- Email: [email protected]
Historical data in the design and interpretation of trials with newly diagnosed patients
Reviews / Epilepsy Research 68 (2006) 19–94
monotherapy. Three additional unpublished trials were discovered by direct inquiry of companies with new AEDs in development. All studies used similar inclusion criteria, including history of refractory partial seizures, occurrence of at least 3–4 seizures/month, and presence of one or two antiepileptic drugs at baseline. The drugs under study included gabapentin, tiagabine, lamotrigine, topiramate, felbamate, oxcarbazepine, and a “drug in development”. The analysis used percent of patients exiting the trial as an outcome measure. Four nearly identical exit criteria were used in nine of ten trials. The tenth trial used only three exit criteria, and only one matched a criterion used in the other studies. Therefore, this trial was excluded from the analysis. Another trial was excluded because its eligibility criteria allowed enrollment of less refractory patients. Trial duration varied from 112 to 188 days. For the sake of conformity, analysis of outcome was restricted to 112 days after the start of drug taper. One study withdrew baseline AEDs before randomization. This study fit in with the studies in which withdrawal occurred after randomization although, as might be expected, the withdrawal rate was slightly faster, and the overall exit rate was higher. This could be related to the fact that there was only one rather than two drugs present during the withdrawal period (the pseudoplacebo and the drug being withdrawn). Withdrawal from carbamazepine did not increase the likelihood of exit. There was no change in outcome when studies performed earlier were compared to studies performed more recently, even though new AEDs had been introduced in the marketplace. Rate of withdrawal differed, causing slight differences in shape of curves, but overall percent exit remained surprisingly consistent. The percent exiting was uniformly high, ranging from 74.9 to 96.2%, with half the studies having an estimated percent exiting between 83.3 and 87.5%. The estimate of the combined percent exit based on the mixed effects model is 86.1 with a 99% confidence interval from 78.6 to 93.6%. Similarly, a non-iterative approach yielded an estimate of 85.8% with a 99% confidence interval from 76.9 to 94.7%.
3. Conclusions These results indicate that the behavior of patients randomized to pseudo placebo, and subsequently with-
77
drawn towards monotherapy is predictable and reproducible. Over 70% will worsen and meet exit criteria. These data may be able to serve as a historical control for future monotherapy studies, obviating the need for a placebo/pseudo placebo arm.
Reference Gilliam, F.G., Veloso, F., Bomhof, M.A., et al., 2003. A dosecomparison trial of topiramate as monotherapy in recently diagnosed partial epilepsy. Neurology 28, 196–202. Karlawish, J.H., French, J., 2001. The ethical and scientific shortcomings of current monotherapy epilepsy trials in newly diagnosed patients. Epilepsy Behav. 2, 193–200. Wang, S., French, J.A., Pledger, G., Temkin, N., 2002. Meta-analysis of withdrawal to monotherapy studies: basis for historical control. Epilepsia 43 (Suppl. 7), 189 (Abstract).
doi:10.1016/j.eplepsyres.2005.09.027 Historical data in the design and interpretation of trials with newly diagnosed patients夽 Michel Baulac Abstract An International League Against Epilepsy (ILAE) subcommission is exploring the possibility of utilizing historical data from treated and control (untreated or under-treated) patients who have been enrolled in monotherapy trials in newly diagnosed epilepsy. Active-control trials have not convincingly distinguished between equivalent effectiveness and equivalent ineffectiveness. Thus, there is insufficient historical evidence of sensitivity to drug effects. Noninferiority trials are based upon experience acquired with previous trials. Optimizing the exploitation of historical data obtained in newly diagnosed patients may be a way to improve the validity of future trials. A concern about use of historical controls is lack of assurance that historical populations and newly recruited populations are similar. The best approach to pooling and modeling 夽 This paper summarizes the work and the project of a subcommission of the International League Against Epilepsy, including J. French, P. Williamson, A. Marson, E. Perucca, T. Tomson and M. Baulac.
