Systematic reviews of complementary therapies - an annotated bibliography. Part 2: herbal medicine.

BACKGROUND: Complementary therapies are widespread but controversial. We aim to provide a comprehensive collection and a summary of systematic reviews of clinical trials in three major complementary therapies (acupuncture, herbal medicine, homeopathy). This article is dealing with herbal medicine. Potentially relevant reviews were searched through the register of the Cochrane Complementary Medicine Field, the Cochrane Library, Medline, and bibliographies of articles and books. To be included articles had to review prospective clinical trials of herbal medicines; had to describe review methods explicitly; had to be published; and had to focus on treatment effects. Information on conditions, interventions, methods, results and conclusions was extracted using a pre-tested form and summarized descriptively.
RESULTS: From a total of 79 potentially relevant reviews pre-selected in the screening process 58 met the inclusion criteria. Thirty of the reports reviewed ginkgo (for dementia, intermittent claudication, tinnitus, and macular degeneration), hypericum (for depression) or garlic preparations (for cardiovascular risk factors and lower limb atherosclerosis). The quality of primary studies was criticized in the majority of the reviews. Most reviews judged the available evidence as promising but definitive conclusions were rarely possible.
CONCLUSIONS: Systematic reviews are available on a broad range of herbal preparations prescribed for defined conditions. There is very little evidence on the effectiveness of herbalism as practised by specialist herbalists who combine herbs and use unconventional diagnosis.

Introduction

In this second part of our series on systematic reviews in complementary therapies we report our findings on herbal medicines. Herbal medicines (defined as preparations derived from plants and fungi, for example by alcoholic extraction or decoction, used to prevent and treat diseases) are an essential part of traditional medicine in almost any culture [1]. In industrialized countries herbal drugs and supplements are an important market. Some countries like Germany have a long tradition in the use of herbal preparations marketed as drugs and figures for prescriptions and sales are stable or slightly declining [2]. In the US and the UK herbal medicinal products are marketed as "food supplements" or "botanical medicines". In recent years sales of such products have been increasing strongly in these countries [3,4]. In the Third World herbs are mainly used by traditional healers [5].

Methods

A detailed description of the methods used in this review of reviews is given in the first part of this series [6]. For searches in Medline 50 single plant names and the 'exploded' term 'medicinal plants' were combined with the standard search strategy for systematic reviews. As a specific intervention-related inclusion criterion we required that reports reviewed prospective (not necessarily controlled) clinical trials of substances extracted from plants in humans. Reviews dealing with single substances (e.g., artemisin derivatives) derived from plants were excluded on the grounds that such agents are comparable to conventional drugs. Disease-oriented reviews including a variety of interventions were included only if they reviewed at least 4 herbal medicine trials.

Results

From a total of 79 potentially relevant reviews preselected in the literature screening process, 58 (published in 65 papers) met the inclusion criteria [7-71]. Eleven reports were not truly systematic reviews (not meeting inclusion criterion 2) [72-82], 5 dealt with isolated substances of plant origin [83-87] and 4 were excluded for other reasons (one disease- focused review with less than 4 herbal medicine trials [88], one review not on preventative or therapeutic use [89], two reviews not truly herbal medicine [90,91]).

More than half of the reports reviewed gingko, hypericum or garlic preparations. No less than 13 systematic reviews dealed with ginkgo (Ginkgo biloba) extracts (see table 1). Seven of these reviewed trials (total number of trials covered in any of the reviews 15) in patients with intermittent claudication [7-13]. Most of these reviews concluded that ginkgo extracts were significantly more effective than placebo in increasing measures like walking distance but the clinical relevance of the effects was felt to be moderate by some reviewers. The five reviews dealing with dementia and cerebral insufficiency (total number of trials included about 50) all draw positive conclusions [13-17]. However, many of the older trials were in patients with minor cognitive impairment and more evidence is needed to decide whether ginkgo extracts have clinically relevant beneficial effects in more severe forms of dementia. Finally, one review found that ginkgo extracts might be effective in the treatment of tinnitus [18] and another found insufficient evidence for efficacy in patients with macular degeneration [19].

