Articleshttp://neurology.thelancet.com Vol 5 December 2006 1029Eﬀect of nimodipine on outcome in patients with traumaticsubarachnoid haemorrhage: a systematic reviewMervyn D IVergouwen, MarinusVermeulen,Yvo BW E M RoosSummaryBackground Despite several randomised controlled trials, there is still much debate whether nimodipine improvesoutcome in patients with traumatic subarachnoid haemorrhage. A 2003 Cochrane review reported improved outcomewith nimodipine in these patients; however, because the results of Head Injury Trial (HIT) 4 were only partly presentedthere is still discussion whether patients with traumatic subarachnoid haemorrhage should be treated with this drug.Here, we present data from all head-injury trials, including previously unpublished results from HIT 4.Methods We systematically searched PubMed and EMBASE databases using the following combinations of variables:“nimodipine” or “calcium antagonist” with “traumatic subarachnoid haemorrhage”, “head injury”, “head trauma”,“brain injury”, or “brain trauma”. Bayer AG and all principal investigators or corresponding authors of the identiﬁedstudies were contacted for additional information.Findings Five manuscripts were identiﬁed, describing the results of four trials. We obtained additional data from HIT1, 2, and 4. In total, 1074 patients with traumatic subarachnoid haemorrhage were included. The occurrence of pooroutcome was similar in patients treated with nimodipine (39%) and those treated with placebo (40%); odds ratio was0·88 (95% CI 0·51–1·54). Mortality rates did not diﬀer between nimodipine (26%) and placebo (27%) treated patients(odds ratio 0·95; 95% CI 0·71–1·26).Interpretation Our results do not lend support to the ﬁnding of a beneﬁcial eﬀect of nimodipine on outcome inpatients with traumatic subarachnoid haemorrhage as reported in an earlier Cochrane review.IntroductionIn patients with aneurysmal subarachnoid haemorrhage,nimodipine has been shown to be eﬀective in theprevention of ischaemic complications after haemorrhageresulting in improved outcome.1Since nimodipine isbelieved to have neuroprotective properties, the drug wassuggested to be of potential beneﬁt in patients withtraumatic head injury.The ﬁrst report on nimodipine treatment in patientswith severe head injury dates from 1984.2The ﬁrstrandomised controlled trials investigating the eﬀect ofnimodipine in head injury the Head Injury Trials (HIT) 1and 2 were published in the early 1990s.3,4In the overallanalysis of these studies, no reduction in poor outcomewas shown. However, in a subgroup analysis of HIT 2, apossible beneﬁcial eﬀect was noted in patients withtraumatic subarachnoid haemorrhage. Because of theseﬁndings, HIT 3 was initiated in which only head-injurypatients with traumatic subarachnoid haemorrhage wereincluded.5The HIT 3 study showed a small but signiﬁcantbeneﬁcial eﬀect in nimodipine-treated patients. To conﬁrmthis favourable eﬀect in a much larger group of patients,HIT 4 was started. Unexpectedly, HIT 4 showed asigniﬁcant increase in poor outcome in nimodipine-treatedpatients. Possibly because of this disappointing result, theresults of HIT 4 have never been published in detail, butsome of the results were presented at a conference.In 2003, a Cochrane review on the eﬀects of calciumantagonists on outcome in patients with head injury waspublished, which concluded that there is no beneﬁcialeﬀect of nimodipine on outcome in head-injury patients.6However, in the subgroup of patients with traumaticsubarachnoid haemorrhage nimodipine was shown toimprove outcome.Because the results of HIT 4 were only partly presented,there is still much debate as to whether patients withtraumatic subarachnoid haemorrhage should be treatedwith nimodipine. Here we present for the ﬁrst time thedata of all head-injury