Dose Equivalents for Second-Generation Antipsychotics: The Minimum Effective Dose Method

Abstract

Background: Clinicians need to know the right antipsychotic dose for optimized treatment, and the concept of dose equivalence is important for many clinical and scientific purposes. Methods: We refined a method presented in 2003, which was based on the minimum effective doses found in fixed-dose studies. We operationalized the selection process, updated the original findings, and expanded them by systematically searching more recent literature and by including 13 second-generation antipsychotics. To qualify for the minimum effective dose, a dose had to be significantly more efficacious than placebo in the primary outcome of at least one randomized, double-blind, fixed-dose trial. In a sensitivity analysis, 2 positive trials were required. The minimum effective doses identified were subsequently used to derive olanzapine, risperidone, haloperidol, and chlorpromazine equivalents. Results: We reviewed 73 included studies. The minimum effective daily doses/olanzapine equivalents based on our primary approach were: aripiprazole 10 mg/1.33, asenapine 10 mg/1.33, clozapine 300 mg/40, haloperidol 4 mg/0.53, iloperidone 8 mg/1.07, lurasidone 40 mg/5.33, olanzapine 7.5 mg/1, paliperidone 3 mg/0.4, quetiapine 150 mg/20, risperidone 2 mg/0.27, sertindole 12 mg/1.60, and ziprasidone 40 mg/5.33. For amisulpride and zotepine, reliable estimates could not be derived. Conclusions: This method for determining antipsychotic dose equivalence entails an operationalized and evidence-based approach that can be applied to the various antipsychotic drugs. As a limitation, the results are not applicable to specific populations such as first-episode or refractory patients. We recommend that alternative methods also be updated in order to minimize further differences between the methods and risk of subsequent bias.

Introduction

Dose equivalence estimates of antipsychotics are important for many reasons. Clinicians need such information when they switch from one antipsychotic to another one or when they combine antipsychotic drugs. Trialists and meta-analysts need dose equivalence estimates for the design of fair comparisons of antipsychotic drugs,^1,2 and whenever the doses of several antipsychotics have to be converted into 1 unit. Finally, the concept is also important for cost calculations and treatment guidelines.

In a comprehensive review, Patel et al³ listed various approaches to define dose equivalence including: the classical flexible-dose chlorpromazine equivalent method published by Davis in 1974,⁴ the dose-response curve method to define near-to-maximum doses by Davis and Chen,⁵ methods based on the maximum licensed doses of the various drugs,^6,7 the concept of daily defined doses (DDDs) of the World Health Organization,⁸ and various (expert) consensus methods (eg, Kane et al⁹, Gardner et al¹⁰, Andreasen et al¹¹). Their review made it clear that a gold standard method does not exist.

In 2003, Woods¹² introduced an approach in which he identified the minimum effective doses of the 5 second-generation antipsychotics (SGAs) available at that time based on their placebo-controlled, fixed-dose studies, and he subsequently derived chlorpromazine equivalents. Patel et al³ found that it was the most frequently used method (714 cites found in a Web of Science search on July 25, 2013). Since the original article,¹² which made recommendations for 5 SGAs and haloperidol based on 15 studies, several further SGAs have been developed and multiple further placebo-controlled trials have been published. In the current systematic review, we therefore (a) refined the method by operationalizing the decision-making process, (b) updated the material, and (c) included 8 additional SGAs.

Methods

Woods¹² identified the lowest fixed dose of each SGA that was consistently more efficacious than placebo in the primary outcome (usually the mean change of the Positive and Negative Syndrome Scale [PANSS¹³] total score or the Brief Psychiatric Rating Scale [BPRS¹⁴] total score from baseline in the intention-to-treat dataset). The identified doses were then used to calculate equivalence ratios. For example, if the minimum effective dose of olanzapine was 7.5 mg/d and that of asenapine was 10 mg/d, the asenapine dose that is equivalent to 1 mg/d olanzapine would be 10/7.5 = 1.33 mg/d. We made several modifications (supplementary webappendix 1 presents Wood’s¹² original method):

1. We primarily calculated olanzapine equivalents, because the minimum effective haloperidol dose is less well understood (see “Results” section). For ease of reference, we also present haloperidol and risperidone equivalents. In keeping with Woods,¹² we did not investigate the minimum effective chlorpromazine dose, as the chlorpromazine literature is composed entirely of small, old trials,¹⁵ which differ in methodological rigor from more recent SGA trials. However, chlorpromazine equivalents were estimated based on the original finding by Davis that 1.6 mg haloperidol corresponds to 100 mg chlorpromazine.⁴ Woods assumed that 2 mg haloperidol corresponds to 100 mg chlorpromazine,¹² which was based on an APA (American Psychiatric Association) guideline statement¹⁶ and not empirical evidence.