Reviews / Epilepsy Research 68 (2006) 19–94
78
data may be an Individual Patient Data (IPD) approach as (1) analysis of data from active-control trials, (2) demonstration that the subpopulation carrying an intermediate probability of recurrence is the most sensitive to drug effect, by comparison to similar subpopulations who received a placebo or a pseudoplacebo in historical studies. This subpopulation could become a reference for patient selection or stratification in future trials to improve assay sensitivity. Contents 1. 2. 3. 4. 5. 6.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . Noninferiority objective with active control trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Historical controls as external placebo arm? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Potential sources of historical data . . . . . . . Potential strategies . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
78 78 80 80 80 81 81
1. Introduction Patients with newly diagnosed epilepsy constitute an ensemble of clinical situations where a monotherapy treatment is actually employed in medical practice. Since the use of placebo or pseudo-placebo as control treatment is generally considered unethical in newly diagnosed patients, active control trials are usually carried out. They are adapted to the clinical objectives of monotherapy treatment, with follow-up durations permitting assessment of long term efficacy and tolerability, seizure freedom being the primary objective in these patients. However, existing active control trials that have been designed to demonstrate the superiority of a new AED over an established comparator (usually carbamazepine), have failed to do so. It turned out that showing noninferiority was a more realistic objective. But due to methodological problems, the validity of noninferiority designs regarding demonstration of efficacy is debated and not recognized by every regulatory body. A few studies have included a placebo or pseudoplacebo arm, however, and they contribute to some amount of historical control data. Would it be possible to utilize this historical data to circumvent some of
the difficulties that are inherent to trials in newly diagnosed patients? This reflection follows the example of the proposal regarding conversion to monotherapy in refractory patients (French, 2005).
2. Noninferiority objective with active control trials 2.1. Noninferiority trials, to be valid, must meet several requirements. There must be evidence, direct or indirect, that the comparator is showing its usual level of efficacy. The trial should be strictly comparable to earlier trials that demonstrated the efficacy of the control agent. These similarly designed trials should have been consistently able to find a difference between an active drug and a less active treatment. This ability to distinguish an effective treatment from a less effective or ineffective one defines assay sensitivity. Such historical evidence of sensitivity to drug effects is insufficient regarding trials with newly diagnosed patients. The notion that an established treatment, carbamazepine for instance, reliably shows superiority to placebo is insufficiently documented. Active-control trials, so far, have not convincingly distinguished between equivalent effectiveness and equivalent ineffectiveness, although a few trials demonstrated superiority of the active control over the test drug (Brodie et al., 2002). 2.2. A noninferiority demonstration is also based on the predefinition of a noninferiority margin, which represents the maximum loss of effect considered as clinically acceptable versus the comparator. This margin should be determined on statistical reasoning, clinical judgment and past experience with placebo-controlled trials. In an objective of demonstrating efficacy, this margin cannot be greater than the smallest effect size that the active control would be reliably expected to have compared with placebo. In newly diagnosed epilepsy, the respective positions of the active control (carbamazepine most of times) and placebo can only be extrapolated, which makes difficult the determination of a valid noninferiority margin. Noninferiority trials are therefore strongly based upon experience acquired with previous, similar trials. Optimizing the exploitation of historical data obtained in newly diagnosed patients may be a way to improve the validity of future trials (Table 1).
Table 1 Summary of trials in newly diagnosed epilepsy that employed a placebo or pseudoplacebo group GBP 3600/GBP 300 (Beydoun, 1999)
GBP 900, 1800/300, open CBZ (Chadwick et al., 1998)
OXC 1200/Placebo (Sachdeo et al., 2001)
TPM 200–500/TPM 25–50 (Gilliam et al., 2003)
TPM 400/TPM 50 (Arroyo et al., 2005)
Epilepsy type Time since 1st seizure
Partial
Partial
Partial
Partial Median 8.5 months
Previously treated
40% with baseline AED
No baseline AED
No baseline AED
40% with baseline AED
Primary outcome
Time to 2nd PS
Time to 3rd PS
Time to 2nd seizure
Exit criteria for efficacy N patients Adult children Duration Titration Control group N controls
1GTC, 2PS, SE 146 Range 7–80 years 4 months
1GTC, 3PS, SE 292 Range 12–86 years 8 months
Time to exit, 1st seizure 1 PS 67 Range 10–69 years 3 months
Partial/generalized 1 month or untreated relapse No baseline AED except emergency Time to 1st seizure
GBP 300 mg 70
GBP 300 mg 72
Placebo 35
1GTC, 2PS, SE 252 15% <16 years Mean 6 months Yes TPM 25–50 g 125
1 GTC, 1PS 470 32% <16 years Median 9 months Yes TPM 50 mg 234
Baseline sz in control group
30% 1 sz in 3 months 70% 2–10 sz in 6 months
Median 5.5 sz/month 15 sz/3 months
1 sz: 38%/3 months 2 sz: 24%/3 months >3 sz: 38% 3/month
1 or 2 in last 3 months
Seizure free control group Primary endpoint
49% (4 month) Non significant
12% (3 month) OXC 1200 > Placebo
39–43% (∼6 month) Non significant
71% (6 month) TPM 400 > 50 mg
900 > 1800 = CBZ > 300 mg
Reviews / Epilepsy Research 68 (2006) 19–94
Trial
GBP: gabapentin; OXC: oxcarbazepine; TPM: topiramate; PS: partial seizures; SE: status epilepticus; GTC: generalized tonic-clonic seizure; Sz: seizure.
