Pfizer Australia, 3842 Wharf Road, West Ryde, Sydney, NSW 2114, Australia
Pfizer Limited UK, Tadworth, Surrey, KT20 7NS, UK
Department of Statistics, Macquarie University, Sydney, NSW 2109, Australia
Abstract
Background
To compare the efficacy of pregabalin and gabapentin at comparable effective dose levels in patients with refractory partial epilepsy.
Methods
Eight randomized placebo controlled trials investigating the efficacy of pregabalin (4 studies) and gabapentin (4 studies) over 12 weeks were identified with a systematic literature search. The endpoints of interest were "responder rate" (where response was defined as at least a 50% reduction from baseline in the number of seizures) and "change from baseline in seizurefree days over the last 28 days (SFD)". Results of all trials were analyzed using an indirect comparison approach with placebo as the common comparator. The basecase analysis used the intentiontotreat last observation carried forward method. Two sensitivity analyses were conducted among completer and responder populations.
Results
The basecase analysis revealed statistically significant differences in response rate in favor of pregabalin 300 mg versus gabapentin 1200 mg (odds ratio, 1.82; 95% confidence interval, 1.02, 3.25) and pregabalin 600 mg versus gabapentin 1800 mg (odds ratio, 2.52; 95% confidence interval, 1.21, 5.27). Both sensitivity analyses supported the findings of the basecase analysis, although statistical significance was not demonstrated. All dose levels of pregabalin (150 mg to 600 mg) were more efficacious than corresponding dosages of gabapentin (900 mg to 2400 mg) in terms of SFD over the last 28 days.
Conclusion
In patients with refractory partial epilepsy, pregabalin is likely to be more effective than gabapentin at comparable effective doses, based on clinical response and the number of SFD.
Background
The primary objective of antiepileptic therapy is to obtain complete control of seizures while minimizing the occurrence of adverse events and improving the patient's quality of life
Pregabalin and gabapentin, α2δ ligands, are both licensed as adjunctive treatment for partial epilepsy, however headtohead comparisons of their efficacy and safety have not been conducted. In the absence of direct, prospective, comparative studies to guide medical decisionmaking in the adjunctive treatment of partial epilepsy, alternative approaches are required to assess the relative value of a particular intervention versus other relevant comparators. In recent years, the role of metaanalysis has developed substantially in medical applications
Methods
Identification and study selection
In order to identify relevant publications, a systematic literature review was performed in English on the PubMed and Thomson ISI Web of Science bibliographic databases. For example, the pregabalin search strategy on Pubmed was defined as follows: (("pregabalin"[Substance Name] OR "pregabalin"[All Fields]) AND partial[All Fields] AND ("epilepsy"[MeSH Terms] OR "epilepsy"[All Fields])) AND (Randomized Controlled Trial[ptyp] AND English[lang]). Studies were then included according to the following predetermined conditions:
Analysis
Figure
Flow chart for analysis of responders
Flow chart for analysis of responders. CI = confidence interval; GBP = gabapentin; PGB = pregabalin
To compare highdose pregabalin to highdose gabapentin, the estimated odds ratios for each dose were compared via indirect comparisons, using placebo as the common comparator. This type of analysis also estimates the efficacy of gabapentin at the higher 2400 mg/day by extrapolating the dose response equations, and addresses the lack of clinical trial data at pregabalin 450 mg/day by interpolating the doseresponse equations.
Imputed data for the basecase analysis was derived from the intentiontotreat (ITT) population and used the last observation carried forward (LOCF) method. However, concerns exist regarding whether it is appropriate to use LOCF in analyses involving progressive conditions or in situations where it may not be possible to determine whether missing data are nonrandom. Specifically for epilepsy, LOCF analysis yields seizurefree rates that are higher than the true clinical situation
Responder rates and logistic regression
Logistic regression analyses were used to model doseresponse curves. For each drug, the odds ratio (vs. placebo) was modelled as a function of the drug dose. Based on exploratory analyses of the crude odds ratios (Figures
Here, odds is taken to mean
This estimation approach enabled active doses to have a different slope and gradient than placebo, while producing both estimates within the same equation without excluding placebo. Odds ratios along with corresponding 95% confidence intervals (CIs) were reported for each dose of each drug versus placebo. That is, pregabalin doses versus placebo from pregabalin trials and gabapentin doses versus placebo from gabapentin trials.
