Are we ready for a randomized trial of valproic acid in newly diagnosed glioblastoma?

Valproic acid is a well-established anti-epileptic drug with a still enigmatic mode of action. It has been proposed to work by inhibiting various ion channels or promoting GABA signaling; it also exhibits pharmacodynamic properties that are likely unrelated to its anti-epileptic activity, including inhibition of histone deacetylases. In contrast to classic enzyme-inducing anti-epileptic drugs, valproic acid is a multiple enzyme inhibitor, targeting UDP-glucuronosyltransferase, epoxide hydroxylase, and CYP2C coenzymes, among others.1 The first therapeutic applications of valproic acid in neuro-oncology, beyond as an anti-epileptic drug, stem from the pediatric literature, where it was believed to induce differentiation of tumor cells. In recent years, a growing list of retrospective analyses has demonstrated that glioma patients exposed to valproic acid have a superior outcome;2–4 this list includes an analysis of patients treated within the registration trial of temozolomide for newly diagnosed glioblastoma.5 In the present issue of Neuro-Oncology, De Wit-Kerkhof and colleagues6 studied the effect of valproic acid on seizure control and survival in patients with glioblastoma who were diagnosed and treated from July 1999 until September 2011. Of 291 patients, 181 had seizures and were treated with valproic acid, levetiracetam, or both and had a minimum follow-up of 6 months; 108 of these patients were treated with temozolomide and valproic acid for at least 3 months. The analysis of the anti-epileptic drug's efficacy was limited to patients with a minimum follow-up of 6 months. For the analysis of survival effects, a minimum duration of 3 months of the combination of temozolomide and valproic acid was required. Both these aspects of the study design introduce a significant bias that needs to be considered when analyzing the data; these measures select for favorable-course patients because progression would have led to discontinuation of temozolomide, increased seizure activity and a change in the anti-epileptic drug regimen, or rapid deterioration that is incompatible with further follow-up. Freedom from seizures was achieved with monotherapy in 41 of 100 patients treated initially with valproic acid (41%) and 16 of 37 patients treated initially with levetiracetam (43%). After using these drugs, alone or in combination, 89 of 116 patients (77%) were seizure-free overall. For the survival studies, the authors analyzed 108 patients treated with valproic acid and temozolomide for at least 3 months. They were compared with 57 patients treated with valproic acid for less than 3 months, another anti-epileptic drug, or no anti-epileptic drug at all. These two groups were reportedly similar regarding major patient characteristics, including O6-methylguanine DNA methyltransferase (MGMT) promoter methylation status. Patients treated with valproic acid had a median survival duration of 69 weeks (95% CI, 61.7-67.3 weeks) compared with 61 weeks (95% CI, 52.5-69.5 weeks) in those not receiving valproic acid for 3 months (HR, 0.63; 95% CI, 0.43-0.92) (P = .016). The difference in progression-free survival duration between these groups was not significant (P = .06). Surprisingly, only age and valproic acid treatment, not extent of resection or MGMT promoter methylation status, were significant prognostic factors on multivariate analysis in this subgroup of 165 patients. While this analysis adds to the literature supporting a disease course-modifying role for valproic acid in glioblastoma, it has several limitations: the patients' data were collected over a long time period, MGMT promoter methylation data were available for less than half the patients, and the inclusion criteria for analysis in the subpopulations introduced a bias, as outlined above, that is difficult to estimate but likely has an impact. This bias is illustrated by the fact that neither complete resection nor MGMT status retained prognostic significance on multivariate analysis. Besides small numbers, notably for MGMT, this striking observation is most likely explained by poor-prognosis patients having been a priori eliminated from this analysis by its very design. This is corroborated by data that show a total resection rate of about 80%, which is double the rate usually reported, and a relatively higher rate of MGMT promoter-methylated patients in the valproic acid group. Interestingly, as in the analysis from the EORTC NCIC-trial,5 valproic acid given early in the disease course had no effect on progression-free survival duration, but only an effect on overall survival. This is difficult to explain on a biological basis unless we assume once more that we are not able to reliably measure progression or that valproic acid pretreatment primes glioblastomas for favorable responses to salvage therapies. The present study supports the survival-promoting effect of valproic acid in glioblastoma. However, the analyses supporting this view carry inherent limitations: they are all retrospective, none of them studied in-depth the duration and intensity of exposure to the drug, and the number of patients was relatively small. In the absence of promising pharmacological agents for treating newly diagnosed glioblastoma, it is obvious that there is significant interest in exploring the possible inclusion of valproic acid into the standards of care. However, this would require a randomized trial, which would be a large, rather expensive trial if conducted in the US or Europe without a commercial sponsor to support such an endeavor. This does not mean that such a trial should not be performed if the scientific community believes that it should be done. However, before such a clinical trial concept moves forward, it is mandatory that we use contemporary prospective clinical trial populations for whom co-medication data have been captured with precision, including RTOG0525, CENTRIC, and AVAGlio. A joint analysis of these patient populations that explores the potential impact of co-medication with valproic acid should guide us in determining whether the next step, a randomized phase II trial of valproic acid in newly diagnosed glioblastoma, should be undertaken.