2. Woods¹² did not operationalize the criterion “consistently superior” and sometimes only one trial was available (so that “consistency” could not be verified). Our primary criterion was that a dose was statistically significantly superior to placebo for the primary outcome in one double-blind randomized controlled trial (RCT). To explore the possibility that minimum effective doses had been missed due to insufficient power, we meta-analyzed the results of lower doses when available. Standardized mean differences (effect size [ES]) expressed as Hedges’s g were calculated together with their 95% CIs based on a fixed-effects model (the use of a random effects model did not change any result). Due to various problems in trial methodology in recent years, such as increasing placebo response^17,18 and high dropout rates,¹⁹ it is currently difficult to demonstrate statistical superiority. These developments lead to an increasing number of failed trials (studies in which neither the new drug nor the active comparator was more effective than placebo), so that even well-established compounds such as haloperidol²⁰ or olanzapine²¹ sometimes do not show separation from placebo or that a lower dose was significantly more efficacious than placebo, whereas higher or intermediate doses were not effective.^22,23 In this environment, we feel it is unlikely that superiority to placebo is just a chance finding.

   In order to examine what happens if a more conservative approach is considered, we conducted a sensitivity analysis, which required that a dose was efficacious in a second RCT. In essence, this is in keeping with a rule for the registration of a compound by the Food and Drug Administration (FDA). If a dose was supported by only one trial, the next higher dose that was efficacious in another trial qualified for the secondary criterion.

3. In contrast to Woods,¹² we also included RCTs, which used a dose of the same drug that was presumed to be ineffective by the original authors (eg, olanzapine 1 mg/d).

4. We included fixed-dose studies and trials that were defined as “fixed-dose range” studies by their authors (eg, iloperidone 4–8 vs 10–16 mg/d vs placebo or olanzapine 5 ± 2.5 vs 10 ± 2.5 mg/d vs placebo). For the latter, we conservatively assumed that the upper limit of the dose range was efficacious (Woods¹² used the midpoint of the range).

We examined the following SGAs and haloperidol in adult patients with schizophrenia or schizoaffective disorder: amisulpride, aripiprazole, asenapine, clozapine, iloperidone, lurasidone, olanzapine, paliperidone, quetiapine, risperidone, sertindole, ziprasidone, and zotepine. We excluded studies in special populations such as adolescents, elderly, first-episode patients, stable patients (mainly relapse-prevention studies), patients with predominant negative symptoms, or patients with treatment resistance (except for clozapine, which is licensed for treatment resistance), as these populations might require different doses. As a result, the findings will not be applicable to these populations.

Our literature search was mainly based on the exhaustive searches for 4 systematic reviews on SGAs by our group.^2,24–26 They included the register of the Cochrane Schizophrenia Group (which is compiled by regular systematic searches in more than 15 databases, clinical trial registers, hand searches, and conference proceedings,²⁷ available up to August 2009) and MEDLINE/PubMed, Embase, Cochrane Central, and clinicaltrials.gov (last search: September 2012). We updated the latter searches in June 2013 (for search terms, see supplementary webappendix 2). We also searched the medical reviews that pharmaceutical companies must submit to the FDA, Cochrane reviews comparing SGAs and haloperidol vs placebo,^28–30 and Cochrane reviews on optimum SGA doses.^31,32 Finally, we sent the draft results section of each antipsychotic to its manufacturer with a request for additional unpublished trials and the possibility for comments. Most manufacturers replied (see “Acknowledgments” section), which increased our confidence that no study was missed. All data were extracted by S.L. and independently verified by M.S. The analyses were conducted with Excel 2010 and Comprehensive Meta-analysis Version 2.³³

Results

Supplementary webappendix 3 shows the PRISMA (Preferred Reporting Items for Systematic Reviews) diagram of the search. A description of the 73 included studies is provided in supplementary webappendix 4. The vast majority of the studies lasted 6 weeks (n = 45) or 8 weeks (n = 14), with a range between 3 and 48 weeks and a mean/median of 6.9/6.0 weeks. Table 1 illustrates which doses separated from placebo and the findings are summarized below. Where we describe a dose as being efficacious, this dose was statistically significantly superior to placebo (or a presumed ineffective dose) in the intention-to-treat analyses of the primary outcome based on the original authors’ tests. Table 2 presents minimum effective doses and dose equivalencies.

Table 1.

