R.Perry et al Phytomedicine 19 (2012)825-835 833 comparative trial by Woelk and Schlafke(2010)found thatlavender Furthermore,the majority of the included trials did not pro- was as effective as the anxiolytic drug lorazepam.Of the remaining vide reasons by group for dropouts.This is a frequent phenomenon studies,the findings of one RCT(Muzzarelli et al.2006)could not be in research of this type and there is a common misconception evaluated because it did not provide between-group comparisons that "natural"means safe (Ernst 2007a).Researchers investigating of change in outcome from baseline.The remaining six trials did not botanical products should comply with the CONSORT guidelines demonstrate significant differences between groups.However,all for the reporting of herbal products (Gagnier et al.2006).None results should be considered in the context of their methodological of the included trials of oral lavender (Bradley et al.2009;Kasper limitations. et al.2010:Woelk and Schlafke 2010)provided information that The results suggest limited specific effects of lavender inhalation met more than 9 of the 15 CONSORT statement criteria regarding and massage on anxiety measures.Although reductions in anxiety the extraction and preparation of herbs.Future trials should report measures were observed,methodological issues limit the extent to such information to allow adequate between-trial comparison of which firm conclusions can be drawn.On the other hand,promising the results to be made. evidence regarding ingested lavender preparations was found. The reporting of many of the trials was generally poor;all were Safety described as randomised but only five gave details on randomisa- tion process(Braden et al.2009;Kasper et al.2010;Morris 2002; The current data suggests that lavender is relatively safe, Muzzarelli et al.2006:Woelk and Schlafke 2010)and two on treat- although the gastrointestinal effects of oral lavender being taken in ment allocation concealment(Dunn et al.1995;Soden et al.2004) the long term haven't been assessed.Of a total of 15 randomised tri- Five trials(Braden et al.2009:Dunn et al.1995:Kasper et al.2010: als only 5 trials(Dunn et al.1995:Kasper et al.2010:Kritsidima etal. Kritsidima et al.2010:Soden et al.2004)included power calcu- 2010:Kutlu et al.2008;Woelk and Schlafke 2010)reported infor- lations,although one was underpowered at baseline (Soden et al mation on adverse events.One(Dunn et al.1995)had AEs that were 2004)and two became marginally underpowered following par- not related to the therapy.The remaining ten did not mention AEs at ticipant drop out(Dunn et al.1995;Kasper et al.2010).Most trials all.It is good practice for all trials to report on adverse events even if reported that participants were similar on prognostic indicators, unrelated to the therapy.In these RCTs,most of the AEs mentioned although this was assumed rather than tested in the majority of tri- were minor ailments.Kutlu et al.(2008)had "some"participants als that used healthy participants.Three trials(Braden et al.2009; reporting discomfort from the odour thus were excluded.In the oral Sgoutas-Emch et al.2001:Soden et al.2004)reported differences in lavender trials,Kasper et al.(2010)reported slightly more adverse anxiety levels at baseline.In addition,the trials differed greatly in events from the placebo group than the lavender group;the most terms of the conditions treated,and independent replications are frequently reported AEs were related to infections and infestations. missing for most. followed by gastrointestinal disorders and nervous system disor- Five trials were described as double-blind (Bradley et al.2009: ders.Woelk and Schlafke (2010)reported slightly more adverse Howard and Hughes 2008;Kasper et al.2010;Soden et al.2004; events in the lavender group than the lorazepam group but again Woelk and Schlafke 2010).although in two (Bradley et al.2009: none were described as serious.Since lavender oil showed no seda- Woelk and Schlafke 2010),blinding details were not clearly pro tive effects and has no potential for drug abuse,lavender appears to vided.Five trials(Braden et al.2009;Kritsidima et al.2010;Morris be an effective and well-tolerated alternative to benzodiazepines 2002:Muzzarelli et al.2006:Xu et al.2008)were described as for amelioration of generalised anxiety.Lavender should,however, single-blind with the patient/participant being blinded.As laven- be used with caution particularly as lavender oil can also be a pow der is a particularly distinctive and recognisable odour,blinding is erful allergen and,as a precaution,ingestion should be avoided unrealistic in most of the trials (i.e.trials comparing lavender aro- during pregnancy due to emmenagogue effects (Ernst 2002)and matherapy versus no odour(Bradley et al.2009;Soden et al.2004). when breastfeeding. or alternative odour(Howard and Hughes 2008)).Blinding was pos- Given funding for CAM research is difficult to obtain and our sible in trials that used oral lavender capsules(Kasper et al.2010; review did not identify convincing evidence for lavender aro- Woelk and Schlafke 2010).where smell would not be detected. matherapy,massage or oil-dripping.additional research should Specific odours in general may provide the calmative effects via focus on replication of oral