79
80
Reviews / Epilepsy Research 68 (2006) 19–94
3. Historical controls as external placebo arm? 3.1. If a reliable placebo or pseudoplacebo group outcome can be extracted from previous trials in newly diagnosed epilepsy, how could this data be utilized in future trials? Following the example of the proposal made for conversion to monotherapy, a simple way would be to design a superiority trial where the test drug would have to be superior to the historical placebo group. Historical control groups, however, are generally not recommended for regulatory purposes (ICH, 2000), excepted when the treatment effect is dramatic or the course of the disease highly predictable, but such is not the case with newly diagnosed epilepsy. The major issue is that a strict similarity between historical populations and newly recruited populations cannot be ensured. The historical control group can be dissimilar with respect to a wide range of factors that could affect outcome, including: demographic characteristics, diagnosis criteria, stage or severity, concomitant treatments, observational conditions, evolution of practice. It can be also dissimilar regarding the conditions of blinding of the patients and/or of the investigators. A few examples, however, can be found in literature where historical controls were utilized in the purpose of improving the data submitted to regulatory agencies. In a study comparing drugs for metastatic breast cancer, the primary objective was not reached, and a switch from superiority to noninferiority was made acceptable by the introduction of a virtual placebo arm extracted from earlier studies (Durrleman and Chaikin, 2003). A study on prevention of cardiovascular events, showed a statistical superiority of the test drug, clopidogrel, over aspirin (Fisher et al., 2001). Since this superiority was modest from a clinical perspective, and based on a single study, some supportive data was requested. Several earlier studies comparing aspirin versus placebo were available, and the comparison with their metaanalysis permitted to show superiority of clopidogrel over historical placebo. 3.3. Following these examples, another suggestion about the way to employ an external placebo-control group would be as a “third arm” in designs comparing two active drugs. Noninferiority would be validated by showing that both established comparator and test drug remain superior to the virtual, historical placebo. This, again, would imply a strict similarity in terms of
designs and patients characteristics with the historical trials, and prior evidence of assay sensitivity.
4. Potential sources of historical data 4.1. A few monotherapy trials in recently diagnosed epilepsy have employed a placebo or a pseudoplacebo group. They are summarized in the Table. These trials show a great deal of heterogeneity. The subtherapeutic treatments, placebo, gabapentin 300 mg, topiramate 25 or 50 mg, are different. It is not sure that 50 mg of topiramate is actually a subtherapeutic dosage. There are also differences regarding epilepsy syndromes (generalized seizures were included in one study), age ranges, trial durations, evaluation variables or exit criteria for lack of efficacy (e.g. third, second or first seizure). The numbers of pre-randomization seizures were also markedly different, a parameter which is known to impact study outcomes (Gilliam et al., 2003). 4.2. The percentages of patients remaining seizurefree in the subtherapeutic arms varied from 12 to 71%, over periods varying between 3 and 6 months. These disparities make it very difficult to extract a consistent outcome for the placebo or subtherapeutic groups. Superiority of active treatment over sub-optimal treatment was not demonstrated in two of these five studies, which further raises the question of the capability of trials in newly diagnosed patients to find consistently a difference between active and less active treatments. 4.3. Available data in terms of placebo and pseudoplacebo groups is relatively scarce in newly diagnosed patients. All other potential sources of data should be exploited. The MESS study (Multicentre Study of Early Epilepsy and Single Seizures) in which patients were randomized either to immediate treatment or to no treatment (or deferred treatment) may be a useful source of control data in untreated patients (Marson et al., 2005).