Results of Placebo-Controlled, Fixed-Dose (Fixed-Dose Range) Studies

	Dose Groups in mg/d
Amisulpride	100	400	800	1200
Puech et al 34		−a	+	−
Aripiprazole	2	5	10	15	20	30
Cutler et al 35	−	−	+
McEvoy et al 36			+	+	+
Kane et al 37				+		+
RGH-MD-04 38			+
Potkin et al 39					+	+
Study 94202 20 (F)	−		−			−
Study 93202 40						−
Asenapine	0.4	0.8	1.6	3.2	4.8	10	20
Study 041-002 41	−	−	−
Study 041-013 42				−	−
Kane et al 43						+	+b
Potkin et al 44						+
Hera 041-021 45						−	−
Clozapinec	100	300	600
Simpson et al 46		+	+
Haloperidol	4	4.5	6	8	10	12	15	15±5	16	20
Zimbroff et al 47	+			+					+
Vichaiya et al 48		+
Garry et al 49			−
Simpson et al 50			−
Kane et al 43				+
Barbato et al 51					−
Meltzer et al 52					+
Kane et al 37					+
Study 049 53					−
Garcia et al 54					+
Study 94202 20					−
Arvanitis et al 55						+
Potkin et al 56							+
Study 115 57							+
Beasley et al 58								+
Zborowski et al 59									+
Study 93202 40										+
Marder et al 60										+
Chouinard et al 61										−
Klieser et al 62										−
Iloperidone	4	4–8	8	10–16	12	12–16	20–24	24
Study ILPB202 63	−		−
Potkin et al (Study 1) 56	−		−		+
Potkin et al (Study 2) 56		+		+
Potkin et al (Study 3) 56						−	+
Cutler et al 64								+
Lurasidone	20	40	80	120	160
Study 049 53 (F)	−	−	−
Ogasa et al 65		+		+
Meltzer et al 66		+		+
Nasrallah et al 22		−	+	−
Nakamura et al 67			+
Loebel et al 68			+		+
Study D1001002 69		d	d
Olanzapine	1	5±2.5	10	10±2.5	15	15±2.55
Beasley et al 70	−		+
Beasley et al 58		−		+		+
Beasley et al 71 (F)		−e		−e		−e
Kane et al 72			+f
Marder et al 73			+f
Davidson et al 74			+f
Hirayasu et al 75			+f
Meltzer et al 66					+
Barbato et al 51					+
Corrigan et al 76					+
Patil et al 77					+
AstraZeneca 78					+
Schmidt et al 79					+
Hera 041-021 45					+
Kinon et al 21					−
Paliperidone	1.5	3	6	9	12	15
Coppola et al 80	−		−
Canuso et al 81			−g		+g
Davidson et al 74		+		+		+
Kane et al 72			+	+	+
Marder et al 73			+		+
Hirayasu et al 75			+
Canuso et al 82					+h
Quetiapine	75	150	<250	250	300	400	450	600	750	800
Immediate release (IR)
Arvanitis et al 55	−	+			+			+	+
Small et al 83			−i
Fabre et al 84				+
Lindenmayer et al 23					−			−
Kahn et al 85						+
King et al 86							+j
11915A 87								+
11916A 88								+
Loebel et al 68								+
Canuso et al 82										−k
Potkin et al 89								−l		−l
Cutler et al 90										−
Extended release (XR)
Kahn et al 85						+		+		+
Lindenmayer et al 23					−			+		−
Cutler et al 90						−		−		−
Risperidone	2	4	6	8	10	12	16
Study 0204 60,61,91	+m		+m		+m		+m
RIS-USA-72 92		+		+
Peuskens et al 93		+n		+n		−n	+n
Bose et al 94		+
Potkin et al 89			+o
Potkin et al 39			+
Barbato et al 51			+
Casey et al 95			+
Study 041-002 41			+
Potkin et al 44			−
Potkin et al 56				+p
Study D1001002 69		q
Sertindole	8	12	16	20	24
Zimbroff et al 47		+		+	+
Van Kammen et al 96	−	−		+
Zborowski et al 59				+	+
Hale et al 97			+r	−r	+r
Ziprasidone	10	40	80	120	160	200
Daniel et al 98			+		+
Goff et al 99 (F)	−	−			−s
Keck et al 100		−		+
Study 104 101	−	−	−
Study 115 57		+		+		+
Cutler et al 64					+
Zotepine	75	150	300
Cooper et al 102			+t
Cooper et al 103			+t
Knoll 104	u	u	u

Note: (+) = statistically significantly better than placebo/very low dose in the primary outcome. (−) = not statistically significantly better than placebo/very low dose in the primary outcome. The light grey background illustrates the minimum effective dose based upon our main (primary) criterion (at least 1 trial with a statistically significant difference compared to placebo/low dose). The dark grey background illustrates the minimum effective dose based upon our more conservative secondary criterion (at least 2 positive studies). If a dose met both criteria, only the dark grey background was used. F = failed study, ie, a study in which neither the new drug nor the active comparator was more effective than placebo. A study is only marked as failed in the row for the new drug, not in the row for the active comparator.

^aOnly significantly better than 100 mg/d in an uncontrolled test.

^bSignificantly better in the mixed model for repeated measurements analysis, not in the last-observation-carried-forward analysis.

^cAs only one study was available, a minimum effective dose could not be defined based on our secondary criterion.