trials, including previouslyunpublished results from the HIT 4 trial.MethodsSearch strategy and selection criteria of studiesWe systematically searched the electronic PubMed andEMBASE databases up to 2006 (week 10) for the followingcombinations of variables: “nimodipine” or “calciumantagonist” with “traumatic subarachnoid haemorrhage”,“head injury”, “head trauma”, “brain injury”, or “braintrauma”. Randomised controlled studies from the English,German, and French published work were included.Animal studies were excluded. Bayer AG and all principalinvestigators or corresponding authors of the identiﬁedstudies were contacted for additional information. Allrandomised controlled trials investigating the eﬀect ofthe calcium antagonist nimodipine in patients with headinjury were included, irrespective of dose, route ofadministration, and duration of treatment.The Glasgow outcome scale was measured 6 monthsafter head injury.7Primary outcome was a poor outcome,deﬁned as death, vegetative state, or severe disability.Lancet Neurol 2006; 5: 1029–32Published OnlineOctober 17, 2006DOI:10.1016/S1474-4422(06)70582-8See Reﬂection and Reactionpage 993Department of Neurology,Academic Medical Centre,Meibergdreef 9, Amsterdam,Netherlands(M D IVergouwen MD,MVermeulen MD,Y BW E M Roos MD)Correspondence to:Mervyn D IVergouwenm.email@example.com
Articles1030 http://neurology.thelancet.com Vol 5 December 2006Favourable outcome was deﬁned as moderate disabilityand good recovery. Secondary outcome was mortality.Methodological qualityOne of the investigators (MDIV) assessed themethodological quality of the studies by means of theJadad scale.8This scale is a validated 5 point scale inwhich the following items are assessed: randomisation(0–2 points), double-blinding (0–2 points), andwithdrawals and drop-outs after randomisation(0–1 point). Higher values on the Jadad scale representhigher quality studies. All trials scoring 1 or 2 points forrandomisation were included in the analysis.Statistical analysisData were analysed according to the intention-to-treatprinciple and processed in Review Manager 4.2 assupplied by the Cochrane Collaboration. Odds ratios forpoor outcome were calculated with a random-eﬀectsmodel because of possible heterogeneity between studies.Mortality odds ratios were calculated with a χ² analysisbecause only crude mortality rates were available.Statistical uncertainty was expressed by 95% CIs.Role of the funding sourceThere was no external funding source or sponsor involvedin the study design, data collection, data analysis, datainterpretation, or writing of the report. The correspondingauthor had full access to all the data in the study and hadresponsibility for the ﬁnal decision to submit forpublication.ResultsBy searching the electronic databases of PubMed andEMBASE ﬁve manuscripts were identiﬁed, describing theresults of four trials.3–5,9,10The principal investigators ofHIT 1 and 2 were contacted for additional data on outcomeand access was granted to both original datasets. For HIT3, only published data were available. Bayer AG wascontacted for the data of HIT 4, but they advised us tocontact the principal investigator of HIT 4. The principalinvestigator of HIT 4 granted us permission to analyseand publish the results of HIT 4, which were described ina table of the pooled outcome scores of all patients of allfour HITs. Finally, the corresponding author of the studyby Pillai and co-workers10was contacted for additionaldata. However, the original data were unavailable.In HIT 1, 352 patients with head injury were included.3The presence of traumatic subarachnoid haemorrhagewas not registered during the study, but was assessedafterwards by a panel of four observers.9CT scans couldbe retrieved for 257 patients, of whom traumaticsubarachnoid haemorrhage was present in 