5. Potential strategies 5.1. A project attempting to better exploit historical data from earlier trials in newly diagnosed patients was undertaken recently (ad hoc ILAE sub commission). It consists in performing meta-analyses of existing data and in addressing three main questions:
Reviews / Epilepsy Research 68 (2006) 19–94
a. Can existing data be synthesized and modeled to reliably predict the outcome for patients receiving active AED treatment and for patients receiving no treatment or a subtherapeutic treatment (placebo and pseudo-placebo)? b. Can these models be used to select a population for whom there is the greatest effect of AED, and would hence have ‘assay sensitivity’ in future trials? c. Can the models be used, as external validation with historical data, to help interpret future actively controlled AEDs monotherapy studies: (i) by demonstrating that the actively treated patients responded to treatment as predicted? (ii) by comparing the outcomes between actively treated patients and historical controls? (iii) by helping to predefine a relevant noninferiority margin when a noninferiority analysis versus active comparator is contemplated. 5.2. As outlined above, there is heterogeneity among existing trials with respect to patient characteristics, trial design and choice of outcomes. The best approach to pooling and modeling data will therefore be an Indvidual Patient Data (IPD) approach (Williamson et al., 2000; Smith et al., 2005). This will allow a thorough investigation of the influence of patient factors upon outcome. This approach also allows the creation of a uniform set of outcome data. A first phase will consist of analysing data from actively treated patients, previously included in randomized controlled trials or in randomized studies. The hypothesis is that distinct subpopulations can be identified, for instance subpopulations in whom there is a low, a high or an intermediate risk of seizure relapse. As a second step, it may be shown that the subpopulation carrying an intermediate probability of recurrence is the most sensitive to drug effect, by comparison to similar subpopulations who received a placebo or a pseudoplacebo in historical studies. This subpopulation could hence become a reference for patient selection or stratification in future trials, and help improving assay sensitivity.
6. Conclusion In conclusion, several supportive techniques may be envisioned for trials in newly diagnosed patients, based on the incorporation of more historical data in the
81
design, analysis and interpretation of active control trials. All potential randomized sources of historical data should be exploited. Identifying different subpopulations with different outcomes may be a way to optimize patient recruitment or stratification in future trials.
Reference Arroyo, S., Dodson, W.E., Privitera, M.D., Glauser, T.A., Naritoku, D.K., Dlugos, D.J., Wang, S., Schwabe, S.K., Twyman, R.E., for the EPMN-106/INT-28 Investigators, 2005. Randomized dosecontrolled study of topiramate as first-line therapy in epilepsy. Acta Neurol. Scand. 112, 214–222. Beydoun, A., 1999. Monotherapy trials with gabapentin for partial epilepsy. Epilepsia 40 (Suppl. 6), S13–S16, discussion S73–S74. Brodie, M.J., Wroe, S.J., Dean ADP, Holdich TAH, Whitehead, J., Stevens, J.W., 2002. Efficacy and safety of remacemide versus carbamazepine in newly diagnosed epilepsy: comparison by sequential analysis. Epilepsy Behav. 3, 140–146. Chadwick, D.W., Anhut, H., Greiner, M.J., et al., 1998. A doubleblind trial of gabapentin monotherapy for newly diagnosed partial seizures. Neurology 51, 1282–1288. Durrleman, S., Chaikin, P., 2003. The use of putative placebo in active control trials: two applications in a regulatory setting. Stat. Med. 22, 941–952. Fisher, L.D., Gent, M., Buller, H.R., 2001. Active control trials: how would a new agent compare with placebo: a method illustrated with clopidogrel, aspirin and placebo. Am. Heart J. 141, 26–32. French, J., this issue. Gilliam, F.G., Veloso, F., Bomhof, M.A., et al., 2003. A dose comparison trial of topiramate as monotherapy in recently diagnosed partial epilepsy. Neurology 60, 196–202. International Commission on Harmonization, 2000. Note for guidance of the choice of control group in clinical trial. ICH topic E10. Marson, A., Jacoby, A., Johnson, A., Kim, L., Gamble, C., Chadwick, D., on behalf of the Medical Research Council MESS Study Group, 2005. Immediate versus deferred antiepileptic drug treatment for early epilepsy and single seizures: a randomized controlled trial. Lancet 365, 2007–2013. Sachdeo, R., Beydoun, A., D’Souza, J., et al., 2001. Oxcarbazepine (Trileptal) as monotherapy in patients with partial seizures. Neurology 57, 864–871. Smith, C.T., Williamson, P.R., Marson, A.G., 2005. Investigating heterogeneity in an individual patient data meta-analysis of time to event outcomes. Stat. Med. 24, 1307–1319. Williamson, P.R., Marson, A.G., Tudur, C., Hutton, J.L., Chadwick, D., 2000. Individual patient data meta-analysis of randomized anti-epileptic drug monotherapy trials. J. Eval. Clin. Pract. 6, 205–214.