^dThe results of this study comparing lurasidone 40 and 80 mg/d with placebo were not obtained.

^eLow-dose olanzapine (1 mg/d) as a comparator, not placebo.

^fThe highly significant statistical tests of these 4 paliperidone studies, which used olanzapine 10 mg/d only for “assay sensitivity,” were made post hoc tests by us.

^g12 mg/d could be reduced to 9 and 6 mg/d and to 3 mg/d.

^hThe initial target dose 9 mg/d could be increased to 12 mg/d. The 2-week monotherapy phase of the study was used for the evaluation.

ⁱQuetiapine was more efficacious than placebo at doses higher than 250 mg/d (average 360 mg/d).

^jLow-dose quetiapine 50 mg/d as a control, not placebo.

^kThe initial target dose 600 mg/d could be increased to 800 mg/d. The 2-week monotherapy phase of the study was used.

^lParticipants with ≤70 kg bodyweight received 400 mg/d, which could be increased to 600 mg/d, and those with >70 kg received 600 mg/d, which could be increased to 800 mg/d. The 2-week monotherapy phase results are presented.

^mThe US American and Canadian parts of the same study have been published separately.

ⁿLow-dose 1 mg/d risperidone as a comparator, not placebo.

^oParticipants with ≤70 kg bodyweight received 4 mg/d and those with >70 kg received 6 mg/d. The 2-week monotherapy phase results are presented.

^pThe fixed-dose range was 6–8 mg/d.

^qThe risperidone dose of this unpublished study was 4 mg/d.

^rLow-dose sertindole 8 mg/d as the comparator, not placebo.

^sThe comparator was ziprasidone 4 mg/d, not placebo. Ziprasidone 160 mg/d and haloperidol were only efficacious in the Clinical Global Impression scale, not in the Brief Psychiatric Rating Scale (BPRS).

^tThe dose could be reduced to 150 mg/d.

^u75, 150, and 300 mg/d were compared with placebo, the results were not obtained.

Table 2.

Minimum Effective Doses of Second-Generation Antipsychotic Drugs and Dose Equivalents

Drug	Minimum Effective Dose	OLA 1 mg Equivalent	RIS 1 mg Equivalent	HAL 1 mg Equivalent	CPZ 100 mg Equivalent
Amisulpride	−	−	−	−	−
Aripiprazole	10	1.33 (1)	5 (2.5)	2.5 (2.2)	4 (3.6)
Asenapine	10	1.33 (1)	5 (2.5)	2.5 (2.2)	4 (3.6)
Clozapine	300?	40 (30)	150 (75)	75 (67)	120 (107)
Haloperidol	4 (4.5)	0.53 (0.45)	2 (1.13)	1	1.6
Iloperidone	8a (12)	1.07a (1.2)	4a (3)	2a (2.7)	3.2a (4.3)
Lurasidone	40	5,33 (4)	20 (10)	10 (8.9)	16 (14.2)
Olanzapine	7.5 (10)	1	3.75 (2.5)	1.88 (2.2)	3 (3.6)
Paliperidone	3 (6)	0.4 (0.6)	1.5 (1.5)	0.75 (1.3)	1.2 (2.1)
Quetiapine	150 (250)	20 (25)	75 (62.5)	37.5 (55.6)	60 (88.9)
Risperidone	2 (4)	0.27 (0.4)	1	0.5 (0.9)	0.8 (1.4)
Sertindole	12 (16)	1.60 (1.6)	6 (4)	3 (3.6)	4.8 (5.7)
Ziprasidone	40 (80)	5.33 (8)	20 (20)	10 (17.8)	16 (28.4)
Zotepine	−	−	−	−	−

Note: We present the minimum effective doses and the doses in milligrams that are equivalent to 1 mg/d olanzapine (OLA), 1 mg/d risperidone (RIS), 1 mg/d haloperidol (HAL), and 100 mg/d chlorpromazine (CPZ). Numbers in parentheses are the results of the sensitivity analysis (2 positive trials). The results of the sensitivity analysis are only shown if they deviated from the primary analysis. (−): means that no recommendation could be made. (?): the result is very questionable because it is based on a single small study comparing clozapine 300 mg/d with 100 mg/d, but not with placebo.

^aExcluding patients with schizoaffective disorder, the minimum effective iloperidone dose based on the primary criterion is 12 mg/d; the resulting equivalence doses (mg/d) are: olanzapine 1.6, risperidone 6, haloperidol 3, and chlorpromazine 4.8.

Amisulpride

The single dose-finding study in acute schizophrenia compared amisulpride 400, 800, and 1200mg/d with a dose of 100 mg/d,³⁴ which the original authors had presumed to be subtherapeutic. It showed a bell-shaped dose-response curve, where 400 and 800 mg/d were associated with the largest reduction in symptoms, however 400 mg/d was significantly more efficacious than 100 mg/d only in an unadjusted test. Amisulpride 100 mg/d was not actually an ineffective dose as it produced a mean 18.4 BPRS total score reduction from baseline (approximately equivalent to 28 points PANSS total score reduction¹⁰⁵). It is therefore impossible to derive a minimum effective dose based on our method (see tables 1 and 2).