71 (36 placebopatients, 35 nimodipine patients).In HIT 2, 852 patients with head injury were enrolled.4The presence of traumatic subarachnoid haemorrhagewas assessed by the investigators and afterwards re-assessed by a review committee. Although theparticipating investigators identiﬁed traumaticsubarachnoid haemorrhage in 149 placebo patients andin 119 nimodipine patients, the review committeeidentiﬁed traumatic subarachnoid haemorrhage in145 placebo patients and in 123 nimodipine patients.HIT 1 (n=71) HIT 2 (n=268) HIT 3 (n=121) HIT 4 (n=577) Pillai, et al (n=37)Placebo Nimodipine Placebo Nimodipine Placebo Nimodipine Placebo Nimodipine Placebo NimodipinePoor outcome 25 (69%) 26 (74%) 83(56%) 53 (45%) 28 (46%) 15 (25%) 75 (26%) 98 (34%) 11 (61%) 13 (68%)Favourable outcome 11 (31%) 9 (26%) 66(44%) 66 (55%) 33 (54%) 45 (75%) 212 (74%) 192 (66%) 7 (39%) 6 (32%)Total 36 35 149 119 61 60 287 290 18 19Data are number (%).Table 1: Glasgow outcome scale of head-injury patients with traumatic subarachnoid haemorrhageFavours treatment Favours placebo10·50·20·1 52 10OR (random) 95% CI Weight % OR (random) 95% CIControl n/NTreatment n/NStudy or subcategoryHIT1HIT2HIT3HIT4PillaiTotal (95% CI)26/3553/11915/6098/29013/1952323/3683/14928/6175/28711/1855115·1125·7719·8028·3011·02100·001·27 [0·45–3·59]0·64 [0·39–1·04]0·39 [0·18–0·85]1·44 [1·01–2·06]1·38 [0·36–5·34]0·88 [0·51–1·54]Total events: 205 (treatment), 222 (control)Test for heterogeneity: χ²=13·35, df=4 (p=0·010), I²=70·0%Test for overall eﬀect: Z=0·44 (p=0·66)Comparison: Nimodipine vs placeboOutcome: Poor outcomeFigure: Comparison of nimodipine versus placebo for the odds of poor outcome deﬁned as death, vegetative state, or dependency on the Glasgow outcomescale at 6 months after head injury.
Articleshttp://neurology.thelancet.com Vol 5 December 2006 1031In HIT 3 only head-injury patients with traumaticsubarachnoid haemorrhage were enrolled.5In total123 patients were randomised. A review committee couldnot conﬁrm the presence of traumatic subarachnoidhaemorrhage in 26 patients. However, these cases wereincluded for the intention-to-treat analysis. Of these123 patients, two patients were lost to follow-up at6 months.In HIT 4, 592 patients were enrolled, of whom 577 werevalid for the intention-to-treat analysis (287 placebopatients, 290 nimodipine patients; unpublished data).Why 15 patients were excluded for the intention-to-treatanalysis is not known. A CT scan review committeeidentiﬁed traumatic subarachnoid haemorrhage in525 patients (259 placebo patients, 266 nimodipinepatients). The HIT 4 data, which were released earlier ata conference where data of the CT scan review committeewere shown, were used for the Cochrane review. For thepresent systematic review we used the data of the HIT 4intention-to-treat analysis, thus including 577 patientsinstead of 525 patients.In the study by Pillai and co-workers,1097 patients withsevere diﬀuse head injury were enrolled, of whom 39 hadtraumatic subarachnoid haemorrhage on CT scan. Ofthese 39 patients, follow-up was missing in two. ThereforeGlasgow outcome scores at 6 months were available for37 patients (18 placebo patients, 19 nimodipine patients).In total, 1074 head-injury patients with traumaticsubarachnoid haemorrhage were included for thissystematic review. In the primary analysis, the eﬀect ofnimodipine on the incidence of poor outcome was studied.For this analysis all ﬁve studies supplied data on theincidence of poor outcome as well as unpublished data ofHIT 4.4,5,9,10Table 1 and the ﬁgure show the dichotomisedGlasgow outcome scores. The occurrence of poor outcomewas similar in nimodipine and placebo-treated patients(39% in the nimodipine group and 40% in the placebogroup: odds ratio 0·88 [95% CI 0·51–1·54]).In the secondary analysis, we studied the eﬀect ofnimodipine on the incidence of mortality. No data onmortality were available in the study of Pillai and co-workers.10Pooled mortality rates of all four HIT studiesincluding all patients valid for intention-to-treat analysisare listed in table 2. Thus, for the secondary analysis, datafor 1037 patients were available. The mortality rates weresimilar in nimodipine and placebo-treated patients (26%in the nimodipine group and 27% in the placebo group:odds ratio 0·95 [95% CI 0·71–1·26]).The methodological quality of the included studies wasassessed with the use of the Jadad scale (table 3). The CTscans of HIT 1 were assessed afterwards by a panel ofobservers who were unaware of drug allocation andoutcome (G Teasdale, personal communication). TheJadad score for HIT 4 was based on limited data. Themedian score was 3 (range 2–5). As all trials scored 1 or 2points for randomisation, no studies were excluded forthis systematic review.DiscussionThe occurrence of poor outcome and mortality rates didnot diﬀer between patients treated with nimodipine andthose treated with placebo. Our results contrast withthose of the Cochrane review published in 2003, whichshowed that nimodipine signiﬁcantly reduced pooroutcome and mortality in patients with traumaticsubarachnoid haemorrhage. The present analysis diﬀersin three aspects from the Cochrane analysis. First, werestudied the original data of HIT 1 so that mortality ratesof HIT 1 could be included in the present analysis.Second, we managed to obtain additional data from theHIT 4 study and included these data in the analysis of allpatients with traumatic subarachnoid haemorrhage.Interestingly, in the Cochrane review, the HIT 4 datawere only included in the overall analysis of patients withtraumatic head injury and not in the subgroup analysisof patients with traumatic subarachnoid haemorrhage.This is remarkable since HIT 4 only included patientswith traumatic subarachnoid haemorrhage. Therefore,the Cochrane review was based on less than half of thepatients of our review. We also included the results of anadditional study.By contrast with patients with traumatic subarachnoidhaemorrhage, nimodipine has been proven to improveoutcome in patients with aneurysmal subarachnoidhaemorrhage.1Several factors can explain the diﬀerencein eﬀect of nimodipine on outcome in patients withaneurysmal compared with traumatic subarachnoidhaemorrhage. Aneurysmal and traumatic subarachnoidhaemorrhages are diﬀerent entities. In patients withaneurysmal subarachnoid haemorrhage, secondarycerebral ischaemia can occur in the weeks after thehaemorrhage. Patients who have secondary ischaemiahave an increased risk of poor outcome. This secondaryRandomisation(0–2 points)Double-blinding(0–2 points)Withdrawals/drop-outs (1 point)Total (0–5 points)HIT 1 2 2 1 5HIT 2 2 2 0 4HIT 3 1 2 0 3HIT 4* 1 1 0 2Pillai 2 1 0 3*Based on limited data.Table 3: Methodological quality of the included studies using the Jadad scalePlacebo NimodipineSurvival 390 (73%) 374 (74%)Death 143 (27%) 130 (26%)Total 533 504Data are number (%).Table 2: Pooled mortality rates for all patients in all four HIT studiesvalid for intention-to-treat analysis
Articles1032 http://neurology.thelancet.com Vol 5 December 2006cerebral ischaemia is less common in traumaticsubarachnoid haemorrhage than in aneurysmalsubarachnoid haemorrhage.11,12Nimodipine is probablyonly eﬀective for secondary cerebral ischaemia and mighttherefore be eﬀective only in patients with aneurysmalsubarachnoid haemorrhage. For a long time, the beneﬁcialeﬀects of nimodipine in subarachnoid haemorrhage havebeen attributed to a beneﬁcial eﬀect on vasospasm.However, nimodipine exerts a ﬁbrinolytic eﬀect in patientswith aneurysmal subarachnoid haemorrhage, an eﬀectwhich was also observed in other calcium antagonistsfrom the