Indirect comparisons of pregabalin and gabapentin were performed on the placeboadjusted results of these direct comparisons. By only comparing the placeboadjusted effects, this method of adjusted indirect comparison may preserve randomization and account for different baseline risks and other prognostic factors for participants in different trials
The log odds ratio for the indirect comparison between pregabalin and gabapentin was obtained by subtracting the log odds ratios of pregabalin versus placebo, from the log odds ratios of gabapentin versus placebo.
Since the log odds ratios of pregabalin versus placebo and gabapentin versus placebo were estimated from different studies, they were statistically independent. Thus, the variance of the log odds ratios between pregabalin and gabapentin was obtained by summing the variances for the log odds ratios of each active treatment versus placebo, and 95% CIs were constructed from these variances.
Odds ratios for the indirect comparisons are presented on the original scale by taking the exponential of the estimate of the log odds ratio and bounds of the 95% Cl for the log odds ratio. For ease of interpretation odds ratios were converted to relative risk (RR) using the formula of Sutton et al
where Risk_{c }indicates the probability of achieving a response in the placebo group
Change from baseline in seizurefree days over the last 28 days
This analysis used the same approach as the responder analysis except that the change from baseline in SFD is a continuous variable. That is, the same model was used as in equation (1) with the exception that the outcome variable was change from baseline in SFD. In line with good statistical practice, baseline SFD was included as a covariate. Analysis of covariance techniques were used to estimate the doseresponse curve, using PROC GLM (SAS Version 8.0). Using this model, estimates of the treatment difference between active doses and placebo along with 95% CIs were obtained.
Since mean treatment differences between active doses and placebo for pregabalin and gabapentin were analyzed using separate dose response equations, the final analyses required indirect comparison between pregabalin and gabapentin for the mean difference adjusted by the results of the direct comparisons with placebo
Since the mean difference of each active treatment versus placebo was estimated from different studies, they were statistically independent. Thus, the variance of the mean difference between pregabalin and gabapentin was obtained by adding the variances for the mean difference of each active treatment versus placebo and 95% CIs were constructed from these variances.
Doseresponse curves were estimated for the change from baseline in SFD for pregabalin and gabapentin separately.
Results
Eight original research articles were identified that fulfilled the data extraction phase (Table
Summary of the multicenter, doubleblind, randomized placebocontrolled trials included in the basecase and sensitivity analyses.
Individual Studies
Pregabalin Trials
Daily Dose (Titration Period)
No. of Patients
Gabapentin Trials
Daily Dose (Titration Period)
No. of Patients
ITT
Completers
ITT
Completers
Beydoun et al 2005
600 mg (1 week)
214
156
UK Gabapentin Study Group, 1990
1200 mg (2 weeks)
61
54
Placebo
98
81
Placebo
66
61
Arroyo et al 2004
150 mg (3 days)
99
88
The US Gabapentin Study Group No. 5, 1993
1200 mg (23 days)
101
95
600 mg (1 week)
92
69
1800 mg (23 days)
54
53
Placebo
96
84
Placebo
98
96
French et al 2003
150 mg
86
81
Anhut et al 1994
900 mg (2 days)
109
100
300 mg
90
71
600 mg
89
61
1200 mg (2 days)
52
50
Placebo
100
87
Placebo
109
100
Elger et al 2005
600 mg
137
80
Sivenius et al 1991
900 mg (2 days)
36
32
1200 mg (2 days)
17
16
Placebo
73
56
Placebo
34
30
Aggregated Studies
Treatment in Pregabalin Trials
Daily Dose
No. of Patients
Treatment in Gabapentin Trials
Daily Dose
No. of Patients
ITT
Completers
ITT
Completers
Placebo
367
308
Placebo
307
287
LowDose Pregabalin
150 mg
185
169
LowDose Gabapentin
900 mg
145
132
MidDose Pregabalin
300 mg
90
71
MidDose Gabapentin
1200 mg
231
215
HighDose Pregabalin
600 mg
532
366
HighDose Gabapentin
1800 mg
54
53
All studies had a 12week doubleblind maintenance phase but the durations of the titration phases were variable.
^{a}In addition to the doses listed above, for trial 1008034 patients were also recruited into a 50 mg dose group. This dose was not found to be therapeutic and is not registered for use in Australia.
^{b}No titration phase.
^{c}In addition to the doses listed above, for trial 1008157 patients were also recruited into a titrated dose arm with total daily doses ranging from 150 mg to 600 mg, depending on individual requirement. This treatment group was not comparable with the other fixeddose treatment groups and was excluded from the analyses.
^{d}In addition to the doses listed above, for trial 9455 patients were also recruited into a 600 mg dose group. This dose was not found to be therapeutic and is not registered in Australia.