Aripiprazole

Seven fixed-dose, placebo-controlled studies^20,35–40 examined aripiprazole doses between 2 and 30 mg/d. The dose of 2 mg/d was ineffective in 2 studies.^20,35 In a single trial, 5 mg/d was more efficacious than placebo only at a few visits, but not at endpoint (primary outcome).³⁵ Meta-analyses of the single 5 mg/d dose³⁵ and the two 2 mg/d doses^20,35 (n = 379, ES: 0.18, 95% CI: −0.03 to 0.39, P = .10) or of the 2 mg/d dose groups^20,35 (n = 285, ES: 0.13, 95% CI: −0.10 to 0.36, P = .28) were also not significant. A dose of 10 mg/d was more efficacious than placebo in 3 trials^35,36,38 and is the minimum effective dose based on the primary criterion and the sensitivity analysis. This is different from the earlier recommendation by Woods¹² (15 mg/d) for whom 2 recent studies^35,38 were not yet available (see tables 1 and 2). Higher doses consistently separated from placebo, as well, except for 2 unpublished studies.^20,40 Study 94202 failed as no aripiprazole dose and also not the active comparator haloperidol (10 mg/d) was found to be effective.²⁰ Study 93202 was negative, because although the active comparator haloperidol (20 mg/d) showed efficacy for the primary outcome, aripiprazole (30 mg/d) separated from placebo only in secondary outcomes.⁴⁰

Asenapine

Five placebo-controlled, fixed-dose trials^41–45 examined asenapine doses between 0.4 and 20 mg/d. In 2 initial studies, doses up to 4.8 mg/d were not found to be efficacious,^41,42 although there was a trend when the 4.8 and 3.2 mg/d doses⁴² were meta-analytically combined (n = 182, ES: 0.29, CI: −0.02 to 0.59, P = .06). An ongoing study is therefore re-examining 5 mg/d.¹⁰⁷ A dose of 10 mg/d is currently the minimum effective dose based on both criteria, because it separated from placebo in 2^43,44 out of 3 studies.^43–45 There is one additional negative study, in which asenapine with a dose of neither 10 nor 20 mg/d was efficacious, while the active comparator olanzapine 15 mg/d was effective,⁴⁵ and another ongoing study is re-examining 10 and 20 mg/d doses.¹⁰⁸

Clozapine

Placebo-controlled, dose-finding studies on clozapine were not available. As clozapine is restricted only for use in patients with treatment resistance, we included a small study in this patient group, which found that clozapine 600 and 300 mg/d were better than 100 mg/d.⁴⁶ Arguably, 300 mg/d is the minimum effective dose based on our primary criterion. The sensitivity analysis could not be conducted, because only one study was available.

Haloperidol

Haloperidol has been frequently used as an additional active comparator in placebo-controlled trials of SGAs, usually at fixed doses that are nowadays considered to be relatively high such as 8 mg/d,⁴³ 10 mg/d,^{20,37,51–54} 12 mg/d,⁵⁵ 15 mg/d,^56,57,99 15 ± 5 mg/d,⁵⁸ 16 mg/d,⁵⁹ and 20 mg/d.^40,60–62 These doses were generally statistically significantly more efficacious than placebo apart from a few studies, which failed as no treatment was better than placebo^{20,51,53,62,99}; 20 mg/d were also not efficacious in Klieser and Lehmann⁶² and in Chouinard et al.⁶¹ However, the evaluation of the low end of the dose range for haloperidol is suboptimal. In a seven-arm sertindole study, 4, 8, and 16 mg/d haloperidol were all more efficacious than placebo.⁴⁷ This finding is corroborated by a study from 1971 in which haloperidol doses up to 4.5 mg/d were superior to placebo,⁴⁸ while 2 other small (52 and 24 participants, respectively), old studies found only a trend for superiority of haloperidol up to 6 mg/d compared with placebo.^49,50

Thus, 4 mg/d is the minimum effective haloperidol dose according to our primary criterion, while 4.5 mg/d fulfills that of the sensitivity analysis.