Table 1

Systematic reviews of clinical trials of ginkgo biloba extracts

					Features
Author Year	Indication	Intervention	Comparisons	Studies	1/2/3/	Results	Author's Conclusion
					4/5
Ginkgo (Ginkgo biloba)
Pittler 2000	intermittent	ginkgo	placebo	8 RCT	y/y/y/	Increase of pain-free walking	Evidence for a modest benefit of
[7]	claudication				y/y	distance over placebo after 12	uncertain clinical relevance
						or 24 weeks 34 m (95%CI 26–
						43 m)
Moher 2000	intermittent	ginkgo*	placebo	5 RCT	y/y/y/	Increase of pain-free walking	Inconsistent results from the few
[8]	claudication				n/y	distance over placebo after 24	available small studies do not
						weeks 32 m (95%CI 14–50 m)	allow firm conclusions
Ernst 96 [9]	intermittent	ginkgo	placebo,	10	p/ p/ n/	Most studies low quality.	Available evidence promising but
	claudication	extract	other drugs	RCT/CCT	n/n	Increase of walking distance	further high quality research
		EGb761				compared to placebo 24 to 160	needed
						m. At least similar
						effectiveness compared to
						other drugs.
Schneider 92	intermittent	ginkgo	placebo,	7 RCT/CCT	?/n/n/	mean effect size d = 0.75	Effectiveness over placebo clearly
[10]	claudication		other	(vs. plac.), 2	y/y	(95%CI 0.44–1.07) over	shown
			treatment	RCT/CCT		placebo
				(other)
Letzel 92	intermittent	ginkgo	ginkgo vs.	5 RCT	?/p/n/	Pooled increase of walking	Ginkgo extract EGb761 more
[11]	claudication	extract	plac.,	ginkgo	y/y	distance: 45% over placebo for	effective than placebo and
		EGb 761	pentoxifyllin	9 RCT		gingko and 57% for	similarly effective as pentoxifyllin
			vs. plac.	pentoxifyllin		pentoxifyllin
Kleijnen 91	intermittent	ginkgo	ginkgo vs.	15	y/y/y/	Many trials low quality. All trials	Ginkgo seems effective for
[12]	claudication		plac.,	RCT/CCT	n/n	with positive results. Evidence	intermittent claudication but further
			pentoxifyllin	(ginkgo), 5		similar as for pentoxifyllin	high quality studies are needed
			vs. placebo	RCT/CCT
				pentoxif.
Weiss 91	cerebral ins.,	ginkgo	placebo	17RCT/CCT	?/p/p/	10 of 12 interpretable trials on	Effectiveness for both conditions
[13]	intermittent	extract		(cerebral	n/n	cerebral insufficieny and all 4	biometrically shown
	claudication	EGb761		ins.), 8		interpretable trials on
				RCT/CCT		intermittent claudication with
						significant positive results
Ernst 99 [14]	dementia	ginkgo	placebo	9 RCT	y/y/y/	Results collectively suggest	Encouraging findings warranting
					y/n	that ginkgo is more effective for	large scale trials
						dementia than placebo
Oken 98 [15]	Alzheimer	ginkgo	placebo	4 RCT	y/y/n/	Significant effect over placebo	Clinical relevance of the observed
	dementia				y/y	for cognitive function (Hedges	effects has to be confirmed in
						g= 0.41, 95%CI 0.22–0.61)	further research
Hopfenmüller	cerebral	ginkgo	placebo	10 RCT, 1	n/ n/ n/	Global response (based on	Ginkgo extract superior to placebo
94 [16]	insufficiency	extract LI		CCT	y/y	symptom scores): OR 1.98
		1370				(95%C11.39–2.57) in favour of
						Ginkgo
Kleijnen 92	cerebral	ginkgo	ginkgo vs.	40 RCT/	y/y/y/	Many trials low quality. Virtually	Ginkgo seems effective for
[17]	insufficiency		plac.	CCT	n/n	all trials reported positive	cerebral insufficiency but further
			hydergine	(ginkgo), 4		results. Evidence similar as for	high quality studies are needed
			vs. plac.	RCT/CCT		hydergine
				(hydergine)
Ernst 99 [18]	tinnitus	ginkgo	placebo,	5 RCT	y/y/y/	3 trials favour ginkgo over	Results suggest that extracts of
			other		y/n	placebo, 1 no difference, in one	ginkgo biloba are effective in
			treatment (1			trial ginkgo better than another	treating tinnitus
			trial)			treatment
Evans 2000	macular	ginkgo	placebo	1 RCT	y/y/y/	one small trial reporting	Insufficient evidence to
[19]	degeneration				y/-	improvement	recommend ginkgo for age-related
							macular degeneration

