The authors of this chapter have been involved in research on various aspects of homeopathy for many years: two of us (PF, WBJ) are doctors who practise homeopathy. This chapter aims to describe homeopathy and the scientific issues it raises, to point out some paths that have been explored and abandoned as blind alleys and to highlight others that seem more promising. We do not attempt to cover all primary studies in depth. We believe we have referred to all systematic reviews and meta-analyses published, and some landmark studies, particularly those that have been replicated. The references to primary research are not comprehensive, but intended to give a flavour of the work. In view of the importance of a range of corroborative evidence, we will cover clinical research, biological models and basic research.

Homeopathy (also spelt homoeopathy and homœopathy; Greek homoios = same or similar, pathos = suffering) is a system of medicine based on the idea of ‘let like be cured by like’, in Latin: similia similibus curentur. It was founded by the German physician Samuel Christian Hahnemann (1755–1843) in 1796 with his seminal ‘Essay on a new principle for ascertaining the curative powers of drugs’ with a few glances at those hitherto employed (Hahnemann 1796).

Coincidentally, this was the same year that Edward Jenner first vaccinated against smallpox. It also coincided with the period of so-called ‘heroic medicine’, which advocated treatments including bleeding and purging in ‘heroic’ doses – as large as the patient could tolerate and not infrequently larger, resulting in the death of the patient. Heroic medicine originated in the work of John Brown of Edinburgh (1735–1788), but had many influential advocates, including Benjamin Rush (1745–1813), a signatory of the American Declaration of Independence. It is notable that, although heroic medicine was briefly influential and widely accepted, it was rapidly abandoned, whereas homeopathy is still alive and well. Some historians of medicine attribute this to the fact that homeopathy was based on empirical observation, whereas heroic medicine was based on a purely theoretical model (Schwanitz 1983). A reaction against heroic medicine may, in part, explain the adoption by homeopaths of increasingly small doses.

Hahnemann delineated two main approaches to medical treatment: the homeopathic and the allopathic or enantiopathic method (based on ‘contraria contrariis’: opposites oppose). It was Hahnemann who coined the words ‘homeopathy’ and ‘allopathy’. The latter term is now sometimes, incorrectly, used to refer to all conventional medicine.

References to both these methods of treating disease can be found in the Hippocratic corpus, for instance:

Homeopathy rapidly gained popularity in Europe in the first half of the 19th century. This seems to have been due to two main factors: its success in some high-profile cases and its success in epidemics (Jütte 1998). For instance, Radetzky, Field Marshal of the Austro-Hungarian Empire, declared ‘the disease of my eye in 1841 was cured exclusively by the homoeopathic treatment of my staff physician, Dr Hartung’, while famous violinist Paganini and the great naturalist Charles Darwin, among other prominent individuals, spoke highly of homeopathy. Darwin was initially very sceptical about high dilutions, but derived great benefit from treatment by Dr James Gully, a homeopath and hydrotherapist. Subsequently Darwin conducted experiments on the effects of highly dilute Ammonium phosphate on the carnivorous plant Drosera (Ullman 2009). (For an account of well-known users of homeopathy, see Ullman 2007).

The other factor in the rise of homeopathy was its success in the epidemics, particularly of cholera, which swept Europe during the 19th century. Homeopaths were much more successful in saving lives than their conventional counterparts. While in some conventional hospitals up to 74% of patients died, in homeopathic settings the figure was 4–11% (Glaz 1991). Homeopaths and the public usually attributed these successes to homeopathic treatment but historical analyses and modern data raise the question of whether the difference in outcomes (of which there is little doubt) was due to homeopaths curing their patients, or conventional physicians killing theirs! The mainstay of conventional treatment of cholera at that time was bloodletting which, as is now recognized, is just about the worst thing to do in a condition characterized by hypovolaemic shock. They also made frequent use of toxic medicines such as the mercury salt calomel but did not use rehydration (Leary 1994). A modern study of homeopathy in cholera showed no difference in death rates between homeopathy and placebo, because there were no deaths (Gaucher et al. 1994).