doi:10.1016/j.eplepsyres.2005.09.028
monotherapy. Three additional unpublished trials were discovered by direct inquiry of companies with new AEDs in development. All studies used similar inclusion criteria, including history of refractory partial seizures, occurrence of at least 3–4 seizures/month, and presence of one or two antiepileptic drugs at baseline. The drugs under study included gabapentin, tiagabine, lamotrigine, topiramate, felbamate, oxcarbazepine, and a “drug in development”. The analysis used percent of patients exiting the trial as an outcome measure. Four nearly identical exit criteria were used in nine of ten trials. The tenth trial used only three exit criteria, and only one matched a criterion used in the other studies. Therefore, this trial was excluded from the analysis. Another trial was excluded because its eligibility criteria allowed enrollment of less refractory patients. Trial duration varied from 112 to 188 days. For the sake of conformity, analysis of outcome was restricted to 112 days after the start of drug taper. One study withdrew baseline AEDs before randomization. This study fit in with the studies in which withdrawal occurred after randomization although, as might be expected, the withdrawal rate was slightly faster, and the overall exit rate was higher. This could be related to the fact that there was only one rather than two drugs present during the withdrawal period (the pseudoplacebo and the drug being withdrawn). Withdrawal from carbamazepine did not increase the likelihood of exit. There was no change in outcome when studies performed earlier were compared to studies performed more recently, even though new AEDs had been introduced in the marketplace. Rate of withdrawal differed, causing slight differences in shape of curves, but overall percent exit remained surprisingly consistent. The percent exiting was uniformly high, ranging from 74.9 to 96.2%, with half the studies having an estimated percent exiting between 83.3 and 87.5%. The estimate of the combined percent exit based on the mixed effects model is 86.1 with a 99% confidence interval from 78.6 to 93.6%. Similarly, a non-iterative approach yielded an estimate of 85.8% with a 99% confidence interval from 76.9 to 94.7%.
3. Conclusions These results indicate that the behavior of patients randomized to pseudo placebo, and subsequently with-
77
drawn towards monotherapy is predictable and reproducible. Over 70% will worsen and meet exit criteria. These data may be able to serve as a historical control for future monotherapy studies, obviating the need for a placebo/pseudo placebo arm.
Reference Gilliam, F.G., Veloso, F., Bomhof, M.A., et al., 2003. A dosecomparison trial of topiramate as monotherapy in recently diagnosed partial epilepsy. Neurology 28, 196–202. Karlawish, J.H., French, J., 2001. The ethical and scientific shortcomings of current monotherapy epilepsy trials in newly diagnosed patients. Epilepsy Behav. 2, 193–200. Wang, S., French, J.A., Pledger, G., Temkin, N., 2002. Meta-analysis of withdrawal to monotherapy studies: basis for historical control. Epilepsia 43 (Suppl. 7), 189 (Abstract).
doi:10.1016/j.eplepsyres.2005.09.027 Historical data in the design and interpretation of trials with newly diagnosed patients夽 Michel Baulac Abstract An International League Against Epilepsy (ILAE) subcommission is exploring the possibility of utilizing historical data from treated and control (untreated or under-treated) patients who have been enrolled in monotherapy trials in newly diagnosed epilepsy. Active-control trials have not convincingly distinguished between equivalent effectiveness and equivalent ineffectiveness. Thus, there is insufficient historical evidence of sensitivity to drug effects. Noninferiority trials are based upon experience acquired with previous trials. Optimizing the exploitation of historical data obtained in newly diagnosed patients may be a way to improve the validity of future trials. A concern about use of historical controls is lack of assurance that historical populations and newly recruited populations are similar. The best approach to pooling and modeling 夽 This paper summarizes the work and the project of a subcommission of the International League Against Epilepsy, including J. French, P. Williamson, A. Marson, E. Perucca, T. Tomson and M. Baulac.