Features: 1 = comprehensive search, 2 = explicit inclusion criteria, 3 = formal quality assessment, 4 = summary of results for each included study, 5 = meta-analysis; y = yes, p = partly, n = no, - = not applicable, ? = unclear review on all pharmacologic treatments for the respective condition RCT = randomized controlled trials, CCT = non-randomized controlled trials, CS = cohort studies, UCS = uncontrolled studies; OR = odds ratio, RR = rate ratio

The effectiveness of St. John's wort (Hypericum perforatum) extracts in depression was investigated in nine reviews [20-30] (total number of trials covered 29; see table 2). Mainly due to slight differences in the inclusion criteria (for example, restriction to trials with a minimum of 6 weeks observation or with a minimum quality score) the respective study collections differed to a considerable amount. However, the conclusions were very similar. Hypericum extracts have been shown to be superior to placebo in mild to moderate depressive disorders. There is growing evidence that hypericum is as effective as other antidepressants for mild to moderate depression and causes fewer side effects but further trials are still needed to establish long-term effectiveness and safety.

Table 2

Systematic reviews of clinical trials of hypericum and garlic preparations

					Features
Author	Indication	Intervention	Comparisons	Studies	1/2/3/	Results	Author's Conclusion
Year					4/5
St John's wort (Hypericum perforatum)
Gaster	depression	hypericum	placebo and	8 RCT	p/y/p/	4 placebo-controlled trials with	Data suggest that hypericum is
2000 [20]			antidepressants		y/n	positive results, in 4 trials	superior to placebo, insuffcient
						standard antidepr. tended to be	evidence re equivalence with
						slightly better	antidepressants
Williams	depression	hypericum	placebo and	14 RCT	y/y/n/	Treatment response: RR 1.9	Data suggest that hypericum is
2000 &		(and other	antidepressants		y/y	(95%C11.2–2.8) vs. placebo and	superior to placebo, insuffcient
Mulrow 98		drugs)				1.2 (1.0–1.4) vs. antidepressants	evidence re equivalence with
[21,22]							antidepressants
Kim 99 [23]	depression	hypericum	placebo and	6 RCT	p/y/y/	Treatment response: RR 1.48	Hypericum more effective than
			antidepressants		y/y	(95%C11.03–1.92) vs. placebo	placebo and similarly effective as
						and 0.98 (0.67–1.28) vs.	low dose antidepressants; quality
						antidepressants	problems
Stevinson	depression	hypericum	placebo and	6 RCT	y/y/y/	Only trials published after Linde	Data confirm findings of earlier
99 [24]			antidepressants		y/n	96; trials show effects better	trials, but still insuff. evidence to
						than placebo/similar to	assess equivalence with
						antidepressants	antidepressants
Linde 98 &	depression	hypericum	placebo and	27 RCT	y/y/y/	Treatment response: RR 2.47	Hypericum more effective than
96 [25,26]			antidepressants		y/y	(95%C11.69–3.61) vs. placebo	placebo. Inadequate evidence to
						and 1.01 (0.87–1.16) vs.	assess equivalence with
						antidepressants	antidepressants
Volz 97	depression	hypericum	placebo and	15	p/p/n/	Most placebo-controlled trials	A therapy with hypericum of mild
[27]			antidepressants	RCT/CCT	n/n	positive; similarly effective as	and moderate depression can be
						(not adequately dosed)	attempted. Further studies needed
						antidepressants