There are several distinct types of homeopathy. The main types are individualized or classical homeopathy, clinical homeopathy and isopathy. In individualized homeopathy typically a single homeopathic medicine is selected on the basis of the total ‘symptom picture’ displayed by a patient, including mental, general and constitutional features. In clinical homeopathy, one or more homeopathic medicines are administered for standard clinical situations or conventional diagnoses – sometimes several homeopathic medicines are combined in a fixed (complex) formulation. Isopathy is the use of homeopathic dilutions of allergens or causative infectious or toxic agents. Related medical systems which use homeopathic medicines include Homotoxicology, founded by HH Reckeweg and based on interpreting disease as an expression of the defensive effort of the organism against pathogenic toxins and detoxification with homeopathic medicines, and Anthroposophical Medicine, an approach founded by R Steiner and I Wegmann, integrating conventional medicine with the influence of soul and spirit.

Homeopathy is one of the most controversial forms of complementary and alternative medicine. Throughout its history of over 200 years it has been the focus of debate, often heated, with strong opinions expressed for and against. As long ago as 1846 John Forbes denounced homeopathy as ‘ludicrously absurd’ and an ‘outrage to human reason’ and any effect of homeopathy therefore impossible. More recently it has been claimed that ‘Accepting that infinite dilutions work would subvert more than conventional medicine; it wrecks a whole edifice of chemistry and physics’ (Vandenbroucke & de Craen 2001). Homeopathy has been described as ‘implausible’ and this has been used as a justification for scepticism concerning the claimed results of clinical trials of homeopathy (Kleijnen et al., 1991 and Shang et al., 2005). Commentators have discussed homeopathy in terms of the ‘growth of truth’, including the relationship between bias, background knowledge and the concordance of clinical results with laboratory science findings, concluding that ultimately ‘the proof of the pudding is in the eating’: the ultimate proof of the validity of a scientific or medical idea is the extent to which it changes reality (Vandenbroucke 2005).

Others, more favourably disposed towards homeopathy, have examined the scepticism surrounding homeopathy in Bayesian terms, pointing to the importance of Bayesian ‘prior beliefs’ based in specific paradigms and the need for a range of sources of evidence to change deep-rooted assumptions (Rutten 2008). Still others have contended that requiring greater burden of proof for homeopathy than for other scientific hypotheses is itself unscientific: the same level of supporting evidence should be accepted for all scientific developments. If a lower level of proof is set for hypotheses that fit prior beliefs, science is biased in favour of such beliefs (Chaplin 2007). But such a stance describes an ideal; in reality, all humans, including scientists, are to some extent, Bayesians: they hold firm beliefs. Their perceptions are coloured and structured by what they see and deem possible. Hence the dispute about homeopathy is likely to be solved not only by data, but also by a convincing model that can alter such broader scientific belief systems. At present, no such universally accepted model exists.

Despite the controversies which have long surrounded it, homeopathy is popular in diverse regions of the world, including the Indian subcontinent, parts of Latin America and western Europe. In other regions it has experienced seemingly terminal declines, for instance in the USA during most of the 20th century and Eastern Europe and Russia during the communist period, only to rise again at the end of the 20th century. And its use is growing in regions as different as South Africa and Japan. It is clear that what has sustained homeopathy above all is not scientific evidence, but the fact that it was, and remains, popular with patients (Dinges 2002).

Central to the controversy surrounding homeopathy is the lack of a ‘plausible’ mechanism of action for the very high dilutions often used in homeopathy. Hahnemann initially used doses similar or somewhat smaller than those used in contemporary conventional medicine, but he gradually reduced the size of his doses to include ‘ultramolecular dilutions’ which are the focus of most of the argument. These dilutions are prepared by a process known as potentization, which involves repeated dilutions, usually in steps of 1:10 or 1:100, with succussion (vigorous shaking) between each dilution. Dilutions are denoted for instance 5x (in the Anglo-American convention) or 5dH (European convention) for the 5th decimal (i.e. 5 times 1:10 dilutions) or 30c/cH for the 30th centesimal (1:100 dilutions). ‘H’ denotes the Hahnemannian method, in which clean glassware (test tubes and pipettes) is used for each step of dilution (German Homoeopathic Pharmacopoeia 1993). This remains the most widely used dilution scale, but other dilutional scales are also used. In the Korsakovian method (denoted K), a single piece of glassware is used; this is emptied and refilled, the liquid adhering to the walls of the vessel in which it is diluted is equivalent to one part, to which 99 parts of new liquid is added. Otherwise it is similar to the method described above (British Homoeopathic Pharmacopoeia 1993). In the 50 millesimal or quinquagintamillesimal dilution method (denoted LM or Q), dilution is in steps of approximately 1:50 000; this is achieved by absorbing the dilution on to small lactose granules, which are then redissolved (German Homoeopathic Pharmacopoeia 1993).