Reviews / Epilepsy Research 68 (2006) 19–94
78
data may be an Individual Patient Data (IPD) approach as (1) analysis of data from active-control trials, (2) demonstration that the subpopulation carrying an intermediate probability of recurrence is the most sensitive to drug effect, by comparison to similar subpopulations who received a placebo or a pseudoplacebo in historical studies. This subpopulation could become a reference for patient selection or stratification in future trials to improve assay sensitivity. Contents 1. 2. 3. 4. 5. 6.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . Noninferiority objective with active control trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Historical controls as external placebo arm? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Potential sources of historical data . . . . . . . Potential strategies . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
78 78 80 80 80 81 81
1. Introduction Patients with newly diagnosed epilepsy constitute an ensemble of clinical situations where a monotherapy treatment is actually employed in medical practice. Since the use of placebo or pseudo-placebo as control treatment is generally considered unethical in newly diagnosed patients, active control trials are usually carried out. They are adapted to the clinical objectives of monotherapy treatment, with follow-up durations permitting assessment of long term efficacy and tolerability, seizure freedom being the primary objective in these patients. However, existing active control trials that have been designed to demonstrate the superiority of a new AED over an established comparator (usually carbamazepine), have failed to do so. It turned out that showing noninferiority was a more realistic objective. But due to methodological problems, the validity of noninferiority designs regarding demonstration of efficacy is debated and not recognized by every regulatory body. A few studies have included a placebo or pseudoplacebo arm, however, and they contribute to some amount of historical control data. Would it be possible to utilize this historical data to circumvent some of
the difficulties that are inherent to trials in newly diagnosed patients? This reflection follows the example of the proposal regarding conversion to monotherapy in refractory patients (French, 2005).
2. Noninferiority objective with active control trials 2.1. Noninferiority trials, to be valid, must meet several requirements. There must be evidence, direct or indirect, that the comparator is showing its usual level of efficacy. The trial should be strictly comparable to earlier trials that demonstrated the efficacy of the control agent. These similarly designed trials should have been consistently able to find a difference between an active drug and a less active treatment. This ability to distinguish an effective treatment from a less effective or ineffective one defines assay sensitivity. Such historical evidence of sensitivity to drug effects is insufficient regarding trials with newly diagnosed patients. The notion that an established treatment, carbamazepine for instance, reliably shows superiority to placebo is insufficiently documented. Active-control trials, so far, have not convincingly distinguished between equivalent effectiveness and equivalent ineffectiveness, although a few trials demonstrated superiority of the active control over the test drug (Brodie et al., 2002). 2.2. A noninferiority demonstration is also based on the predefinition of a noninferiority margin, which represents the maximum loss of effect considered as clinically acceptable versus the comparator. This margin should be determined on statistical reasoning, clinical judgment and past experience with placebo-controlled trials. In an objective of demonstrating efficacy, this margin cannot be greater than the smallest effect size that the active control would be reliably expected to have compared with placebo. In newly diagnosed epilepsy, the respective positions of the active control (carbamazepine most of times) and placebo can only be extrapolated, which makes difficult the determination of a valid noninferiority margin. Noninferiority trials are therefore strongly based upon experience acquired with previous, similar trials. Optimizing the exploitation of historical data obtained in newly diagnosed patients may be a way to improve the validity of future trials (Table 1).
Table 1 Summary of trials in newly diagnosed epilepsy that employed a placebo or pseudoplacebo group GBP 3600/GBP 300 (Beydoun, 1999)
GBP 900, 1800/300, open CBZ (Chadwick et al., 1998)
OXC 1200/Placebo (Sachdeo et al., 2001)
TPM 200–500/TPM 25–50 (Gilliam et al., 2003)
TPM 400/TPM 50 (Arroyo et al., 2005)
Epilepsy type Time since 1st seizure
Partial
Partial
Partial
Partial Median 8.5 months
Previously treated
40% with baseline AED
No baseline AED
No baseline AED
40% with baseline AED
Primary outcome
Time to 2nd PS
Time to 3rd PS
Time to 2nd seizure
Exit criteria for efficacy N patients Adult children Duration Titration Control group N controls
1GTC, 2PS, SE 146 Range 7–80 years 4 months
1GTC, 3PS, SE 292 Range 12–86 years 8 months
Time to exit, 1st seizure 1 PS 67 Range 10–69 years 3 months
Partial/generalized 1 month or untreated relapse No baseline AED except emergency Time to 1st seizure
GBP 300 mg 70
GBP 300 mg 72
Placebo 35
1GTC, 2PS, SE 252 15% <16 years Mean 6 months Yes TPM 25–50 g 125
1 GTC, 1PS 470 32% <16 years Median 9 months Yes TPM 50 mg 234
Baseline sz in control group
30% 1 sz in 3 months 70% 2–10 sz in 6 months
Median 5.5 sz/month 15 sz/3 months
1 sz: 38%/3 months 2 sz: 24%/3 months >3 sz: 38% 3/month
1 or 2 in last 3 months
Seizure free control group Primary endpoint
49% (4 month) Non significant
12% (3 month) OXC 1200 > Placebo
39–43% (∼6 month) Non significant
71% (6 month) TPM 400 > 50 mg
900 > 1800 = CBZ > 300 mg
Reviews / Epilepsy Research 68 (2006) 19–94
Trial
GBP: gabapentin; OXC: oxcarbazepine; TPM: topiramate; PS: partial seizures; SE: status epilepticus; GTC: generalized tonic-clonic seizure; Sz: seizure.