ITT = intentiontotreat.
Logistic regression analysis of ≥50% reduction in baseline seizures: Basecase analysis (ITT LOCF)
In the basecase analysis, each dose of pregabalin was significantly different from placebo, with the magnitude of the difference increasing with dose (Table
Doseresponse curves for pregabalin and gabapentin for response (basecase analysis)
Doseresponse curves for pregabalin and gabapentin for response (basecase analysis). The 450 mg pregabalin and 2400 mg gabapentin doses were not studied in any of the trials; the results plotted here were calculated from the doseresponse equations. CI = confidence interval
Odds ratios, relative risks and corresponding 95% confidence intervals (CI) for pregabalin and gabapentin versus placebo, and for pregabalin versus gabapentin in the indirect comparison using placebo as the common comparator (basecase analysis using the ITT LOCF approach)
Dose Comparison
Odds Ratio
95% CI
Relative Risk ^{ a }
95% CI
Pregabalin
150 mg vs. Placebo
2.75
1.73, 4.38
2.34
1.61, 3.27
300 mg vs. Placebo
4.62
3.16, 6.77
3.39
2.60, 4.29
450 mg vs. Placebo^{b}
6.26
4.33, 9.06
4.10
3.25, 5.02
600 mg vs. Placebo
7.77
5.32, 11.34
4.63
3.72, 5.58
Gabapentin
900 mg vs. Placebo
2.21
1.33, 3.68
1.97
1.29, 2.90
1200 mg vs. Placebo
2.54
1.64, 3.93
2.20
1.54, 3.04
1800 mg vs. Placebo
3.08
1.64, 5.81
2.55
1.54, 3.92
2400 mg vs. Placebo
3.54
1.48, 8.49
2.82
1.41, 4.85
Pregabalin vs. Gabapentin
150 mg vs. 900 mg
1.24
0.62, 2.48
1.21
0.64, 2.16
300 mg vs. 1200 mg
1.82
1.02, 3.25
1.68
1.02, 2.65
450 mg^{a }vs. 1800 mg
2.03
0.98, 4.23
1.84
0.98, 3.21
600 mg vs. 1800 mg
2.52
1.21, 5.27
2.19
1.19, 3.69
600 mg vs. 2400 mg^{a}
2.19
0.85, 5.69
1.96
0.86, 3.87
^{a}These relativerisk values would change with any change in assumption regarding the placebo response rate, currently assumed to be 10%.
^{b}450 mg pregabalin and 2400 mg gabapentin were not studied in any of the trials; these results were calculated from the doseresponse equation.
ITT = intentiontotreat; LOCF = last observation carried forward
Similar to pregabalin, each dose of gabapentin was significantly different from placebo, with the magnitude of the difference increasing with dose (Table
Table
Analysis of completers
These datasets were restricted to those patients who completed the clinical trials and from this completer patient population, responders were identified. The percentage of patients discontinuing the trials was greater at the higher doses of pregabalin than gabapentin trials. In addition, a greater percentage of placebo patient withdrew from the pregabalin than gabapentin trials.
The odds ratios and 95% CIs derived from the pregabalin and gabapentin doseresponse curves among completers are summarized in Table
Odds ratios, relative risks and corresponding 95% confidence intervals (CI) for pregabalin and gabapentin versus placebo, and for pregabalin versus gabapentin in the indirect comparison using placebo as the common comparator (sensitivity analyses)
Dose Comparison
Odds Ratio
95% CI
Relative Risk ^{ a }
95% CI
Analysis of Completers
Pregabalin
150 mg vs. Placebo
2.63
1.61, 4.30
2.26
1.52, 3.23
300 mg vs. Placebo
4.23
2.82, 6.36
3.20
2.39, 4.14
600 mg vs. Placebo
6.81
4.50, 10.30
4.31
3.33, 5.34
Gabapentin
900 mg vs. Placebo
2.45
1.44, 4.17
2.14
1.38, 3.17
1200 mg vs. Placebo
2.72
1.72, 4.30
2.32
1.60, 3.23
1800 mg vs. Placebo
3.15
1.64, 6.04
2.59
1.54, 4.02
Pregabalin vs. Gabapentin
150 mg vs. 900 mg
1.07
0.52, 2.21
1.06
0.55, 1.97
300 mg vs. 1200 mg
1.56
0.84, 2.87
1.48
0.85, 2.42
600 mg vs. 1800 mg
2.16
1.00, 4.68^{b}
1.94
1.00, 3.42
Analysis of Responders
Pregabalin
150 mg vs. Placebo
2.85
1.76, 4.63
2.41
1.64, 3.40
300 mg vs. Placebo
3.60
2.41, 5.37
2.86
2.11, 3.74
600 mg vs. Placebo
4.54
3.04, 6.77
3.35
2.52, 4.29
Gabapentin
900 mg vs. Placebo
2.34
1.38, 3.96
2.06
1.33, 3.06
1200 mg vs. Placebo
2.69
1.71, 4.24
2.30
1.60, 3.20
1800 mg vs. Placebo
3.28
1.72, 6.28
2.67
1.60, 4.11
Pregabalin vs. Gabapentin
150 mg vs. 900 mg
1.22
0.60, 2.49
1.19
0.63, 2.17
300 mg vs. 1200 mg
1.34
0.73, 2.45
1.30
0.75, 2.14
600 mg vs. 1800 mg
1.38
0.65, 2.96
1.33
0.67, 2.47
^{a}These relativerisk values would change with any change in assumption regarding the placebo response rate, currently assumed to be 10%.