The use of such dilutions creates a fundamental problem: the Italian aristocrat-scientist Count Amedeo Avogadro proposed in 1811 that the volume of a gas is proportional to the number of atoms or molecules regardless of the nature of the gas. The value of Avogadro’s constant, the number of particles (atoms or molecules) in a mole of a pure substance was calculated by the Austrian scientist Johann Josef Loschmidt in 1865 as 6.022 × 10²³.

The inescapable implication is that dilutions above 23x/dH or 12c/cH (corresponding to dilutions of 10⁻²³ and 10⁻²⁴ respectively) are very unlikely to contain a molecule of the starting substance. Homeopathic medicines are used in both low dilutions, where the original substance is materially present, and in high dilutions, in which material quantities of the starting substance are much less likely to be present. In fact the ‘molecular threshold’ is usually crossed before this dilution, depending on factors including initial concentration and molecular weight. Classical pharmacological actions in vivo (‘classical pharmacological action’ is defined as interaction between pharmacologically active molecules and receptors) have been reported with dilutions as high as 10⁻²²mol/l and repeatedly with diilutions of 10⁻¹⁷–10⁻¹⁸ (Eskinazi 1999). By definition dilutions beyond the ‘molecular threshold’ cannot. Many homeopathic medicines are of biological origin, and consist of complex mixtures of molecules often of unknown molecular weight and concentration, and in any case, even for pure substances of known molecular weight and concentration, the probability of a molecule of the initial substance being present can only be expressed in terms of probability. Homeopathic medicines in which a molecule of the starting substance is unlikely to be present are referred to as ‘ultramolecular’ or ultra-low dilutions (ULD), or beyond the reciprocal of Avogadro’s number (BRAN).

It is important to remember that the first principle of homeopathy is that of similarity, and in this area there is substantial overlap with other areas of science, including hormesis (the stimulatory or beneficial effects of small doses of toxins) (Stebbing, 1982, Calabrese and Blain, 2005, Calabrese et al., 2006a and Calabrese et al., 2006b), rebound effects, dose-dependent reverse effects and paradoxical pharmacology (Bond, 2001, Teixeira, 2006 and Teixeira, 2007). The recent concept of ‘postconditioning hormesis’ refers to a small stimulus exerting a beneficial effect after a biological system has experienced a harmful stress of similar nature (Calabrese et al. 2007). These phenomena have in common that they are secondary, reverse or paradoxical effects of drugs and toxins in living organisms as a function of dose or time.

In order to understand the effects of drugs on healthy humans, Hahnemann conducted volunteer experiments, and homeopathy is the first form of medicine that aspired to base itself purely on empirical clinical trial evidence. As Hahnemann himself acknowledged, he was not the first to propose testing drugs on healthy volunteers – this had been proposed by the Swiss naturalist Albrecht von Haller (1708–1777) – but Hahnemann was the first to propose the systematic use of medicines on such data.

Hahnemann conducted a total of 99 Prüfungen (tests) of a wide range of substances on healthy individuals, and published these in his Materia Medica Pura and Chronic Diseases. Prüfung is traditionally translated into English as ‘proving’, but the more recent term ‘homeopathic pathogenetic trial’ (HPT) is more apt.

In the early 19th century homeopathy diffused widely around the world: it was introduced to the USA by German immigrants, most notably Dr Constantine Hering (1800–1880), who settled in Philadelphia, PA, USA, in 1833. Dr John Martin Honigberger (1795–1869), physician to Maharaja Ranjit Singhji of the Punjab, was the first to practise homeopathy in India, in 1839.

The scientific issues raised by homeopathy can be classified as follows:

These questions are ordered approximately from basic science to pragmatic. The fundamental claim of homeopathy is question 4, but the controversial aspects are questions 1–3. Most of the recent debate about homeopathy has revolved around question 3, the issue of placebo, and we will address this question first.

Systematic reviews and meta-analyses of randomized controlled trials (RCTs) are considered the strongest form of research evidence, sometimes referred to as type 1 evidence. A number of systematic reviews and meta-analyses of homeopathy as a whole or for specific conditions or of particular homeopathic medicines have been published.

