Subjects
Patients between 18 and 80 years of age with
definite or
probable VP were eligible, as defined by the following modified diagnostic criteria [
6]: short spells of vertigo lasting from seconds to a few minutes; disturbance of gait or postural instability during the attacks; attacks occur at rest or can be provoked by hyperventilation, certain head positions or changes in head position; no central ocular motor disorder; treatment response to antiepileptic drugs (not applicable at presentation); additional criteria: reduced hearing or unilateral tinnitus during the attacks or permanently; increase of detectable vestibular/cochlear deficit during the course of the disease.
Definite VP was identified when four diagnostic criteria or three plus one additional criterion were fulfilled;
probable VP was identified when three diagnostic criteria or two plus one additional criterion were fulfilled, and
possible VP was identified in patients who fulfilled fewer criteria.
Exclusion criteria included additional vestibular or ocular motor disorders such as benign paroxysmal positional vertigo or vestibular migraine at the time of initial presentation, and other neurologic disorders affecting the CNS. Pregnant or nursing women were not included.
Subjects were recruited from the outpatient Dizziness Unit of the Neurological Clinic of Munich University Hospital. Eligible subjects received written and oral information about the study, including possible side effects of OXA. Written informed consent was obtained from all subjects prior to inclusion.
The study was open for recruitment from August 2005 until December 2011. Last patient last visit (LPLV) was on November 10th 2011. The study was stopped because of the slow recruitment and the high number of drop-outs, since the minimum number of subjects stated in the protocol (at least 40 patients) was reached.
Participants were randomized to receive OXA in the first, and matching placebo in the second treatment period or vice versa. Each treatment period was 3 months long with a 1-month wash-out period in between. Study medication for the first period was dispensed at inclusion; for the second period, it was dispensed at the 2nd visit, scheduled during the washout phase. The 3rd and last visit was scheduled after the 2nd treatment period. Study medication was administered in an ascending therapeutic scheme (first week: one 300 mg tablet per day, second week: 300 mg b.i.d., third and subsequent weeks: 300 mg t.i.d.). Patients received written and oral instructions on how to take the medication and were asked to record the number of pills taken in the vertigo diary.
Patients’ self-report of attacks
Participants were instructed to document vertigo attacks, duration, intensity (1–10, one being the lowest), and accompanying symptoms in structured diary forms collected at each visit. They were asked to also enter days without symptoms by just filling in the date.
Monitoring of side effects
Side effects were assessed by handing out standardized forms to the patients, listing the known side effects of OXA (from the Tegretal® SmPC, September 2004) ordered by organ system, along with instructions and contact information. As the electrolyte blood levels (especially sodium), blood count and liver function might change under therapy with sodium channel blockers, they were examined weekly for the first month and monthly for the rest of each treatment period by the patients’ general practitioner. At the follow-up visits, the forms were collected, and subjects were also asked if they had had any (including unexpected) side effects or no side effects at all. In the event of severe side effects or any kind of rash, patients were instructed to contact the investigators immediately.
In the protocol, attack frequency, duration, intensity, triggers and concomitant symptoms, as well as adverse reactions were listed as main outcomes. With regard to data quality of the patient-reported outcomes (see Data management), and to multiple testing, it was decided before unblinding the treatment allocation to reduce outcomes and to define the endpoints as follows: primary endpoint is the number of days with one or more attacks among the observed days (ATD). Secondary endpoints are: number of attacks during the observed days (NAT); median (for each day) duration of attacks. Adverse events are reported descriptively. The endpoints are compared for OXA versus placebo treatment period (including the time under ascending dose).
The sample size was set to at least 40 patients according to the protocol, the significance level was set to alpha = 5%.
Randomisation, concealment, and blinding
A simple randomisation with a computer-generated list of random numbers, which was generated in the pharmacy, was used to allocate the participants.
OXA (Tegretal®, Novartis Pharma GmbH, Nuremberg) 300 mg tablets were over-encapsulated with lactose monohydrate as filling material at the pharmacy of the Munich University Hospital. Identical capsules with the same filling material were used as placebo. OXA- and placebo-capsules were packed in bottles labeled with the random numbers and the order in which they should be taken (e.g., 14 A and 14 B together making up a kit for one participant), according to the randomisation list.
Participants were enrolled and examined by physicians from the neurologic outpatient clinic, mostly by one of the authors (KH). The investigational medical product was kept in numbered kits, which were assigned to the participants in ascending order.