^{b}Not significant (rounded up to 1.00).
Doseresponse curves for pregabalin and gabapentin for response (sensitivity analyses)
Doseresponse curves for pregabalin and gabapentin for response (sensitivity analyses). CI = confidence interval
When these results are subject to indirect comparison using placebo as the common comparator, some differences were evident between the basecase and completer analysis. The magnitude of effects in favor of pregabalin over gabapentin at all doses in the basecase analysis are only retained for the highdose comparison (i.e., pregabalin 600 mg vs. gabapentin 1800 mg) in the completer analysis (Table
Analysis of responders
In the analysis of responders, and consistent with the basecase analysis, each dose of pregabalin and gabapentin was significantly different from placebo (Table
When the responder data are subject to indirect comparison using placebo as the common comparator, there were no statistically significant differences between pregabalin and gabapentin at any dose level (Table
Analysis of change from baseline in seizurefree days over the last 28 days (SFD)
Pregabalin and gabapentin (at all dose levels) were associated with change from baseline increases in SFD relative to placebo (Table
Adjusted means and mean difference between pregabalin and placebo, and gabapentin and placebo, for seizurefree days over the last 28 days
Dose
Mean Change in SFD
S.E.
Mean Difference Between Active and Placebo
95% CI for Mean Difference
Pregabalin
150 mg
5.40
0.35
2.29
1.87, 2.72
300 mg
5.71
0.36
2.61
2.12, 3.10
450 mg^{a}
5.90
0.36
2.80
2.28, 3.32
600 mg
6.03
0.36
2.93
2.38, 3.47
Placebo
3.10
0.39


Gabapentin
900 mg
4.42
0.36
1.32
0.82, 1.82
1200 mg
4.48
0.37
1.37
0.85, 1.89
1800 mg
4.56
0.37
1.45
0.90, 2.00
2400 mg^{a}
4.61
0.38
1.51
0.94, 2.08
Placebo
3.10
0.38


^{a}Note that 450 mg pregabalin and 2400 mg gabapentin were not studied in any of the trials; these results were calculated from the doseresponse equation.
CI = confidence interval; S.E. = standard error; SFD = seizurefree days over the last 28 days
Doseresponse curves for pregabalin and gabapentin for change from baseline in seizurefree days in last 28 days
Doseresponse curves for pregabalin and gabapentin for change from baseline in seizurefree days in last 28 days. The 450 mg pregabalin and 2400 mg gabapentin doses were not studied in any of the trials; these results were calculated from the doseresponse equations.
The indirect comparisons shown in Table
Mean difference in change from baseline in seizurefree days over the last 28 days for pregabalin versus gabapentin from the indirect comparison using placebo as the common comparator
Dose Comparison
Mean Difference
95% CI
150 mg vs. 900 mg
0.97
0.25, 1.69
300 mg vs. 1200 mg
1.24
0.46, 2.02
450 mg^{a }vs. 1800 mg
1.35
0.52, 2.18
600 mg vs. 1800 mg
1.48
0.62, 2.34
600 mg vs. 2400 mg^{a}
1.42
0.55, 2.29
^{a}450 mg pregabalin and 2400 mg gabapentin were not studied, thus, these results were calculated from the doseresponse equation.
Discussion
The objective of this analysis was to utilize existing data to compare the efficacy of pregabalin and gabapentin in order to better inform healthcare decisionmakers of the relative benefits of the two treatments in the current absence of direct headtohead comparisons. A randomized, doubleblind, flexibledose trial (ClinicalTrials.gov Identifier: NCT00537940) of pregabalin versus gabapentin as adjunctive therapy in patients with partial seizures is ongoing although the results will not be available until at least 2012.