A review of clinical trials in homeopathy reported from 1975 to 2002 found 93 studies comparing homeopathy with placebo or other treatment (Mathie 2003). Positive effects of homeopathy were found in 50. The evidence favoured a positive treatment effect of homeopathy in: allergic rhinitis, childhood diarrhoea, fibromyalgia, influenza, pain, side-effects of radio-/chemotherapy, sprains and upper respiratory tract infection. Reviewing 12 systematic reviews of homeopathy for specific medical conditions, Jonas et al. (2003) reached similar conclusions: homeopathy may be effective for allergies, childhood diarrhoea, influenza and postoperative ileus, but not for treatment of migraine, delayed-onset muscle soreness or prevention of influenza.

Single RCTs of homeopathy have been conducted in clinical areas, including asthma (White et al. 2003), life-threatening sepsis (Frass et al. 2005) and stomatitis induced by cancer chemotherapy (Oberbaum et al. 2001), chronic fatigue syndrome (Weatherley-Jones et al. 2004), premenstrual syndrome (Yakir et al. 2001), postpartum bleeding (Oberbaum et al. 2005) and Arnica for various clinical conditions (Tveiten et al., 1998 and Stevinson et al., 2003). Most of these have yielded positive results.

In some clinical situations, both RCTs and clinical observational studies have been conducted, providing a fuller picture of the possible role of homeopathy. Such areas include upper respiratory tract and ear infections in children (de Lange de Klerk et al., 1994, Frei and Thurneysen, 2001a and Jacobs et al., 2001), attention deficit hyperactivity disorder (Frei and Thurneysen, 2001b and Frei et al., 2005) and homeopathy for symptoms related to cancer treatment (Balzarini et al., 2000, Thompson and Reilly, 2003 and Jacobs et al., 2005a).

The group led by David Reilly has published an impressive series of trials of isopathy for respiratory allergies, including hayfever, perennial rhinitis and allergy, including a meta-analysis (Reilly and Taylor, 1985, Reilly et al., 1986, Reilly et al., 1994 and Taylor et al., 2000). To date there has been no independent positive replication of these findings. Lewith et al. (2002) studied 242 patients with asthma who were sensitive to house dust mite; they were treated with three doses over 24 hours of house dust mite 30c and followed for 16 weeks. At the end of the study there was no statistically significant improvement in the treated group compared to placebo, though there was one clear difference between the groups: there was significantly more variation in the verum group than in the placebo group (Hyland & Lewith 2002). There were important differences between the design of Lewith et al.’s trial and those of the Reilly group. There were also differences in the objectives: the Reilly group’s objective was to demonstrate efficacy compared to placebo, whereas Lewith’s study was oriented to showing efficacy in asthma. Lewith’s trial was multicentric, primary care-based, and of longer duration, with less rigorous diagnosis and entry criteria than the Reilly group’s work. However, comparable studies in Norway with potentized birch pollen in subjects who were allergic to birch pollen did not demonstrate clear-cut effects (Aabel, 2000 and Aabel, 2001).

Migraine is another area in which there have been replicated studies: a study on 60 patients by Brigo & Serpelloni (1991), using classical individualized homeopathy, was reported as showing a highly significant positive result. But attempted replications have been much less positive. The first, a close replication, showed weak effects approaching significance (Whitmarsh et al. 1997). In a subsequent study with individualized unrestricted homeopathy that included patients with all types of chronic headaches – not only migraine – no significant difference was found between homeopathy and placebo (Walach et al. 1997). Another study, conducted in Norway in 68 migraine patients, again used individualized homeopathy. Pain diaries showed no difference, but the neurologist rating was significantly in favour of homeopathy (Straumsheim et al 2000). However, an uncontrolled observational study reported major improvement in quality of life in patients undergoing homeopathic treatment for headache (Muscari-Tomaioli et al. 2001). Another observational study showed large effect sizes (d = 2.44) for improvement in headache days over the course of a year and in quality of life (Witt et al. 2009a).

Another model with independent replication is Arnica 30x to prevent delayed-onset muscle soreness. The results of two studies in the Oslo marathon have been pooled, and a small but significant effect on muscle soreness, but none on muscle enzymes, was shown (Tveiten & Bruset 2003). However, a larger-scale study replication, mostly on runners in the London marathon, was negative (Vickers et al. 1998). More recently the results of three linked studies of Arnica 30x in different types of knee surgery have yielded a positive result (Brinkaus et al. 2006). The models are so divergent, however, that no common quantitative measure can be derived.