Ernst 95	depression	hypericum	placebo and	11 RCT	y/y/y/	Most of 8 placebo-controlled	Hypericum is superior to placebo
[28]			antidepressants		y/n	trials positive. 3 trials against	and seems equally effective as
						standard medication with similar	standard medication
						effects
Volz 2000	mild to	hypericum	fluoxetine	17+9	n/y/n/	No direct comparison of	Response rates are similar;
[29]	mod.			CCT	y/n	hypericum and fluoxetine	findings difficult to interpret
	depression					available. Mean depression	because of the indirect comparison
						score (HAMD) reduction in
						hypericum trials 53%, in
						fluoxetine trials 55%
Friede 98	anxiety in	hypericum	placebo,	8 RCT	?/y/y/	Trials collectively show reduction	Hypericum is effective for
[30]	depressed		amitriptyline		y/n	of anxiety symptoms over	depressed patients with anxiety
	p.					placebo. Only 1 trial vs
						amitriptyline
Garlic (Allium sativum)
Lawrence	cardiovasc.	garlic	mainly placebo;	45 RCT	y/y/y/	37 trials consistently show small	Insufficient data to draw conclusion
2000 [31]	risk factors		no & other		y/y	short-term effects over placebo	regarding clinical cardiovascular
			treatment			for cholesterol reduction. No	outcomes. Garlic preparations may
						consistent effects on blood	have small, positive, short-term
						pressure, promising effects re	effects on lipids
						platelet aggregation and
						fibrionolytic activity
Stevinson	hyperchol-	garlic	placebo	13 RCT	y/y/y/	Pooled total cholesterol	Available data suggest that garlic is
2000 [32]	esterolemia				y/y	reduction over placebo 0.41	superior to placebo. The size of the
						(95% Cl -0.66 to -0.15) mmol/l;	effect is modest. The use of garlic
						when analysis restricted to high	for hyperchol. is therefore of
						quality trials 0.11 (-0.30 to 0.08)	questionable value
Silagy 94 &	cholesterol	garlic	placebo	16 RCT	y/p/y/	Pooled cholesterol reduction	Meta-analysis suggests positive
Neil 96	lowering				y/y	over placebo 0.65 (95% Cl 0.53–	effects but reviewers are sceptic
[33,34]						0.76) mmol/l	(low quality; own replication
							negative)
Warshafsky	cholesterol	garlic	placebo	5 RCT	p/y/y/	Pooled cholesterol reduction	Available evidence supports the
93 [35]	lowering				y/y	over placebo 0.59 (95%Cl 0.44–	use of garlic as one modality to
						0.74) mmol/l	decrease cholesterol levels
Silagy 94	lowering	dried garlic	placebo, other	8 RCT	y/p/y/	Pooled reduction over placebo:	Garlic maybe of some clinical use
[36]	blood	(Kwai)	treatment		y/y	SBP 7.7 (95% Cl 4.3–11.0), DBP	in subjects with mild hypertension.
	press.					5.0 (2.9–7.1) mm Hg	Further research needed
Kleijnen 91	cardiovasc.	garlic	placebo	18	p/p/y/	Most studies with shortcomings.	No clear conclusion drawn
[37]	risk factors	supplements		RCT/CCT	y/n	The majority of trials with pos.
						results but inconsistent effect
						sizes
Kleijnen 89	cardiovasc.	garlic &	unclear	10 RCT,	y/p/n/	All trials with severe	Inadequate evidence to justify
[38]	risk factors	onions		8 CCT	y/n	shortcomings. Fresh garlic with	supplementation, further research
						beneficial effcts, onions and	needed
						commercially available
						supplements yielded
						contradictory results
Jepson 97	lower limb	garlic	placebo	1 RCT	y/y/y/	Walking distance not	Insufficient evidence
[39]	atheroscler.				y/-	significantly different between
						groups

legend see table 1

Eight reviews have been performed on garlic (Allium sativum) for cardiovascular risk factors [31-38] (total number of trials covered about 50) and lower limb atherosclerosis [39] (see table 2). A modest short-term effect over placebo on lipid-lowering seems to be established but the clinical relevance of these effects is uncertain. Data from randomised trials on cardiovascular mortality are not available. Effects on blood pressure seem to be at best minor. The available results on fibrinolytic activity and platelet aggregation are promising but insufficient to draw clear conclusions. A specific problem in research on garlic is the great variety of garlic preparations used: the exact content of bioactive ingredients in these is often unclear.

Three reviews (covering a total of about 30 trials) have been performed on preparations containing extracts of Echinacea (Echinacea purpurea, pallida or angustifolia), two of which by the same study group [40-43]. The results suggest that Echinacea preparations may have some beneficial effects mainly in the early treatment of common colds. Similar to garlic a major problem is the high variaton of bioactive compounds between different Echinacea preparations. Cranberries (Vaccinium macrocarpon) for urinary tract infections [44,45], mistletoe (Viscum album) for cancer [46-48], peppermint (Mentha piperita) oil for irritable bowel syndromes [49,50] and saw palmetto (Serenoa repens) for benign prostate hyperplasia [51-53] have each been subject to two reviews. For saw palmetto there is good evidence for efficacy over placebo while for the other three the data are inconclusive (see table 3).

Table 3

Systematic reviews of clinical trials of herbal medicines (at least 2 reviews per herb)

					Features
Author	Indication	Intervention	Comparisons	Studies	1/2/3/	Results	Author's Conclusion
Year					4/5
Echinacea (Echinacea purpurea, angustifolia and pallida)
Barrett	upper resp.	echinacea	placebo	13RCT	y/p/y/	Overall quality modest. All 4	Echinacea may be beneficial for
99 [40]	infections	(incl.			y/n	prevention studies show only	early treatment of acute upper
		combinations)				minor trends, 8 of 9 treatment	respiratory infections; little evidence
						studies with generally positive	to support the prolonged use for
						results	prevention
Melchart	common	echinacea	placebo, no	16 RCT	y/y/y/	Minor effects in prevention and	Echinacea extract can be efficacious
99 [41]	cold	(incl.	treatment		y/p	treatment, promising effects in	for the common cold, but evidence
		combinations)				early treatment. Heterogen.	insufficient for recommendations
						preparations
Melchart	immuno-	echinacea	placebo, no	18 RCT, 8	y/y/y/	Most studies low quality. Most	Echinacea extracts can be
94	stimulation	(incl.	treatment	CCT	y/n	studies show immunostimulating	efficacious immunostimulators, but
[42,43]		combinations)				effects	evidence insufficient for
							recommendations
Cranberries (Vaccinium macrocarpon)
Jepson	urinary	cranberries	placebo	4 RCT	y/y/y/	In 3 of 4 trials cranberries effective	Insufficient evidence, further research
98 [44]	tract inf.				y/n	for at least one of the outcomes of	needed
	(prevent)					interest
Jepson	urinary	cranberries		O RCT	y/y/-/	No trials meeting the inclusion	No evidence available
98 [45]	tract inf.				-/-	criteria
	(treatm.)
Mistletoe (Viscum album)
Kleijnen	cancer	mistletoe	placebo, no	11	y/y/y/	Most studies low quality. Most	Insufficient evidence to recommend
94 [46]			treatment	RCT/CCT	n/n	studies show longer survival with	mistletoe outside of clinical trials
						mistletoe but not the best trial
Kiene 89	cancer	mistletoe	no treatment,	2 RCT, 33	y/n/n/	Most studies low quality. 9 of 12	Available evidence supports positive
[47,48]			none	CCT, 11	y/n	interpretable studies suggest	effects of mistletoe
				other		positive effects on survival
				studies
Peppermint (Mentha piperita)
Jailwala	irritable	1. peppermint	placebo	1. 3 RCT	p/y/y/	Chinese herbal therapy trial rated	In both cases efficacy not clearly
2000*	bowel	oil		2. 1 RCT	n/n	as positive, one of three	established
[49]	syndr.	2. Chinese				peppermint oil trials rated as
		herbal				positive
		therapy
Pittler 98	irritable	peppermint	placebo,	8 RCT	y/y/y/	Global improvement rates	The role of peppermint oil for IBS
[50]	bowel	oil	other		y/y	significantly higher compared to	has not been established beyond
	syndr.		treatment			placebo. Quality of trials doubtful	reasonable doubt
Saw palmetto (Serenoa repens)
Boyle	ben.	Permixon®	placebo,	11 RCTs,	?/n/n/	peak urine flow 2.20 (95% Cl 1.20–	Despite some limitations strong
2000 [51]	prostate	(saw	other	2 UCS	y/y	3.20) ml/s increase over placebo;	evidence that the extract tested has
	hyperplasia	palmetto)	treatment			significant decrease nocturia	beneficial effects
Wilt 2000	ben.	saw palmetto	placebo,	14 RCT	y/y/y/	Saw palmetto superior to placebo	Evidence suggests that saw
&98	prostate		other	(plac),	y/y	for nocturia, self rating, peak urine	palmetto improves urological
[52,53]	hyperplasia		treatment	5 RCT		flow; similar effects as finasteride	symptoms and flow measures.
				(other)			Further studies needed