Since all trials included in this analysis were placebocontrolled, placebo was used as the common comparator. Our basecase findings, based on ITT LOCF analysis, were that 12 weeks of mid and highdosage pregabalin was more efficacious than corresponding dosages of gabapentin in terms of responder rates, and that all dose levels of pregabalin were more efficacious than corresponding dosages of gabapentin in terms of SFD. A sensitivity analysis of responder rates in the completer population confirmed the benefit of mid and highdose pregabalin, respectively, over mid and highdose gabapentin, although the small sample sizes meant that the comparisons were not able to detect statistically significant betweentreatment differences. Nevertheless, the data generated in the basecase analysis for the comparison between highdose pregabalin and gabapentin was reproduced in the analysis of completers, inferring that pregabalin 600 mg is twice as likely to induce a response as gabapentin 1800 mg. In the analysis of responders, the trend in favor of pregabalin remained but statistical separation between the two drugs was not detected. One possible explanation for the observed efficacy advantage in favor of mid and highdose, but not lowdose, pregabalin relative to corresponding doses of gabapentin, is inherent differences in the pharmacokinetic profiles of the two drugs. Unlike pregabalin, gabapentin exhibits saturable absorption at clinically relevant dosages resulting in nonlinear pharmacokinetics
Since the two relative efficacy estimates for pregabalin and gabapentin were based on indirect comparisons, there is a greater potential for bias than if the comparisons were based on a direct prospective, randomized controlled trial comparing the two agents. Prospective, randomized controlled trials versus placebo were included in this indirect comparative analysis although it is unlikely that randomization would have held in its entirety across the studies. As a result, there was a risk that patients assigned to the different trials were not comparable regarding certain demographics and clinical characteristics. Thus, in order to minimise any bias resulting from this risk, we included in our indirect comparison only those studies that reported and were well matched regarding baseline patient characteristics (e.g., gender, age, duration of disease, baseline seizure rate). Of note, pregabalin was demonstrated to be more efficacious than gabapentin in this analysis despite one of the pregabalin studies including patients with an extremely high baseline seizure rate (21 to 25 seizures per month)
Other sources of potential bias relevant to this analysis were that the included studies used slightly different designs to measure treatment effects, and that there was a relatively large time span over which the studies were conducted and published (1990 to 2005). The use of slightly different designs to measure treatment effects is not expected to impact on the conclusions of the analysis. The timing bias is more likely to be unfavourable for pregabalin. In particular there was significant growth between 1990 and 2005 in the number of alternative addon treatments for partial epilepsy such that if patients felt they were not adequately responding to pregabalin then they were probably more likely to switch to an alternative treatment option.
All three efficacy analyses used in this study provide useful information as to the relative merits of pregabalin and gabapentin. ITT LOCF analysis will tend to overestimate response rates to pharmacotherapies in partial epilepsy trials
The results generated by this analysis are consistent with findings from two metaanalyses of randomized, controlled doubleblind trials, in which adjunctive pregabalin (RR 3.56; 95% CI 2.60 to 4.87; n = 1397) and adjunctive gabapentin (OR 1.93; 95% CI 1.37 to 2.71, n = 997) were significantly more likely than placebo to provide a ≥50% reduction in baseline seizures in patients with treatmentresistant partial epilepsy
Conclusions
The current analysis demonstrates that pregabalin, administered twice or three times daily for up to 12 weeks, is likely to be a more efficacious adjunctive treatment for partial epilepsy than gabapentin three times daily. Although an indirect comparison has certain inherent limitations, we believe that the performed analysis has relevancy for clinical decisionmaking in the treatment of partial epilepsy, in particular until the results of the direct headtohead comparison become available.
Competing interests
This study was funded by Pfizer Inc. Philippa Delahoy and Sally Thompson are fulltime employees of Pfizer Inc. Ian Marschner was an employee of Pfizer Inc. for part of the time that this study was conducted.
Authors' contributions
PD and IM designed the study. PD conducted the systematic review and metaanalysis, supported by IM. All authors were involved with interpreting the data statistically and clinically. All authors drafted, read and approved the final manuscript.
Acknowledgements
Editorial and administrative support for the development of this paper was provided by Malcolm Darkes and Beth Young of UBC Scientific Solutions, and was funded by Pfizer Inc.
Prepublication history
The prepublication history for this paper can be accessed here: