Study design and study population
The trial was designed as a randomized, double-blind, double-dummy, parallel-group, single-centre trial at the Department of Otorhinolaryngology of the Královské Vinohrady University Hospital, Prague, Czech Republic. It was registered in the public clinical trials register ISRCTN under number 68772788.
We enrolled male and female patients aged ≥ 30 years with unilateral or bilateral chronic or subchronic tinnitus of at least 3 months’ duration. Subjects were eligible for participation if their tinnitus was maskable (by noise masking), the degree of annoyance by tinnitus was rated at least 3 on an 11-Point Box Scale (type of numeric analogue scale) at screening and baseline visits, they scored at least 5 on the abridged Tinnitus Questionnaire (Mini-TQ)  at baseline and had given informed consent.
Patients were excluded from the study if they had acute or chronic otitis media, vestibular neuritis or drug-induced tinnitus, if they were taking any other treatment for tinnitus, if they had severe cardiovascular, renal or hepatic disorders, malignant diseases, insulin-dependent diabetes mellitus or gastro-intestinal disorders leading to impaired drug absorption. Any drugs taken to treat tinnitus had to be discontinued at least 8 weeks (Ginkgo extracts at least 12 weeks before baseline. Patients who needed drugs that could possibly interfere with the effects of the investigational treatments (e.g. due to agonistic or antagonistic action on common pharmacodynamics pathways), who were taking anticoagulants or who were known to be allergic to the investigational drugs were also excluded. Female patients of childbearing age were only included under safe contraception.
Randomization and interventions
The random allocation sequence was generated by the sponsor using a validated computer program matching drug numbers to treatments in a 1:1 ratio. The randomisation sequence was concealed by using identical labels and packages for both treatments with ascending drug numbers. The list matching drug numbers with treatments was unavailable to persons involved in conducting the study. Double-blinding was achieved by the double-dummy technique, i.e. all patients received the same number of tablets, either EGb 761® and pentoxifylline-like placebo or pentoxifylline and EGb 761®-like placebo. Tablets containing active drug and the corresponding placebo tablets were indistinguishable in texture, colour, shape and size.
For the duration of 12 weeks, the subjects randomized to receive EGb 761® took one film-coated tablet of 120 mg EGb 761® together with one pentoxifylline-like placebo tablet twice a day; those randomized to pentoxifylline took one extended-release tablet of 600 mg pentoxifylline together with one EGb 761®-like placebo tablet twice a day. EGb 761®Footnote 1 is a dry extract from Ginkgo biloba leaves (35–67:1), extraction solvent: acetone 60% (w/w). The extract is adjusted to 22.0–27.0% ginkgo flavonoids calculated as ginkgo flavone glycosides and 5.0–7.0% terpene lactones consisting of 2.8–3.4% ginkgolides A, B, C and 2.6–3.2% bilobalide and contains less than 5 ppm ginkgolic acids.
The therapeutic effects of EGb 761® and pentoxifylline were assessed using tinnitus-related rating scales as well as assessments of psychological symptoms and functioning.
Two separate 11-Point Box Scales for tinnitus loudness (extending from 0 = no tinnitus at all to 10 = extremely loud tinnitus) and annoyance by tinnitus (extending from 0 = not annoying at all to 10 = unbearably annoying) were filled in every day. The evaluation of the scales was based on mean weekly values per subject. The Mini-TQ  is an abridged, 12-item version of the Tinnitus Questionnaire (TQ) . It was designed to reflect tinnitus-related psychological distress and to investigate the dimensions of the complaint about tinnitus such as subjective perception, coping attitudes and beliefs about tinnitus.
The Hospital Anxiety and Depression Scale (HADS) [12, 13] was designed to assess the presence and severity of mild, even sub-syndromal degrees of anxiety and depression. Since no somatic items are included, the scale is feasible to measure depression in somatic illnesses. The Sheehan Disability Scale (SDS)  is a brief 3-item self-rating tool, designed to measure the extent to which three major sectors in the patient’s life (work/school, social life, family life) are impaired by panic, anxiety, phobic or depressive symptoms.
Safety and tolerability of both investigational products were assessed by physical examination, otological examination, ECG measurements, laboratory tests and vital signs measurements.
For each of the effectiveness outcomes the EGb 761® group was compared to the pentoxifylline group using descriptive data analysis methods. The comparison of the treatment groups with respect to the 11-Point Box Scales and the Mini-TQ total score was performed using an analysis of covariance (ANCOVA) with treatment as the factor and the baseline value (in the case of 11-Point Box Scales: average value of the week until baseline) of the respective effectiveness variable as a covariate. The confidence intervals of the differences in the least square means (LS means) were computed to compare the effectiveness of the treatments. For the HADS and the SDS changes from baseline were compared between the EGb 761® group and the pentoxifylline group by the Wilcoxon rank sum test. Changes for ordinal variables over time were modelled with generalized estimating equations (GEEs) for ordinal responses.
Descriptive statistics were computed to describe the empirical distributions; 95% confidence intervals were calculated within the treatment groups and between the pentoxifylline and the EGb 761® group. Since no confirmatory hypotheses were formulated, an adjustment of the type-one error rate was not performed and no formal sample size calculation was done. All p values presented are two-sided and should be interpreted in an exploratory sense. The presented results are based on the full analysis set (FAS) using the last observation carried forward method (LOCF) to replace missing values. Furthermore, all analyses were also performed based on observed cases (OC). The results based on the OC were generally very consistent with those for the FAS and are therefore not presented.