legend see table 1

Single systematic reviews have been published on aloe (Aloe vera) [54], artichoke (Cynara scolymus) leave extract [55], evening primrose (Oenothera biennis) oil [56], feverfew (Tanacetum parthenium) [57], ginger (Zingiber officinialis) [58], ginseng (Panax ginseng) [59], horse chestnut (Aesculus hippocastanum) seeds [60], kava (Piper methysticum) [61], milk thistle (Silybum marianum) [62], a fixed combination of three herbal extracts [63], rye-grass pollen (Secale cereale) extract [64,65], tea tree (Melaleuca alternafolia) oil [66], and valerian (Valehana officinalis) root [67] (see table 4). The only review which focused on a herbal intervention which is not marketed as a drug or food supplement was on cabbage leaves for breast engorgement and included a single small-scale trial [68]. Chinese herbal therapy for atopic eczema [69] and a variety of herbs for lowering blood glucose [70] and for analgesic and anti-inflammatory purposes [71] have also been reviewed. For some of these herbal preparations the evidence is promising but further studies are considered necessary to establish efficacy in almost every case.

Table 4

Systematic reviews of clinical trials of herbal medicines

					Features
Author	Indication	Intervention	Comparisons	Studies	1/2/3/	Results	Author's Conclusion
Year					4/5
Vogler 99	various	aloe	placebo, other	6 RCT,4	y/y/y/	Positive results for genital	Promising results, but overall
[54]			& no treatment	CCT	y/n	herpes, psoriasis, hyper-	evidence insufficient
						lipidemia, diabetes;
						contradictory for wound healing
Pittler 98	cholesterol	artichoke	placebo	1 RCT	y/y/y/	Effects over placebo only in the	More trials needed
[55]	lowering	leave			n/n	subgroup of participants with
		extract				serum cholesterol > 210 mg/dl
Morse 89	atopic	evening	placebo	9	?/n/n/	Epogam significantly better	No conclusion drawn
[56]	eczema	primrose oil		RCT/CCT	y/y	than placebo for most
		(Epogam)				outcomes
Vogler 98	migraine	feverfew	placebo	5 RCT	y/y/y/	Majority of trials favor feverfew	Effectiveness has not been
[57]					y/n	over placebo	established beyond reasonable
							doubt
Ernst 2000	nausea and	ginger root	placebo,	6 RCT	y/y/y/	2 of 3 trials on postoperative	Evidence promising but insufficient
[58]	vomiting		metoclopramide		y/p	nausea positive (best	to draw firm conclusions
						negative), trials on
						seasickness, morning sickness
						and chemotherapy-induced
						nausea positive
Vogler 99	various	ginseng root	placebo, other	16 RCT	y/p/y/	Contradictory results re.	The efficacy of ginseng root extract
[59]		extract	treatment (1		y/n	physical performance (7 trials),	is not established beyond
			trial)			psychological function (5),	reasonable doubt for any of these
						immunomodulation (2),	indications
						positive results in diabetes and
						herpes simplex (1 trial
						respectively)
Pittler 98	venous	horse	placebo, other	13 RCT	y/y/y/	Significant effects over placebo	horse chestnut seeds seem to be
[60]	insufficieny	chestnut	treatment		y/n	and similar effects compared to	effective; further tials needed
		seeds				other treatments	(confirmation, long-term results,
							combination)