lavender.Future trials of oral laven- associative memory evocation rather than any direct pharmaco- der would benefit from adopting a good trial design and stringent logical action.This possibility needs to be addressed in future trials reporting to enable replication:this would include adopting a perhaps by comparing lavender with an alternative odours in addi- randomised design with allocation concealment,triple blind (if tion to 'no odour'conditions,as considered in Howard and Hughes possible)and indistinguishable placebos.Intention-to-treat anal- (2008).Also,a single-blind design would mean the outcome asses- yses and a priori power calculations should be conducted and sor was aware of the condition and given that they reported on reported.All withdrawals,dropouts and adverse events should be treatment effectiveness in one trial(Dunn et al.1995).blinding of fully reported giving the number and reason by group.It is essen- outcome assessor is essential to reduce the effect of demand char- tial to get good tolerability and safety data for all modes of lavender acteristics or the Hawthorne effect.Failure to blind may therefore application,thus longer-term follow ups would be required for oral have increased the risk of bias. lavender before it is recommended as an alternative treatment for Withdrawals/dropouts were generally under-reported.In par- anxiety. ticular,the trials that induced an anxiety state were particularly poor at reporting this information and in many cases it was Limitations assumed there were no dropouts due to the short duration of the tri- als,although this was not explicitly stated.Three trials(Kritsidima As described above,the methodological issues observed in the et al.2010:Morris 2002:Muzzarelli et aL 2006)used intention-to- included studies are notable and limit the extent to which the anxi- treat analyses (ITT),two (Kritsidima et al.2010:Muzzarelli et al. olytic properties of lavender can be evaluated.The methodological 2006)were the result of not having any dropouts.A further four issues of the lavender inhalation trials are most difficult to manage; trials(Braden et al.2009;Morris 2002:Motomura et al.2001:Toda trying to ensure that consistent levels of lavender odour is released and Morimoto 2008)also appeared to use ITT analyses due to no between groups,or when attempting successful replication of such drop-outs but again this was only assumed due to the short dura- trials,can be particularly hard to control.In addition,although the tion of the trial. search strategy was thorough,some clinical trials may not haveR. Perry et al. / Phytomedicine 19 (2012) 825–835 833 comparative trial byWoelk and Schlafke (2010)found thatlavender was as effective as the anxiolytic drug lorazepam. Of the remaining studies,the findings of one RCT (Muzzarelli et al. 2006) could not be evaluated because it did not provide between-group comparisons of change in outcome from baseline. The remaining six trials did not demonstrate significant differences between groups. However, all results should be considered in the context of their methodological limitations. The results suggestlimited specific effects oflavender inhalation and massage on anxiety measures. Although reductions in anxiety measures were observed, methodological issues limit the extent to which firm conclusions can be drawn. On the other hand, promising evidence regarding ingested lavender preparations was found. The reporting of many of the trials was generally poor; all were described as randomised but only five gave details on randomisa￾tion process (Braden et al. 2009; Kasper et al. 2010; Morris 2002; Muzzarelli et al. 2006; Woelk and Schlafke 2010) and two on treat￾ment allocation concealment (Dunn et al. 1995; Soden et al. 2004). Five trials (Braden et al. 2009; Dunn et al. 1995; Kasper et al. 2010; Kritsidima et al. 2010; Soden et al. 2004) included power calcu￾lations, although one was underpowered at baseline (Soden et al. 2004) and two became marginally underpowered following par￾ticipant drop out (Dunn et al. 1995; Kasper et al. 2010). Most trials reported that participants were similar on prognostic indicators, although this was assumed rather than tested in the majority of tri￾als that used healthy participants. Three trials (Braden et al. 2009; Sgoutas-Emch et al. 2001; Soden et al. 2004) reported differences in anxiety levels at baseline. In addition, the trials differed greatly in terms of the conditions treated, and independent replications are missing for most. Five trials were described as double-blind (Bradley et al. 2009; Howard and Hughes 2008; Kasper et al. 2010; Soden et al. 2004; Woelk and Schlafke 2010), although in two (Bradley et al. 2009; Woelk and Schlafke 2010), blinding details were not clearly pro￾vided. Five trials (Braden et al. 2009; Kritsidima et al. 2010; Morris 2002; Muzzarelli et al. 2006; Xu et al. 2008) were described as single-blind with the patient/participant being blinded. As laven￾der is a particularly distinctive and recognisable odour, blinding is unrealistic in most of the trials (i.e. trials comparing lavender aro￾matherapy versus no odour (Bradley et al. 2009; Soden et al. 2004), or alternative odour (Howard and Hughes 2008)). Blinding was pos￾sible in trials that used oral lavender capsules (Kasper et