. All
`
`subgroup analyses of QD versus BID dos-
`ing also resulted in statistically significant
`differences, with | exception (P = 0.07).
`However, the primary analysis comparing
`BID to >BID regimens failed to find sig—
`nificant differences in adherence rates,
`
`possibly due to insufficient statistical
`power. Further meta-analyses that include
`
`additional head—to-head comparative tri—
`als would be necessary to establish any
`differences between adherence rates for
`
`BID and >BID regimens.
`
`Independent of research design, indi—
`vidual clinical cases were assumed to be
`
`equal since it was not possible to differ-
`entiate between patients andfor studies
`for severity of illness, number of con-
`comitant drugs taken, or comorbid condi—
`tions, mainly because some investigators
`did not report details. Monane et 3131
`investigated the effects of concomitant
`medications and comorbid conditions on
`
`adherence and reported odds ratios for ad-
`
`herence (defined as 2.80% intake) among
`patients beginning antihypertensive ther-
`apy. A significantly larger proportion of
`patients using calcium channel blockers,
`angiotensin-converting enzyme inhibitors,
`
`or beta—blockers adhered to therapy com-
`pared with users of thiazides, regardless
`of dosage frequency. The authors also re-
`
`ported that adherence was significantly
`higher in hypertensive patients with con-
`comitant ischemic heart disease or con-
`
`gestive heart failure.31
`It was also assumed in this study that
`
`patients presented with equivalent severity
`of disease, whether they were identified as
`having chronic, essential, or mild to mod-
`erate hypertension. Hence, results can only
`be generalized with caution to actual clin-
`ical practice, where patients may present
`with any severity level of hypertension and
`
`with any comorbid condition(s).
`No duplicate studies were included in
`this meta-analysis. However, it is difficult
`to identify unpublished studies. It is also
`possible that some studies were not in-
`dexed in the MEDLINE, Embase, and IPA
`
`databases. The possibility of language bias
`
`also exists, since we included only publi-
`cations in English or French.
`In addition, although both adherence
`rates and differences in adherence rates
`
`appeared to decrease with duration of ther-
`
`apy (adherence rates of 85.7% and 93.8%
`with a difference of 9.1% for therapies
`lasting 3 to 6 months vs rates of 80.0%
`and 86.8% with a difference of 6.9% for
`
`those lasting 12 to 24 months), adherence
`rates could not be established for the very
`long term due to lack of data. Rudd and
`colleagues32 reported an annual decline of
`36% in adherence after the first year of
`
`antihypertensive treatment.
`Some other aspects of patient adher-
`ence were outside the scope of this study,
`such as the effects of timing and com—
`plexity of administration. Data on med-
`
`ication timing or intervals were not avail-
`
`able to perform a subgroup analysis of
`studies reporting time— or timing—specific
`
`313
`
`
`
`adherence rates. Such an adherence rate
`
`may be meaningful in clinical practice as
`it could enable the clinician to assess the
`
`regularity of medication intake, which in
`turn may be crucial to the success of cer-
`
`tain therapies that require strict dosing in-
`
`tervals. Complexity of administration is
`also a factor contributing to differences in
`adherence rates. Adherence rates associ-
`
`ated with the administration of transder—
`
`mal patches may differ from those seen
`with oral therapy. Adherence rates associ-
`
`ated with sublingual administration may
`even differ from those associated with a
`
`standard oral regimen. All articles in-
`
`cluded in the present meta-analysis re-
`ported adherence rates associated with
`solid, oral dosage forms to be swallowed.
`Further research is required for establish-
`ing the role of complexity of administra-
`tion in patient adherence.
`
`Another limitation to this meta-analysis
`
`is that we investigated the association be—
`
`tween daily dose frequency and patient ad-
`herence but did not perform a causal analy~
`sis. Further research is required to establish
`causation and determine the specific as-
`pects of daily dose frequency that have the
`greatest impact on patient adherence.
`The present study was also limited to
`antihypertensive pharmacotherapy, and its
`findings may not be readily generalized to
`other therapies for other disease states, al-
`though one might expect to see compara—
`ble results in other disease conditions that
`
`are both asymptomatic and chronic.
`
`CONCLUSIONS
`
`The results of this meta-analysis demon—
`
`strate that QD dosing regimens of oral an-
`tihypertensive agents are associated with
`higher adherence rates than BID dosing
`or MDD regimens. The analysis compar-
`
`314
`
`CLINICAL T