Pittler 2000	anxiety	kava	placebo	7 RCT	y/y/y/	All trials suggest superiority	Available data suggest that kava is
[61]					p/p	over placebo; 3 trials with data	a treatment option for anxiety.
						for meta-analysis show sign.	Further studies needed
						superiority
Lawrence	liver	milk thistle	placebo, other	33 RCT,	y/y/y/	Variety of conditions studied,	Efficacy is not established.
2000 [62]	diseases		& no treatment	1 CCT	y/y	studies often poor quality.	Possible benefit shown most
						Mixed and inconsistent findings	frequently for aminotransferases.
Ernst 99	musculoskel.	Phytodolor®	placebo, other	10 RCT	y/p/y/	Placebo-controlled trials show	The data suggest that the
[63]	pain	populus,	treatments		y/n	superiority over placebo and	combination is effective in the
		fraxinus,				similar effects as NSAIDs	symptomatic treatment of
		solidago					muskuloskeletal pain
MacDonald	ben. prostata	rye grass	placebo, other	4 RCT	y/y/y/	Signif. improvement over	Available evidence suggests that
2000 &	hyperplasia	pollen	therapy		y/y	placebo in subjective, but not	Cernilton® is well tolerated and
Wilt 2000		extract				objective symptoms; no	modestly improves subjective
[64,65]						differences compared to	symptoms. Further studies needed
						tadenan and paraprost
Ernst 2000	dermatologic	tea trea oil	placebo, other	4 RCT	y/y/y/	2 trials vs. placebo positive, 3	Data promising but insufficient
[66]	conditions		treatment		y/n	trials vs. other treatments
						similar effects
Stevinson	insomnia	valerian root	placebo	9 RCT	y/y/y/	Highly heterogeneous studies	Available evidence is promising but
2000 [67]					y/n	with sometimes contradictory	not fully conclusive. Further,
						and inconsistent findings	rigorous trials needed
Renfrew	breast	cabbage	usual care	1 RCT	y/y/n/	fewer women stopping breast	Further research desirable
84 [68]	engorgement	leaves			y/n	feeding among those receiving
						cabbage leaves
Armstrong	atopic	Chinese	placebo	2 RCT	y/y/n/	2 positive studies by the same	Evidence encouraging but
99 [69]	eczema	herbal			y/n		insufficient given the potential of
		therapy				treat analysis	relevant side effects
Ernst 97	hypoglyc.	all plants	no treatment,	7 RCT, 4	y/p/n/	Most studies low quality. Most	Use of hypoglcemic plant remedies
[70]	activity		placebo, none	CCT, 10	y/n	papers report positive effects	not supported by rigorous
				UCS		on a variety of plants	research. Further studies required
Ernst 2000	analgetic or	various	placebo	18 RCT	y/y/y/	Trials on evening primrose oil,	The results suggest that several
[71]	inflamm.				y/n	blackcurrant seed oil, borage	herbal remedies have potential in
	treatment					oil, harpagophytum, willow	alleviating the pain of rheumatic
						bark, feverfew, and 3	diseases. More research urgently
						combinations; almost all trials	needed
						positive

legend see table 1

Discussion

Our overview shows that a considerable number of systematic reviews on herbal medicines is available. In the majority of cases the reviewers considered the available evidence as promising but only very rarely as convincing and sufficient as a firm basis for clinical decisions. The methodological quality of the primary studies has been criticized by many reviewers.