Iloperidone

There were 5 placebo-controlled fixed-dose (or fixed-dose range) studies, which examined iloperidone doses between 4 and 24 mg/d. Two 4 mg/d doses^56,63 and their meta- analysis were ineffective (n = 293, ES: 0.12, CI: −0.11 to 0.35, P = .30). A dose of 8 mg/d was also ineffective in 2 trials,^56,63 and yet the fixed-dose range of 4–8 mg/d was efficacious in another study.⁵⁶ The upper limit of this dose range, 8 mg/d, therefore, qualifies for the minimum effective dose according to our primary criterion. A dose of 12 mg/d separated from placebo in a further study⁵⁶ and is thus the minimum effective dose based on the sensitivity analysis. Higher doses were efficacious, as well, except for the fixed-dose range of 12–16 mg/d in Study 3005 (Potkin Study 3).⁵⁶

We note that if patients with schizoaffective disorder are excluded (as nowadays required by the FDA⁶³), the minimum effective dose would be 12 mg/d for our primary criterion: In Study 3004,^56,63 neither 4–8 nor 12–16 mg/day was efficacious, in Study 3000 (Potkin Study 1) 12 mg/day remained effiacious, while in the Study 3 of Potkin et al.,^56,63 both 12–16 and 20–24 mg/d were effective.

Lurasidone

There were 7 fixed-dose, placebo-controlled studies, which examined lurasidone doses between 20 and 160 mg/d.^{22,53,65–69} The dose of 20 mg/d was not efficacious in a single study, which failed as neither any lurasidone dose (20, 40, and 80 mg/d) nor haloperidol 10 mg/d was significantly superior to placebo.⁵³ The dose of 40 mg/d was efficacious in 2^65,66 out of 4^22,53,65,66 trials (one of which was a failed study⁵³) and is therefore the minimum effective dose. Higher doses were usually efficacious, except for 80 mg/d in the failed study⁵³ and 120 mg/d in a study reported by Nasrallah et al.²² Data from study NCT00711269/D1001002⁶⁹ (40 and 80 mg/d) are not available. A study that re-examines 20 mg/d is expected to complete in September 2014.¹⁰⁹

Olanzapine

We identified 15 relevant studies, which examined olanzapine doses between 1 and 15 ± 2.5 mg/d.^{21,45,51,58,66,70–79} One olanzapine 1 mg/d⁷⁰ dose group and 2 fixed-dose range groups of 5 ± 2.5mg/d did not separate neither from placebo⁵⁸ nor from the presumed ineffective dose of 1 mg/d olanzapine.⁷¹ However, their meta-analysis was significant (n = 393, ES: 0.20, CI: 0.005–0.40, P = .04). Therefore, 7.5 mg/d (the upper range of 5 ± 2.5mg/d) is the minimum effective dose. Higher olanzapine doses (10, 10 ± 2.5, 15, and 15 ± 2.5 mg/d) were more efficacious than placebo in 13^{45,51,58,66,70,74,75,76–79} out of 15 studies. The exceptions were the recent study on a glutamatergic compound²¹ and the failed study by Beasley et al 1997⁷¹ in which only the highest olanzapine dose (15 ± 2.5mg/d) was more efficacious than 1 mg/d, but only in secondary outcomes. The authors commented that olanzapine 1 mg/d was not completely ineffective.⁷¹ A dose of 10 mg/d olanzapine is the minimum effective dose based on the sensitivity analysis.

Paliperidone

Seven studies were included, which compared paliperidone doses between 1.5 and 15 mg/d^{72–75,80–82} with placebo. Paliperidone 1.5 mg/d was not effective in the single study that examined this dose,⁸⁰ whereas 3 mg/d in a single study⁷⁴ was efficacious and thus is the minimum effective dose based upon the primary criterion.⁷⁴ The 6 mg/d dose did separate from placebo in 3^72,73,75 out of 5 studies^{72,73,75,80,81} and thus meets the criterion of the sensitivity analysis. Doses higher than 6 mg/d were consistently more efficacious than placebo.^{72–74,81,82}

Quetiapine

We included 12 studies on quetiapine immediate release^{23,55,68,82–90} and 3 on extended release,^23,85,90 which examined doses between 75 and 800 mg/d. While 75 mg/d was ineffective,⁵⁵ 150 mg/d separated from placebo in the single study examining it⁵⁵ and is the minimum effective dose based upon the primary criterion. The dose of 250 mg/d qualifies for the criterion of the sensitivity analysis (Fabre et al⁸⁴). Overall determining dose response for quetiapine is not straightforward as higher doses did not separate from placebo in several studies, including quetiapine 800 mg/d arms^23,90 (see table 1).

Risperidone

Twelve relevant studies compared risperidone doses between 2 and 16 mg/d with placebo^{39,41,44,51,56,60,61,69,89,92,94,95} or presumed ineffective risperidone 1 mg/d.⁹³ The dose of 2 mg/d was not more efficacious than placebo in the separately published US part⁶⁰ of the pivotal study 0204, but it was in the Canadian part⁶¹ and overall (see FDA report⁹¹). Consequently, 2 mg/d qualifies as the primary minimum effective dose. Risperidone 4 mg/d met the criterion of the sensitivity analysis. It was more efficacious than risperidone 1 mg/d (a presumed ineffective dose) in a large international study⁹³ and it was more efficacious than placebo in a recent trial on cariprazine.⁹⁴ In RIS-USA-72,⁹² it separated from placebo in the study’s primary outcome (the number of participants with at least 20% PANSS total score reduction), the superiority in the mean PANSS total score change was only of borderline statistical significance (P = .051). The results of a 4 mg/d risperidone group of a lurasidone study have not been published (Study NCT00711269/D1001002⁶⁹). Risperidone doses of 6 mg/d or higher were consistently more efficacious than placebo in all other studies,^{39,41,51,56,89,95} except for 6 mg/d in one asenapine trial,⁴⁴ and 12 mg/d did not separate from risperidone 1 mg/d in the large international study.⁹³

Sertindole

Three placebo-controlled studies^47,59,96 and 1 with low-dose sertindole (8 mg/d) as a comparator⁹⁷ examined doses between 8 and 24 mg/d. The dose of 8 mg/d was not efficacious in the single study comparing it with placebo⁹⁶ and was therefore used as a “presumed ineffective” dose instead of placebo in another study in which it was confirmed as being less efficacious than 16 and 24 mg/d.⁹⁷ The dose of 12 mg/d was more efficacious than placebo in 1⁴⁷ out of 2 studies^47,96 and is therefore the minimum effective dose fulfilling the primary criterion. The dose of 16 mg/d meets the criterion of the sensitivity analysis.⁹⁷ Higher doses of 20 mg/d^47,59,96 and 24 mg/d^47,59,97 were efficacious in 3 trials each.

Ziprasidone

We included 6 studies, which compared ziprasidone doses between 10 and 200 mg/d,^{57,64,98–101} with placebo or ziprasidone 4 mg/d.⁹⁹ 10 mg/d was not efficacious in 2 trials,^99,101 and their meta-analysis was also not significant (n = 127, ES: 0.01, CI: −0.34 to 0.35, P = .96). A dose of 40 mg/d separated from placebo in 1 study⁵⁷ and is therefore the minimum effective dose according to the primary criterion, which is corroborated by a maintenance study, although it was excluded because it was conducted in patients with a stable presentation.¹¹⁰ A dose of 80 mg/d separated from placebo in 1⁹⁸ out of 2 studies^98,101 and therefore—taken together with the positive study on 40 mg/d⁵⁷—meets the criterion of the sensitivity analysis. Higher doses were consistently efficacious in studies that examined them,^57,64,98,100 except 1 failed study⁹⁹ in which only 160 mg/d and haloperidol 15 mg/d separated from the presumed ineffective dose of 4 mg/d ziprasidone, but only for secondary outcomes.

Zotepine

Data on the only dose-finding study that compared zotepine 75, 150, and 300 mg/d with placebo could not be obtained.¹⁰⁴ The only other 2 placebo-controlled trials^102,103 employed fixed doses of 300 mg/d and yet these could be reduced to 150 mg/d. The majority of patients stayed on 300 mg/d in 1 trial, but the proportion doing so in the other trial has not been reported.¹⁰³ In theory, 300 mg/d would be the minimum effective dose, but this finding is very doubtful.

Discussion

We refined, updated, and expanded the frequently used dose equivalence method first presented by Woods in 2003¹² by including 8 other SGAs and by expanding the database from 15¹² to now 73 RCTs. Some minimum effective doses had to be changed, whereas others were corroborated by new supporting evidence.

Authors agree that ideally large RCTs, which compare multiple fixed doses of multiple drugs, would be needed but, given the high number of antipsychotics, it is unlikely that these will ever be available.^3,10,12 The key strength of the method presented by Woods¹² is that an operationalized, evidence-based criterion can be applied to each antipsychotic. A critical limitation is the question how well the minimum effective doses of each drug could be identified by the dose-finding trials. Most mean ESs of the minimum effective doses were in a small range and thus, by and large, comparable: asenapine 10 mg/d, n = 561, ES: 0.31, CI: 0.14–0.47; iloperidone 8 mg/d, n = 585, ES: 0.24, CI: 0.08–0.40; lurasidone 40 mg/d, n = 705, ES: 0.13, CI: −0.02 to 0.28 (excluding 1 failed study,⁵³ the result was significant, n = 569, ES: 0.23, CI: 0.06–0.39); olanzapine 7.5 mg/d, ES: 0.20, CI: 0.005–0.40; risperidone 2 mg/d, ES: 0.32, CI: 0.02–0.62; sertindole 12 mg/d, ES: 0.36, CI: 0.09–0.62; ziprasidone 40 mg/d, ES: 0.21, CI: 0.01–0.40 (an ES for clozapine could not be calculated). The mean ESs of haloperidol 4 mg/d (n = 139, ES: 0.49, CI: 0.16–0.83), paliperidone 3 mg/d (n = 243, ES: 0.60, CI: 0.34–0.86, but 1.5 mg/d was not effective and even tended to be less efficacious than placebo⁸⁰), and possibly also aripiprazole 10 mg/d (n = 497, ES: 0.40, CI: 0.22–0.58) were medium sized, so that future studies might reveal that lower doses are effective, as well. Finally, the minimum effective dose of quetiapine (150 mg/d) appears low, but its ES was high (n = 99, ES: 0.70, CI: 0.29–1.10), and it was numerically higher than all other quetiapine doses including 750 mg/d in the pivotal quetiapine study.⁵⁵ Several negative results for high doses of quetiapine (see table 1) suggest that the dose-response relationship for this antipsychotic is difficult to understand. But the quetiapine example also shows that including ES in the judgment is problematic, because placebo response^17,111 and dropout rates^19,112 have increased over the decades, and they have been made responsible for small ESs in recent studies in which even well-established drugs such as haloperidol or olanzapine do not separate from placebo.^20,21 Consequently, it is difficult to compare ESs from different periods. Moreover, these in part relatively high ESs also demonstrate that the application of the “FDA two studies rule,” as used in the sensitivity analysis, is particularly subject to the small number of trials available, because the doses identified were often higher.

Other limitations are that we excluded studies in specific populations, which may require different doses. RCTs suggest that 2–3 mg/d haloperidol is effective in first-episode patients. Five placebo-controlled, fixed-dose amisulpride studies^113–117 suggest that 50 mg/d is the minimum effective dose for patients with predominant negative symptoms.^113,116 Moreover, most included trials are registrational studies with limited follow-up and generalizability, and the approach does not take side-effects into account, which in practice need to be balanced with efficacy. Statistical significance depends in part on the sample size. Theoretically, very high sample sizes could enable even very low doses to show some efficacy over placebo. We included a few “fixed-dose range” studies. However, if we had excluded them all, only the minimum effective dose for iloperidone would need to be increased to 12 mg/d.

The relative value of the minimum effective dose method should also be considered in light of the limitations of other methods to define dose equivalence (see also Patel et al³ for a detailed review). All methods (including Woods¹²) assume linearity, but the dose relationships between drugs could also be different (eg, quadratic), and the dose-response curves are usually sigmoidal.⁵ Even when consensus methods were used, experts simply assumed linearity.¹¹ Further, as consensus methods^9–11 also depend on the knowledge of experts, we wonder whether the extent of their knowledge is likely to fully include all relevant RCTs as this review included 73 RCTs, some of which were unpublished.

The original method proposed by Davis⁴ estimated chlorpromazine equivalents by calculating the ratio of the mean doses of each antipsychotic and the mean chlorpromazine dose used in flexible-dose trials. This method is not limited by the robustness of the evidence base from which the minimum effective dose is identified, and a larger evidence base may be available. Its major limitation is that drugs can usually only be given in predefined doses ranges which, in extreme cases, may not even include the optimum dose. Moreover, Davis assumed that all first-generation antipsychotics are equally efficacious, but small efficacy differences between SGAs may exist.^2,26

Davis and Chen⁵ constructed dose-response curves from dose-finding studies for each antipsychotic to identify near-to-maximum effective doses (ED95) and median effective doses (ED50). This approach is extensively used in preclinical research and, while this might seem appropriate, it is again hampered by the paucity of available data and inconsistencies of different dose-finding studies of the same drug.

Cochrane meta-analyses on the doses of a few antipsychotics exist.^31,32,118 This is the most comprehensive method, but their aim was not to identify dose equivalence. Rather various efficacy and tolerability outcomes were examined to identify the optimum doses of each drug.

DDDs are the “assumed, average, maintenance doses” of drugs for their main indication in adults.^3,8 They have been developed by the World Health Organization as a technical metric to measure drug utilization rather than to define dose equivalence. Finally, “maximum licensed methods”^6,7 express doses as a percentage of the maximum licensed dose of a drug. For example, the maximum chlorpromazine dose according to the British National Formulary¹¹⁹ is 1000 mg/d, corresponding to 100%. Obviously, whether this method can be used for dose equivalence depends on the premise that the maximum licensed doses are really the maximally efficacious doses. For example, the summary of product characteristics for risperidone indicates that the effective dose range is from 4 to 16 mg/d, although studies have shown that doses higher than 6–8 mg/d are associated only with more side-effects but not with more efficacy.^60,61,93

We conclude that, as with all other methods to define dose equivalence, our update and refinement of the minimum effective dose method has strengths and limitations, but that it will be useful in many situations. We recommend that alternative methods also be updated in order to minimize further differences between the methods and risk of subsequent bias.

Supplementary Material

Supplementary material is available at http:// schizophreniabulletin.oxfordjournals.org.

Funding

Supported in part by a grant from the German Federal Ministry of Education and Research, no FKZ: 01KG 0606, GZ: GFKG01100506.