79
80
Reviews / Epilepsy Research 68 (2006) 19–94
3. Historical controls as external placebo arm? 3.1. If a reliable placebo or pseudoplacebo group outcome can be extracted from previous trials in newly diagnosed epilepsy, how could this data be utilized in future trials? Following the example of the proposal made for conversion to monotherapy, a simple way would be to design a superiority trial where the test drug would have to be superior to the historical placebo group. Historical control groups, however, are generally not recommended for regulatory purposes (ICH, 2000), excepted when the treatment effect is dramatic or the course of the disease highly predictable, but such is not the case with newly diagnosed epilepsy. The major issue is that a strict similarity between historical populations and newly recruited populations cannot be ensured. The historical control group can be dissimilar with respect to a wide range of factors that could affect outcome, including: demographic characteristics, diagnosis criteria, stage or severity, concomitant treatments, observational conditions, evolution of practice. It can be also dissimilar regarding the conditions of blinding of the patients and/or of the investigators. A few examples, however, can be found in literature where historical controls were utilized in the purpose of improving the data submitted to regulatory agencies. In a study comparing drugs for metastatic breast cancer, the primary objective was not reached, and a switch from superiority to noninferiority was made acceptable by the introduction of a virtual placebo arm extracted from earlier studies (Durrleman and Chaikin, 2003). A study on prevention of cardiovascular events, showed a statistical superiority of the test drug, clopidogrel, over aspirin (Fisher et al., 2001). Since this superiority was modest from a clinical perspective, and based on a single study, some supportive data was requested. Several earlier studies comparing aspirin versus placebo were available, and the comparison with their metaanalysis permitted to show superiority of clopidogrel over historical placebo. 3.3. Following these examples, another suggestion about the way to employ an external placebo-control group would be as a “third arm” in designs comparing two active drugs. Noninferiority would be validated by showing that both established comparator and test drug remain superior to the virtual, historical placebo. This, again, would imply a strict similarity in terms of
designs and patients characteristics with the historical trials, and prior evidence of assay sensitivity.
4. Potential sources of historical data 4.1. A few monotherapy trials in recently diagnosed epilepsy have employed a placebo or a pseudoplacebo group. They are summarized in the Table. These trials show a great deal of heterogeneity. The subtherapeutic treatments, placebo, gabapentin 300 mg, topiramate 25 or 50 mg, are different. It is not sure that 50 mg of topiramate is actually a subtherapeutic dosage. There are also differences regarding epilepsy syndromes (generalized seizures were included in one study), age ranges, trial durations, evaluation variables or exit criteria for lack of efficacy (e.g. third, second or first seizure). The numbers of pre-randomization seizures were also markedly different, a parameter which is known to impact study outcomes (Gilliam et al., 2003). 4.2. The percentages of patients remaining seizurefree in the subtherapeutic arms varied from 12 to 71%, over periods varying between 3 and 6 months. These disparities make it very difficult to extract a consistent outcome for the placebo or subtherapeutic groups. Superiority of active treatment over sub-optimal treatment was not demonstrated in two of these five studies, which further raises the question of the capability of trials in newly diagnosed patients to find consistently a difference between active and less active treatments. 4.3. Available data in terms of placebo and pseudoplacebo groups is relatively scarce in newly diagnosed patients. All other potential sources of data should be exploited. The MESS study (Multicentre Study of Early Epilepsy and Single Seizures) in which patients were randomized either to immediate treatment or to no treatment (or deferred treatment) may be a useful source of control data in untreated patients (Marson et al., 2005).