Our summary of the existing studies must be interpreted with caution. What we performed is a systematic review of systematic reviews which inherently bears a large risk of oversimplification. Readers who want to reliably assess the evidence for a given herb for a defined condition should read the respective reviews. Our collection – which to the best of our knowledge is complete up to summer 2000 – is aimed at facilitating the access and giving an idea of the amount of the available evidence. Based on the increase of herbal medicine reviews in recent years we expect that at least ten new publications will become available in the year 2001.

Most of the currently available systematic reviews address herbal preparations which are marketed and widely used in industrialized countries. However, the widespread traditional use of herbs in the Third World is rarely ever investigated and has not been subjected to systematic reviews. The many herbs used in folk medicine or other traditional uses of herbs (for example, hypericum is used for a variety of ailments other than depression including enuresis, diarrhoea, gastritis, bronchitis, asthma, sleeping disorders etc.) seem to be rarely investigated. Furthermore, practitioners of herbal medicine often combine different herbs and use unconventional diagnostic approaches to adapt prescriptions to single patients. It seems likely that these traditional forms of herbal medicine will remain underresearched relative to single herbal preparations due to the lack of financial incentive for sponsors and due to methodological problems.

Herbal medicines products are not, in general, subject to patent protection. This reduces the motivation for drug companies to invest in trials. Many of the existing herbal medicine manufacturers are comparably small companies, often with limited research resources and expertise. Maybe partly for these reasons, the quality of many older herbal medicine trials is low. Furthermore, negative trials which could threaten the company's survival might not become published.

A fundamental problem in all clinical research of herbal medicines is whether different products, extracts, or even different lots of the same extract are comparable and equivalent. This is a major issue in the expert research community and a major obstacle to a reliable assessment for the non-expert. For example, Echinacea products can contain other plant extracts, use different plant species (E. purpurea, pallida or angustifolia), different parts (herb, root, both), and might have been produced in quite different manners (hydro- or lipophilic extraction). Pooling studies that use different herbal products in a quantitative meta- analysis can be misleading. Health care professionals and patients considering to prescribe or take a particluar herbal product should check carefully whether the respective product or extract has been tested in the trials included in a review. On the health food store shelf the high quality, standardized products used in the trials might not be available. Only a herbal medicine expert can judge with some certainty whether the results can be extrapolated to the product of interest.

On the level of health care policies the available systematic reviews more often provide insight into the deficiencies of the evidence than guidance for decision making. Trials on hard endpoints are very rarely available and observation periods have generally been short. The clinical relevance of the observed effects is not always clear.

Herbal medicines are generally considered as comparably safe. While this is probably correct case reports show that severe side effects and relevant interactions with other drugs can occur. For example, hypericum extracts cause considerably fewer side effects than tricyclic antidepressants [92] but can decrease the concentration of a variety of other drugs by enzyme induction [93]. Several reviews summarizing side effects and interactions have been published [94-98].

In conclusion, the systematic reviews collected for this analysis are a good tool to get an overview of the available evidence from clinical trials in the area of herbal medicine. However, applying the findings to patients care is problematic for those who are not experts in herbal medicine. In this case it might be better to directly search the literature for clinical trials of the respective product.

Competing interest

KL, DM, GtR, and AV have been involved in some of the reviews analyzed. These were extracted and assessed by other members of the team.

Pre-publication history

The pre-publication history for this paper can be accessed here: