Current issues of ACP Journal Club are published in Annals of Internal Medicine


Lamotrigine was more effective than carbamazepine, gabapentin, and topiramate for treatment failure in partial epilepsy


ACP J Club. 2007 Nov-Dec;147:74. doi:10.7326/ACPJC-2007-147-3-074

Related Content in this Issue
• Companion Abstract and Commentary: Valproate was more effective than lamotrigine and better tolerated than topiramate in generalized or unclassified epilepsy

Clinical Impact Ratings

Neurology: 7 stars

Source Citation

Marson AG, Al-Kharusi AM, Alwaidh M, et al. The SANAD study of effectiveness of carbamazepine, gabapentin, lamotrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an unblinded randomised controlled trial. Lancet. 2007;369:1000-15. [PubMed ID: 17382827]



In patients diagnosed with partial-onset seizures, how do carbamazepine (CBZ), gabapentin (GPT), lamotrigine (LTG), oxcarbazepine (OXC), and topiramate (TPM) compare?


Design: Randomized controlled trial (Standard and New Antiepileptic Drugs [SANAD] trial, Arm A)

Allocation: Concealed.*

Blinding: Unblinded.*

Follow-up period: Up to 6 years.

Setting: Hospital-based outpatient clinics in the United Kingdom.

Patients: 1721 patients (mean age 38 y, 55% men) who had ≥ 2 clinically definite, unprovoked epileptic seizures in the past year and were recommended to take CBZ over valproate by the recruiting clinician. Exclusion criteria were age ≤ 4 years, acute symptomatic seizures (including febrile seizures), history of progressive neurologic disease, and contraindications to treatment.

Intervention: CBZ (n = 378), GPT (n = 377), LTG (n = 378), OXC (n = 210), or TPM (n = 378).

Outcomes: Treatment failure (i.e., stopping drug because of inadequate seizure control, intolerable side effects, or both or addition of other antiepileptic drugs) and 1-year remission. Secondary outcomes included 2-year remission and time to first seizure.

Patient follow-up: 96% (intention-to-treat analysis).

Main results

For treatment failure, LTG was more effective than CBZ, GPT, and TPM; OXC was more effective than GPT (Table); LTG and OXC did not differ; CBZ did not differ from GPT, TPM, or OXC; TPM did not differ from GPT or OXC. For 1-year remission, CBZ was more effective than GPT; LTG was more effective than GPT; OXC was more effective than GPT (Table); CBZ did not differ from LTG, OXC, or TPM; TPM did not differ from GPT, LTG, or OXC; LTG and OXC did not differ. For 2-year remission, CBZ was more effective than GPT and TPM (Table); OXC and CBZ did not differ; LTG and TPM did not differ. The CBZ group had fewer first seizures than did the GPT and LTG groups (Table) but did not differ from the OXC or TPM groups.


Lamotrigine was more effective than carbamazepine, gabapentin, and topiramate for treatment failure in partial-onset seizures.

*See Glossary.

Source of funding: Health Technology Assessment Programme.

For correspondence: Dr. A.G. Marson, University of Liverpool, Liverpool, England, United Kingdom. E-mail

Table. Comparisons of carbamazepine (CBZ), gabapentin (GPT), lamotrigine (LTG), oxcarbazepine (OXC), and topiramate (TPM) in partial epilepsy†

Outcomes at ≤ 6 y Comparisons Event rates RRR (95% CI) NNT (CI)
Treatment failure LTG vs CBZ 41% vs 47% 17% (2.2 to 30) 13 (8 to 99)
LTG vs GPT 41% vs 56% 26% (14 to 38) 7 (5 to 13)
LTG vs TPM 41% vs 54% 28% (15 to 39) 7 (5 to 13)
OXC vs GPT 44% vs 56% 18% (1.3 to 33) 11 (6 to 138)
First seizure CBZ vs GPT 72% vs 80% 11% (5 to 15) 12 (9 to 25)
CBZ vs LTG 72% vs 79% 7.9% (1.6 to 13) 16 (10 to 80)
1-y remission CBZ vs GPT‡ 70% vs 60% 17% (5.2 to 31) 10 (6 to 31)
LTG vs GPT 67% vs 60% 12% (0.6 to 23) 15 (8 to 275)
OXC vs GPT 64% vs 60% 19% (3.6 to 34) 9 (5 to 47)
2-y remission CBZ vs GPT‡ 46% vs 37% 21% (5.7 to 34) 11 (6 to 41)
CBZ vs TPM‡ 46% vs 39% 16% (0.2 to 29) 14 (7 to 1433)

†Abbreviations defined in Glossary. RRR, RBI, NNT, and CI calculated from hazard ratios and control event rates in article.
‡RBI, NNT, and CI calculated from data in article.


After deciding to begin treatment for patients with epilepsy, physicians have a wide array of drugs from which to choose. Because some drugs are better at preventing specific types of seizures and some types of seizures may be exacerbated by specific drugs, physicians must first attempt to determine whether the seizures are of partial (focal) or generalized onset. The results of the SANAD trial by Marson and colleagues may then be used as a guide for the choice of medications.

Although the unblinded nature of the studies might have introduced some bias, the practical aspects of masking treatment allocation (e.g., developing identical dosage forms for different medications and the different titration schedules for each drug) in such large trials would have been complicated and expensive, if not impossible.

Because the study of partial epilepsy showed that the 5-year probability of remaining on the best drug (LTG) was < 60%, it can be inferred that none of the drug choices are particularly good. The study design considered treatment failure to be either lack of adequate control or intolerable side effects, which reflects the real-world use of anticonvulsants. This is often difficult to measure in studies because patients may consider some adverse effects (e.g., weight gain) to be tolerable if seizure control is complete but intolerable if seizures lead to lifestyle restrictions.

The findings suggest that patients can reasonably start treatment with either CBZ or LTG, but treatment with GPT or TPM is less likely to be successful at the doses compared. The lack of diplopia may indicate that the final dose of CBZ could have been higher, but the design and execution of the trial is otherwise admirable. The differences between CBZ and LTG were not important, suggesting that other factors, such as costs of the drugs and ancillary testing, may be reasons to choose 1 drug first. The study also underscores the likelihood that patients may have to switch drugs 1 or more times, a point that physicians should stress at each visit.

Although VAL is not a useful first-choice drug for partial epilepsy, its inclusion in the study would have provided useful data for making therapeutic decisions.

The cost–benefit analysis presented is dependent on the way that drugs are purchased and may not be applicable in countries other than the United Kingdom.

In the study of treatment of generalized epilepsy, the newer drugs (LTG and TPM) were not better at controlling seizures than the older, established VAL. However, individual patients will require different approaches to treatment, especially when pregnancy is involved and LTG is the better choice (1).

Thomas P. Bleck, MD
Northwestern University and Evanston Northwestern Healthcare
Evanston, Illinois, USA


1. Cunnington M and Tennis P. Lamotrigine and the risk of malformations in pregnancy. Neurology. 2005;64:955-60. [PubMed ID: 15781807]