The German trial on Aciclovir and Corticosteroids in Herpes-simplex-virus-Encephalitis (GACHE): a multicenter, randomized, double-blind, placebo-controlled trial

The German trial of Aciclovir and Corticosteroids in Herpes-simplex-virus-Encephalitis (GACHE) was a prospective, multicenter, randomized, double blind, placebo-controlled, parallel group clinical trial in adult patients with recent onset of HSVE to evaluate the effect of early adjuvant corticosteroids (dexamethasone) on morbidity and mortality in the treatment of adult patients with HSVE [1]. Patients were planned to be recruited in 27 Departments of Neurology of academic medical centers in Germany, Austria and The Netherlands; only centers in Germany actually recruited patients. Recruitment rate was based on the incidence of HSVE (2–4 cases/1,000,000). In addition a survey was conducted in all participating centers estimating the anticipated number of cases. All centers expected between two and five yearly cases of HSVE. Full ethical approval for this study was obtained from all responsible Ethics Committees. An independent data safety monitoring board reviewed the study regularly.

All patients between 18 and 85 years of age with focal neurological signs for not more than 5 days prior to admission to the respective study center were consecutively screened for eligibility and, if considered eligible, invited to participate in the study. Inclusion criteria were age > 18 to < 85 years, laboratory-proven HSVE (detection of HSV-DNA in the cerebrospinal fluid (CSF) by polymerase chain reaction (PCR) or a positive CSF HSV-specific antibody-index) and focal neurological signs for not more than 5 days prior to admission. For women of childbearing potential a negative pregnancy testing in urine was necessary [1]. Exclusion criteria are listed in Table 1. Written informed consent was obtained from each subject or from the subject’s legal representative or designee before any protocol-specific procedures were performed.

Study treatment allocation was done in a 1:1 ratio by the method of minimization considering two factors, study center and GCS (dichotomized: ≤8, > 8). Randomization was performed using the web-based software randomizer (http://​www.​randomizer.​at). The trial was conducted in double-blind fashion with the patient, the treating physician, the observer and all other site personnel as well as the monitor being unaware of the treatment assignment. Dexamethasone and placebo were provided in identically appearing vials and study kits.

All patients were treated with aciclovir intravenously at a dosage of 10 mg/kg body weight every 8 hours with an infusion time of 1 h, if patients had a normal renal function. In case of reduced creatinine clearance (< 60 ml/min) the aciclovir dosage was adapted. In addition, they received the study medication, either intravenous Dexamethasone at a dosage of 40 mg every 24 h for 4 days, or placebo that was identical in appearance to the active drug.

All concomitant medications were allowed with exception of actual or long-standing corticosteroid treatment. Gastric protection with an antacid medication was obligatory during the administration of the study medication.

Functional outcome at 6 months (±14 days) after randomization was assessed by the modified Rankin scale (mRS), a seven-point-scale (0 to 6 points) with 0 indicating normal neurological function, 6 indicating death and the points in between different grades of disability and dependence. This scale is frequently used in ischemic stroke outcome trials.

The primary endpoint was the mRS dichotomized at 0–2 vs 3–6, with 0–2 regarded as favorable outcome without dependency. A mRS of 3 to 6 was interpreted as unfavorable outcome.

Secondary endpoints included mortality after 6 and 12 months, functional outcome after 6 months measured with the Glasgow outcome scale (GOS), functional outcome after 12 months measured with mRS and GOS, quality of life as measured with the EuroQol 5D instrument (EQ-5D) after 6 and 12 months, neuropsychological testing including Mini-Mental test (MMT), Rey Complex Figure Test, Auditory Verbal Learning and Memory Test (AVLT I and II), Digit Test, Trail-Making Test, Word Fluency Test, Hospital Anxiety and Depression Scale (HADS) after 6 months, cranial MRI findings after 6 months, seizures up to day of discharge or at the latest at day 30, and after 6 and 12 months.

The statistical design was that of a group sequential design with a maximum of three stages, rejection boundaries according to O’Brien and Fleming [19] and a potential sample size adjustment for the last stage [20]. Initial assumptions for sample size calculation were a failure rate of 40% in the aciclovir monotherapy group and a failure rate of 25% for the combination therapy of aciclovir plus corticosteroids. For a specified type I error rate of 5% and equally spaced stages this treatment difference can be detected with a power 80% if a maximum of 52 patients per group and stage are included, i.e. the maximum total sample size is 312 patients (ADDPLAN; Release 3). Assuming a rate of loss to follow-up of 15% it was planned to include 62 patients per group in each stage which results in a maximum of 372 patients for the entire study. In the second interim analysis a sample size recalculation based on the pooled data of the first and second stage was planned with the option to adapt the sample size to a maximum of 450 patients for the whole study. For more details see [1].

Due to slow recruitment, the study was terminated prematurely before the first interim analysis. Therefore, the final analysis was done according to a fixed sample size design and all results will be interpreted only descriptively.