al. 2010; Woelk and Schlafke 2010), where smell would not be detected. Specific odours in general may provide the calmative effects via associative memory evocation rather than any direct pharmaco￾logical action. This possibility needs to be addressed in future trials, perhaps by comparing lavender with an alternative odours in addi￾tion to ‘no odour’ conditions, as considered in Howard and Hughes (2008). Also, a single-blind design would mean the outcome asses￾sor was aware of the condition and given that they reported on treatment effectiveness in one trial (Dunn et al. 1995), blinding of outcome assessor is essential to reduce the effect of demand char￾acteristics or the Hawthorne effect. Failure to blind may therefore have increased the risk of bias. Withdrawals/dropouts were generally under-reported. In par￾ticular, the trials that induced an anxiety state were particularly poor at reporting this information and in many cases it was assumedthere werenodropoutsdue to the shortdurationofthe tri￾als, although this was not explicitly stated. Three trials (Kritsidima et al. 2010; Morris 2002; Muzzarelli et al. 2006) used intention-to￾treat analyses (ITT), two (Kritsidima et al. 2010; Muzzarelli et al. 2006) were the result of not having any dropouts. A further four trials (Braden et al. 2009; Morris 2002; Motomura et al. 2001; Toda and Morimoto 2008) also appeared to use ITT analyses due to no drop-outs but again this was only assumed due to the short dura￾tion of the trial. Furthermore, the majority of the included trials did not pro￾vide reasons by group for dropouts. This is a frequent phenomenon in research of this type and there is a common misconception that “natural” means safe (Ernst 2007a). Researchers investigating botanical products should comply with the CONSORT guidelines for the reporting of herbal products (Gagnier et al. 2006). None of the included trials of oral lavender (Bradley et al. 2009; Kasper et al. 2010; Woelk and Schlafke 2010) provided information that met more than 9 of the 15 CONSORT statement criteria regarding the extraction and preparation of herbs. Future trials should report such information to allow adequate between-trial comparison of the results to be made. Safety The current data suggests that lavender is relatively safe, although the gastrointestinal effects of oral lavender being taken in the long term haven’t been assessed. Of a total of 15 randomised tri￾als only 5 trials (Dunnet al. 1995;Kasper et al. 2010;Kritsidima et al. 2010; Kutlu et al. 2008; Woelk and Schlafke 2010) reported infor￾mation on adverse events. One (Dunn et al. 1995) had AEs that were not related to the therapy. The remaining ten did not mentionAEs at all. Itis good practice for alltrials to report on adverse events even if unrelated to the therapy. In these RCTs, most of the AEs mentioned were minor ailments. Kutlu et al. (2008) had “some” participants reporting discomfortfrom the odour thus were excluded. In the oral lavender trials, Kasper et al. (2010) reported slightly more adverse events from the placebo group than the lavender group; the most frequently reported AEs were related to infections and infestations, followed by gastrointestinal disorders and nervous system disor￾ders. Woelk and Schlafke (2010) reported slightly more adverse events in the lavender group than the lorazepam group but again none were described as serious. Since lavender oil showed no seda￾tive effects and has no potential for drug abuse, lavender appears to be an effective and well-tolerated alternative to benzodiazepines for amelioration of generalised anxiety. Lavender should, however, be used with caution particularly as lavender oil can also be a pow￾erful allergen and, as a precaution, ingestion should be avoided during pregnancy due to emmenagogue effects (Ernst 2002) and when breastfeeding. Given funding for CAM research is difficult to obtain and our review did not identify convincing evidence for lavender aro￾matherapy, massage or oil-dripping, additional research should focus on replication of oral lavender. Future trials of oral laven￾der would benefit from adopting a good trial design and stringent reporting to enable replication: this would include adopting a randomised design with allocation concealment, triple blind (if possible) and indistinguishable placebos. Intention-to-treat anal￾yses and a priori power calculations should be conducted and reported. All withdrawals, dropouts and adverse events should be fully reported giving the number and reason by group. It is essen￾tialto get good tolerability and safety data for all modes of lavender application, thus longer-term follow ups would be required for oral lavender before it is recommended as an alternative treatment for anxiety. Limitations As described above, the methodological issues observed in the included studies are notable and limit the extent to which the anxi￾olytic properties of lavender can be evaluated. The methodological issues ofthe lavender inhalation trials are most difficultto manage; trying to ensure that consistent levels of lavender odour is released between groups, or when attempting successful replication of such trials, can be particularly hard to control. In addition, although the search strategy was thorough, some clinical trials may not have