dihydropyridine group.13Therefore, nimodipineexerts a beneﬁcial proﬁbrinolytic eﬀect in patients withaneurysmal subarachnoid haemorrhage and this could bethe mechanism by which the incidence of cerebralischaemia is decreased. In patients with traumaticsubarachnoid haemorrhage the proﬁbrinolytic eﬀectmight not be beneﬁcial, but even deleterious by inducingcerebral and systemic haemorrhages.The Jadad scale was used to assess the methodologicalquality of the included papers. The Jadad score of HIT 4was based on a limited number of data. With the availabledata of HIT 4, the Jadad score was 2. Therefore the largeststudy in this systematic review probably had the lowestquality score. Often, lower quality trials tend to have morepositive results, but this was not found in this analysis.In this systematic review including 1074 patients withtraumatic subarachnoid haemorrhage, we could notconﬁrm the beneﬁcial eﬀects of nimodipine shown in aprevious review, which included 460 patients with thiscondition.ContributorsMDIV, MV, and YBWEMR developed the study protocol. MDIV andYBWEMR managed the data. MDIV did the statistical analysis. MDIV,MV, and YBWEMR prepared the manuscript.Conﬂicts of interestWe have no conﬂicts of interest.AcknowledgmentsWe thank Prof Sir G Teasdale and Prof R Braakman, principal investigatorsof HIT 1 and 2, respectively, for permission to study the original data ofthese trials. We also thank Prof V J Farrell, principal investigator of HIT 4,for permission to use and publish parts of the results of HIT 4.References1 Rinkel GJE, Feigin VL, Algra A, van den Bergh WM, Vermeulen M,van Gijn J. Calcium antagonists for aneurysmal subarachnoidhaemorrhage. Cochrane Database Syst Rev 2005; 1: CD000277.2 Kostron H, Twerdy K, Stampﬂ G, Mohsenipour I, Fischer J,Grunert V. Treatment of the traumatic cerebral vasospasm with thecalcium channel blocker nimodipine: a preliminary report.Neurol Res 1984; 6: 29–32.3 Bailey I, Bell A, Gray J, et al. A trial of the eﬀect of nimodipine onoutcome after head injury. Acta Neurochir (Wien) 1991; 110: 97–105.4 The European Study Group on Nimodipine in Severe Head Injury.A multicenter trial of the eﬃcacy of nimodipine on outcome aftersevere head injury. J Neurosurg 1994; 80: 797–804.5 Harders A, Kakarieka A, Braakman R, the German tSAH StudyGroup. Traumatic subarachnoid hemorrhage and its treatment withnimodipine. J Neurosurg 1996; 85: 82–89.6 Langham J, Goldfrad C, Teasdale G, Shaw D, Rowan K. Calciumchannel blockers for acute traumatic brain injury.Cochrane Database Syst Rev 2003; 4: CD000565.7 Jennett B, Bond M. Assessment of outcome after severe braindamage. Lancet 1975; 305: 480–84.8 Jadad AR, Moore RA, Carroll D, et al. Assessing the quality ofreports of randomized clinical trials: is blinding necessary?Control Clin Trials 1996; 17: 1–12.9 Murray GD, Teasdale GM, Schmitz H. Nimodipine in traumaticsubarachnoid hemorrhage: a re-analysis of the HIT I and HIT IItrials. Acta Neurochir (Wien) 1996; 138: 1163–67.10 Pillai SV, Kolluri VR, Mohanty A, Chandramouli BA. Evaluation ofnimodipine in the treatment of severe diﬀuse head injury: a double-blind placebo-controlled trial. Neurol India 2003; 51: 361–63.11 Fukuda T, Hasue M, Ito H. Does traumatic subarachnoidhemorrhage caused by diﬀuse brain injury cause delayed ischemicbrain damage? Comparison with subarachnoid hemorrhage causedby ruptured intracranial aneurysms. Neurosurgery 1998; 43: 1040–49.12 Steiger HJ, Aaslid R, Stooss R, Seiler RW. Transcranial Dopplermonitoring in head injury: relations between type of injury, ﬂowvelocities, vasoreactivity, and outcome. Neurosurgery 1994; 34: 79–85.13 Roos YBWEM, Levi M, Carroll TA, Beenen LFM, Vermeulen M.Nimodipine increases ﬁbrinolytic activity in patients withaneurysmal subarachnoid hemorrhage. Stroke 2001; 32: 1860–62.