The primary analysis is based on the full analysis set (FAS) following the intention-to-treat principle. It comprises all patients treated with at least one dose of study medication (aciclovir and dexamethasone/placebo) and analyzed in the group randomized to. As sensitivity analysis, the primary endpoint was additionally analyzed as per protocol (PP). Patients were included into this PP analysis only if all inclusion and exclusion criteria were fulfilled and all medical procedures/visits have been carried out according to the protocol (dexamethasone/placebo over 4 days, if not dead: 6-month visit within ±14 days). In case of randomization through randomization list without entering into the randomization tool afterwards, the patient was excluded from the PP set. The affiliation to the PP set in disputable cases was fixed prior to database closure in a separate document.

For the FAS analysis, missing values for the primary endpoint were imputed using the last observation-carried-forward approach. Patients who died between randomization and the 6 months visit were evaluated with mRS = 6 at the 6 months visit. In addition, worst case and best case scenarios were evaluated based on the FAS meaning that in both.

treatment groups the worst (mRS > 2) or best (mRS ≤ 2) outcome was assumed in case of missing values.

Patients who died between randomization and the 12 months visit were evaluated with mRS = 6 at the 12 months visit. HADS defines two subscales, the anxiety score and the depression score. Each score was calculated adding up the values given for seven items. In case of one missing value within a subscale, this value was imputed using the mean value of the six other items. If more than one item was missing within a subscale, no imputation was done. The subscales were categorized into negative (0–7), neutral (8–10) and positive (≥11) and tabulated by treatment. For further secondary endpoints no imputation was done. For the PP set no missing values were imputed.

All secondary endpoints were analyzed based on the FAS and the PP set. In case of categorical variables, p-values correspond to the Chi²-test, in case of continuous variables, p-values are given for the nonparametric U-test. Mortality was evaluated through Kaplan Meier curves and log rank tests were performed.

Due to the small number of randomized patients and the partly high amount of missing data (e.g. for GCS Scale at baseline) no further analyses such as regression analyses were conducted.

The main safety parameter is death which was analyzed as secondary endpoint. Additionally to that, death was analyzed in the same manner based on the safety analysis set which contains all patients who were treated at least for 1 day (aciclovir dexamethasone/placebo) and patients were categorized into groups as treated.

All analyses have been done using validated SAS® Version 9.1 run on a Linux system. Stacked bar graphs for mRS have been created using R Version 3.5.1.

One thousand one hundred twenty-nine patients were screened by 27 centers between November 2007 and December 2012. After exclusion of 1087 patients, finally 41 patients were randomized to dexamethasone plus aciclovir (N = 21) (dexamethasone group) or to placebo plus aciclovir) (N = 20) (placebo group) (Fig. 1). A CONSORT flow chart is presented in Fig. 1. We had to exclude 3 patients from FAS after randomization because of informed consent withdrawal (N = 2) and screening failure (N = 1). Both treatment groups were characterized using descriptive methods based on the FAS and PP set. In this paper we describe the FAS group, for further details of the PP set we refer to the online published data set (see electronical Additional file 1). Table 2 summarizes the demographics for the patients in the FAS which consist of 19 patients in each group. No severe group differences were present at baseline. Patients in the Dexamethasone group were on average 3 years older and had more often neurological signs at onset (78.9% vs. 63.2%). The pre-encephalitis mRS was also minimally higher in the Dexamethasone group (mRS 0: Placebo n = 17, Dexamethasone n = 14; mRS ≥1: Placebo n = 2, Dexamethasone n = 5).

No differences in the dichotomized mRS outcome scale were found between the Placebo and Dexamethasone groups. In both groups, 12 patients had a mRS ≤2, seven patients had > 2 (p = 1.0), respectively. This was confirmed in the sensitivity analyses. Table 3 and Fig. 2 show the mRS at 6-month visit, for both study groups, placebo and dexamethasone.

Results for secondary endpoints are given in Tables 3 and 4. Overall, 3 patients died within 12 months. In the dexamethasone group, one patient died within 6 months. In the placebo group 2 patients died within 12 months. No differences in the mortality rates could be detected (Table 3). Neither after six (p = 0.266), nor after 12 months (p = 0.681). Similarly, no relevant differences in the functional outcome after six and 12 months were found. The GOS after 6 months (p = 0.46) and 12 months (p = 0.773) turned out to be similar in both groups. Furthermore, the quality of life was not different between the groups after 6 months [EQ-5D-Index (p = 0.243) and EQ-5D-VAS (p = 0.184)] and 12 months [EQ-5D-Index (p = 0.812) and EQ-5D-VAS (p = 0.296)] as well as the modified ranking scale after 12 months (p = 0.897). No relevant differences were found in all neuropsychological tests after 6 months (see Table 4). Seizure activity was similar in both groups up to day of discharge or at the latest at day 30 (p = 0.632), and after 6 months (p = 0.713) and 12 months (p = 0.271) as well.