Investigators and patients were blinded to the treatment allocation sequence.
Data management
For analysis, a relational HSQL-Database (version 1.8, front end LibreOffice version 4) was set up and data from (paper) records, vertigo diaries, and adverse event (AE) collection forms were entered. Before data entry, written rules were set up and trained, to ensure consistent handling. This was particularly important for the diaries, which were not always completed by the participants as expected. To summarize, the database form allowed entry of quantitative (preferred where available) or qualitative measures of attacks per day. Duration was entered as the median duration of attacks on each day (in case different durations were recorded by the patient). Since patients did not always document attack-free days, documentation periods (several per patient possible, independent of treatment period) were classified as “attack-free days documented” or “attack-free days not documented”, and entered in the database. This was done to obtain information on data quality to support the blinded decision about confirmatory and exploratory endpoints, and to facilitate ancillary analyses. After data entry, verification against the source data was performed, and queries were generated by a second person, which were recorded in the database along with the answers.
The standardized side effect forms contained a field to record the onset date and duration (rarely documented) of the events, and date fields to specify the time period covered (start and end date). The latter allowed the event to be assigned to the treatment phases according to the following procedure: Forms with an end date not earlier than the start of the first treatment phase and before the start of the second treatment period were assigned to the first treatment period. Forms with an end date not earlier than the start of the second treatment period and at least 14 days after the end of the first treatment period were assigned to the second treatment period. Forms with an end date not earlier than the start of the second, and not more than 14 days after the end of the first treatment period were assigned to the first and/or second treatment period, i.e., OXA and/or placebo (see Table
2).
Table 1
Due to the high drop-out rate, this table gives population descriptives for the periods analyzed
Age 20–73 years, median 52 years, mean 50 years | Age 20–73 years, median 56 years, mean 52 years |
Gender: 6 females, 10 males | Gender: 5 females, 9 males |
Table 2
Frequency of adverse events categorized by organ system
General | Fatigue, feeling weak | 5 | 3 | 2 |
Swelling of eye lids or other parts of the body | 2 | 1 | – |
Immune system | Symptoms of hypersensitivity involving more than one symptom or organ | 2 | – | – |
Skin | Rash | 1 | – | – |
CNS | Vertigo, dizziness | 6 | 4 | 1 |
Headache | 6 | 4 | 3 |
Nausea | 5 | 1 | 2 |
Disturbance of concentration | 1 | 2 | – |
Listlessness | 1 | – | 1 |
Restlessness | 1 | 2 | – |
Emotional instability | 2 | 1 | – |
Tremor | 2 | – | – |
Coordination impaired | 1 | – | – |
Nystagmus | – | 1 | – |
Visual disturbances | | 3 | – | 1 |
Cardiac disorders | Rate and rhythm disturbances | – | – | 1 |
Gastrointestinal disorders | Abdominal pain | 2 | 1 | 1 |
Obstipation | 1 | 1 | – |
Diarrhea | 3 | 1 | – |
Hepatobiliary disorders | | 1 | 1 | – |
Metabolism disorders | Including sodium imbalance | 1 | 1 | 1 |
Statistical analyses
Statistical analysis and figure design was performed using R (version 3.3.2,
www.r-project.org), except for modelling results which were obtained using PROC GENMOD in SAS 9.4 University Edition, SAS Institute, Cary USA. To facilitate reproducible analyses StatWeave (version 0.91_03,
www.stat.uiowa.edu/~rlenth/StatWeave), a literate programming tool was used.
The classical approach for cross-over trials is to take the difference in the outcome parameter between active OXA treatment and placebo to test efficacy. However, this requires data for both treatment periods. Due to the high number of probably informative missings (resulting from drop-outs, periods of insufficiently filled in diaries), excluding such patients might introduce bias. Therefore, we decided before unblinding treatment allocation, to employ analyses with treatment (OXA vs. placebo) as grouping variable capable of including data from one or two treatment periods. Furthermore, this approach avoids reducing diary data by averaging to one number per patient and treatment period, which would raise weighting issues.
For the primary endpoint ATD and for the secondary endpoint NAT a Poisson regression model was used. The models were estimated applying the GEE approach (generalized estimating equations) [
17]. GEE use quasi-likelihood to fit the model, account for clustered data (data within patients are not independent), and estimate a scaling parameter adjusting for overdispersion which can occur in Poisson regression. The results of median attack duration are shown descriptively.