5. Potential strategies 5.1. A project attempting to better exploit historical data from earlier trials in newly diagnosed patients was undertaken recently (ad hoc ILAE sub commission). It consists in performing meta-analyses of existing data and in addressing three main questions:
Reviews / Epilepsy Research 68 (2006) 19–94
a. Can existing data be synthesized and modeled to reliably predict the outcome for patients receiving active AED treatment and for patients receiving no treatment or a subtherapeutic treatment (placebo and pseudo-placebo)? b. Can these models be used to select a population for whom there is the greatest effect of AED, and would hence have ‘assay sensitivity’ in future trials? c. Can the models be used, as external validation with historical data, to help interpret future actively controlled AEDs monotherapy studies: (i) by demonstrating that the actively treated patients responded to treatment as predicted? (ii) by comparing the outcomes between actively treated patients and historical controls? (iii) by helping to predefine a relevant noninferiority margin when a noninferiority analysis versus active comparator is contemplated. 5.2. As outlined above, there is heterogeneity among existing trials with respect to patient characteristics, trial design and choice of outcomes. The best approach to pooling and modeling data will therefore be an Indvidual Patient Data (IPD) approach (Williamson et al., 2000; Smith et al., 2005). This will allow a thorough investigation of the influence of patient factors upon outcome. This approach also allows the creation of a uniform set of outcome data. A first phase will consist of analysing data from actively treated patients, previously included in randomized controlled trials or in randomized studies. The hypothesis is that distinct subpopulations can be identified, for instance subpopulations in whom there is a low, a high or an intermediate risk of seizure relapse. As a second step, it may be shown that the subpopulation carrying an intermediate probability of recurrence is the most sensitive to drug effect, by comparison to similar subpopulations who received a placebo or a pseudoplacebo in historical studies. This subpopulation could hence become a reference for patient selection or stratification in future trials, and help improving assay sensitivity.
6. Conclusion In conclusion, several supportive techniques may be envisioned for trials in newly diagnosed patients, based on the incorporation of more historical data in the
81
design, analysis and interpretation of active control trials. All potential randomized sources of historical data should be exploited. Identifying different subpopulations with different outcomes may be a way to optimize patient recruitment or stratification in future trials.
Reference Arroyo, S., Dodson, W.E., Privitera, M.D., Glauser, T.A., Naritoku, D.K., Dlugos, D.J., Wang, S., Schwabe, S.K., Twyman, R.E., for the EPMN-106/INT-28 Investigators, 2005. Randomized dosecontrolled study of topiramate as first-line therapy in epilepsy. Acta Neurol. Scand. 112, 214–222. Beydoun, A., 1999. Monotherapy trials with gabapentin for partial epilepsy. Epilepsia 40 (Suppl. 6), S13–S16, discussion S73–S74. Brodie, M.J., Wroe, S.J., Dean ADP, Holdich TAH, Whitehead, J., Stevens, J.W., 2002. Efficacy and safety of remacemide versus carbamazepine in newly diagnosed epilepsy: comparison by sequential analysis. Epilepsy Behav. 3, 140–146. Chadwick, D.W., Anhut, H., Greiner, M.J., et al., 1998. A doubleblind trial of gabapentin monotherapy for newly diagnosed partial seizures. Neurology 51, 1282–1288. Durrleman, S., Chaikin, P., 2003. The use of putative placebo in active control trials: two applications in a regulatory setting. Stat. Med. 22, 941–952. Fisher, L.D., Gent, M., Buller, H.R., 2001. Active control trials: how would a new agent compare with placebo: a method illustrated with clopidogrel, aspirin and placebo. Am. Heart J. 141, 26–32. French, J., this issue. Gilliam, F.G., Veloso, F., Bomhof, M.A., et al., 2003. A dose comparison trial of topiramate as monotherapy in recently diagnosed partial epilepsy. Neurology 60, 196–202. International Commission on Harmonization, 2000. Note for guidance of the choice of control group in clinical trial. ICH topic E10. Marson, A., Jacoby, A., Johnson, A., Kim, L., Gamble, C., Chadwick, D., on behalf of the Medical Research Council MESS Study Group, 2005. Immediate versus deferred antiepileptic drug treatment for early epilepsy and single seizures: a randomized controlled trial. Lancet 365, 2007–2013. Sachdeo, R., Beydoun, A., D’Souza, J., et al., 2001. Oxcarbazepine (Trileptal) as monotherapy in patients with partial seizures. Neurology 57, 864–871. Smith, C.T., Williamson, P.R., Marson, A.G., 2005. Investigating heterogeneity in an individual patient data meta-analysis of time to event outcomes. Stat. Med. 24, 1307–1319. Williamson, P.R., Marson, A.G., Tudur, C., Hutton, J.L., Chadwick, D., 2000. Individual patient data meta-analysis of randomized anti-epileptic drug monotherapy trials. J. Eval. Clin. Pract. 6, 205–214.
doi:10.1016/j.eplepsyres.2005.09.028