Paralytic postoperative ileus has also been the subject of repeated studies: patients were treated with Opium and Raphanus C6; the time to resumption of intestinal transit was the outcome. Initial pilot studies were promising, but a large-scale multicentre study in France did not reproduce the initial results (GRECHO 1989). The initial results were so strong, however, that a pooled analysis still yielded significant results (Barnes et al. 1997).

Another area with mixed replications is rheumatology. An initial study of individualized homeopathy in rheumatoid arthritis was positive (Gibson et al. 1980), but independent studies failed to confirm these results (Andrade et al., 1991 and Fisher and Scott, 2001). A trial of a single homeopathic medicine (Rhus toxicodendron) in osteoarthritis with little or no individualization yielded a negative result (Shipley et al. 1983). However a replication, using the same homeopathic medicine but in a different rheumatological condition (fibromyalgia) and incorporating individualization, yielded a positive result (Fisher et al. 1989). Subsequently there have been two further positive results from RCTs of homeopathic treatment of fibromyalgia, but these have used different methods and outcomes: one looked at individualized homeopathy (Bell et al. 2004) and the other at the entire ‘package’ of homeopathic care, including consultation and medicine, compared to normal care (Relton et al. 2009).

In another area the apparently contradictory results of replications may be explicable by crucial methodological differences. Two placebo-controlled RCTs of homeopathy for attention disorder hyperactivity disorder were published at about the same time and were superficially similar: the patients recruited were similar and individualized homeopathy was used in both. However, closer scrutiny reveals that there were important differences in prescribing between the two studies: only one of the five most frequently prescribed homeopathic medicines was the same in the two studies (Frei et al., 2005 and Jacobs et al., 2005b). But, importantly, one of these studies (Frei et al. 2005) was of unusual design: it was in fact a study of withdrawal of treatment. The optimum treatment was determined, then after improvement had occurred, randomly and blindly replaced with placebo for 6-week periods, then resumed. The patients deteriorated when on placebo. The crucial point is that it took a mean of three attempts to find the effective homeopathic medicine. It has been shown that if this study had randomized the patients at the time of the first homeopathic prescription, its results too would have been negative (Frei et al. 2007).

It has been contended that the lack of effect in some trials is due to confounding effects – toothpaste, coffee, diet and other factors believed to antidote homeopathy – or due to patients being unresponsive because of conventional drugs. An attempt was made to conduct a study of untreated patients only, but proved impractical largely because of recruitment problems (Siebenwirth et al. 2009). Another study, in depression, failed because of recruitment problems, reflecting strong patient preference issues related to homeopathy (Katz et al. 2005).

Reviews of RCTs of homeopathy in the following conditions are broadly positive: childhood diarrhoea, influenza (treatment), postoperative ileus, seasonal allergic rhinitis, rheumatic diseases, upper respiratory tract diseases and vertigo. There is replicated RCT evidence that homeopathy may be effective in childhood diarrhoea, fibromyalgia, influenza (treatment), otitis media, seasonal allergic rhinitis, sinusitis and vertigo. There is also some evidence from individual RCTs that homeopathy may be effective in chronic fatigue syndrome, premenstrual syndrome, postpartum bleeding, sepsis and stomatitis.

For anxiety, childhood asthma, insect bites, menopausal symptoms in breast cancer, migraine, muscle soreness, prevention of upper respiratory tract infection, stroke and warts, the current RCT evidence is inconclusive or negative. For other medical conditions the current published evidence is fragmentary.

The evidence overall becomes less positive when filtered for internal trial validity and sample size. It can nevertheless be counterargued that many of these studies were of poor homeopathic quality, for instance lacking individualization or with unrealistic outcomes.

This debate also raises a fundamental methodological issue, concerning an assumption which underpins placebo-controlled studies: that specific and non-specific effects are separable and do not interact and that the total treatment effect is simply the sum of the two components. The effects of a treatment are traditionally divided into specific and non-specific: non-specific effects include explanation, reassurance, advice on diet, exercise and lifestyle. Placebo effects are a subset of non-specific effects: the psychological effects associated with an intervention, which include expectation of benefit, in part a response conditioned by previous experience of treatment, the patient’s perception of the therapist and the setting, credibility of the therapy and others.

There is no a priori reason to suppose that specific and non-specific effects are separable and do not interact. Given the importance of this assumption, it is surprising how little it has been questioned or investigated. When it has been investigated the answer has usually been that this assumption is unsafe. The most comprehensive review of the topic concluded that ‘specific and non-specific effects are sometimes synergistic, and at other times antagonistic, so the implicit model of the randomized clinical trial is too simple’ (Kleinjen et al. 1994). For instance, a trial of analgesia for post dental surgery pain compared intravenous infusion given by a person at the bedside, by a person hidden in an adjacent room or by a preprogrammed syringe pump. The effect differed not only in scale but in sign: the patients who received the active drug given by a person, hidden or not, had more pain than the placebo group; the reverse was true when the drug was given by a pump (Levine & Gordon 1984).

An attempt to investigate interaction between specific and non-specific effects in homeopathy was confounded by differential recruitment between groups (Fisher et al. 2006). However, a subsequent RCT, also of double-dummy design but not placebo-controlled, demonstrated equivalence of individualized homeopathy to the antidepressant fluoxetine (Adler et al. 2009).

Clinical research in homeopathy provides a good example of why research in CAM needs to be multimodal, matched to specific utility benefits and not hierarchical, as described in Chapter 1. The impossibility of proving whether homeopathy in general works using heterogeneous studies was demonstrated statistically in a funnel plot-based sensitivity analysis by Linde et al. (2001). They demonstrated that over 1000 adequately powered placebo-controlled RCTs would be needed. The conclusion is that very meticulous basic research or whole-systems comparative trials should be the main focus of homeopathic research.

A strong case can be made that if the research question is, ‘Do homeopathic medicines exert physiological effects?’ then it is much better to use laboratory experiments. RCTs are a clumsy, expensive and difficult-to-reproduce way to answer the question. If the question is, ‘Does homeopathy benefit patients?’ or ‘How much?’ or ‘At what cost?’ or ‘Is it safe?’, then clinical studies are required. But they should be of pragmatic, not placebo-controlled, design.

Cost-effectiveness studies in the public and insured primary care sectors in France and Germany suggest that integration of homeopathy is associated with better outcomes for equivalent costs. These studies were of ‘quasiexperimental’ design (i.e. different treatments were compared, but patients were not assigned randomly to the different treatments). A third, more recent, study was an RCT – the only one of its kind in homeopathy research to date.

Trichard et al., 2005 and Trichard and Chaufferin, 2004 compared two treatment approaches (‘homeopathic strategy’ versus ‘antibiotic strategy’) used in routine medical practice by allopathic and homeopathic GPs in the management of recurrent acute rhinopharyngitis in 499 children aged between 18 months and 4 years. The GPs using homeopathy had significantly better results in terms of clinical effectiveness, complications, parents’ quality of life and time lost from work, for lower cost to social security. A study based in a public sector clinic in Italy and using data from the official Prescription Archive documented a large reduction in drug costs for patients receiving homeopathic treatment compared to patients who did not (Rossi et al. 2009a).

Witt and colleagues (Witt et al., 2005a and Witt et al., 2005b) compared homeopathic and conventional GPs’ outcomes in chronic diagnoses commonly treated in general practice (adults – headache, low-back pain, depression, insomnia, sinusitis; children – atopic asthma, dermatitis, rhinitis). A total of 493 patients were treated by 101 homeopathic and 59 conventional GPs. The patients treated by the two groups of GPs were generally similar. The conclusion was that patients who sought homeopathic treatment had better outcomes for similar cost. Some of the same authors conducted a 12-month cohort study comparing homeopathic and conventional treatment of eczema in children. The two groups had similar improvements in perception of eczema symptoms (assessed by patients or parents) and disease-related quality of life (Keil et al. 2006). Also in the German context, the cost-effectiveness of the commercial homeopathic preparation Sinfrontal was assessed in a placebo-controlled RCT of adults with acute maxillary sinusitis (Kneis & Gandjour 2009). Sinfrontal led to average incremental savings of 275 euros per patient compared with placebo over 22 days; this was mainly due to reduced absenteeism from work (7.8 versus 12.9 workdays). In an indirect comparison, Sinfrontal had a higher cure rate than antibiotics, at similar or lower cost.

The clinical areas in which research on homeopathy has been done do not match well with those for which it is used in practice. For instance, homeopathy is commonly used to treat psychiatric problems, including anxiety and depression in France, Germany, the UK and USA (Davidson et al., 1998, Unutzer et al., 2000, Trichard et al., 2003 and Becker-Witt et al., 2004), yet as mentioned above, clinical trial evidence is scant. Reasons for this include model simplicity (for instance, isopathic treatment of allergies has been relatively extensively investigated because it is a simple model), expediency and commercial motives. In practice homeopathy is frequently used for clinical problems, for instance dermatological or gastrointestinal conditions where there is little or no RCT evidence. But observational studies look at what happens to patients who receive homeopathic treatment. A comprehensive observational study at the Bristol Homoeopathic Hospital included over 6500 consecutive patients with over 23 000 attendances in a 6-year period (Spence et al. 2005). At follow-up, 70% of patients reported improved health, 50% major improvement. The best treatment responses were reported in childhood eczema or asthma, and in inflammatory bowel disease, irritable bowel syndrome, menopausal problems and migraine. A study at a public-sector clinic in Italy yielded similar results (Rossi et al. 2009b).

In a prospective, multicentre cohort study in Germany and Switzerland, 73% of 3709 patients with 8-year follow-up contributed data. The most frequent diagnoses were allergic rhinitis and headache in adults, and atopic dermatitis and multiple recurrent infections in children. Disease severity decreased significantly (P < 0.001) between baseline, 2 years and 8 years. Younger age, female gender and more severe disease at baseline correlated with better outcomes (Witt et al. 2008). An observational study by the same group followed 82 adults with psoriasis for 2 years. Mean disease duration was 15 years; 96% had had previous treatment. Their psoriasis improved with a large effect size and conventional treatment and health service use were reduced (Witt et al. 2009b).

A 500-patient survey at the Royal London Homoeopathic Hospital showed that many patients were able to reduce or stop conventional medication following homeopathic treatment (Sharples et al. 2003). The size of the effect varied between diagnoses: for skin complaints, for example, 72% of patients reported being able to stop or reduce their conventional medication; there was no reduction for cancer patients. In both these surveys, many of the patients were suffering from difficult-to-treat ‘effectiveness gap’ conditions. (Effectiveness gaps are clinical areas in which available treatments are not fully effective or satisfactory for any reason, including simple lack of effective treatment, adverse effects, acceptability to patients and cost: Fisher et al. 2004.)

Some observational studies have also addressed cost issues. For example, Frenkel & Hermoni (2002) reported that homeopathic intervention led to modest economic savings and modest reductions in the use of medications commonly used to treat allergic conditions and their complications. One homeopathic GP practice in London recorded costs of homeopathic medicines and calculated the costs of conventional drugs that would otherwise have been prescribed for 100 patients (Jain 2003). Average cost savings were £60 per patient. The majority of patients’ health improved and most did not report any side-effects.

Although these observational studies showed significant and persistent outcomes, it is important to remember that the extent to which the observed effects are due to differential dropout, spontaneous improvement, regression to the mean, lifestyle changes, placebo or context effects is unknown and needs clarification in future explanatory studies.

The available evidence suggests that patients’ confidence in the safety of homeopathy is justified: the hazards from homeopathic products are modest in comparison with those of conventional medicine (Kirby 2002). A systematic review of the safety of homeopathy, including a search of the English-language literature between 1970 and 1995, came to the following conclusions: homeopathic medicines may provoke adverse effects, but these are generally mild and transient; adverse effects of homeopathy are underreported; there are cases of ‘mistaken identity’ where herbal medicines were described as homeopathic (Dantas & Rampes 2000). The main risks associated with homeopathy are indirect, relating to the prescriber rather than the medicine (Fisher et al. 2002). In two studies, adverse reactions were observed in approximately 2.7% of patients (Anelli et al., 2002 and Endrizzi et al., 2005); in a third study, 7.8% of homeopathy patients had adverse reactions, compared to 22.3% in the corresponding group receiving conventional treatment (Riley et al. 2001). In addition, laboratory research has reported adverse effects in animals so some caution is needed in assuming all homeopathy is safe (Jonas & Ernst 1999).

A temporary deterioration on starting homeopathic treatment associated with a good long-term prognosis is widely described in the homeopathic literature – homeopathic ‘aggravation’ (or healing reaction) (Vithoulkas, 1986 and Kent, 1990). A systematic review of double-blind randomized trials of homeopathy found no clear evidence of their occurrence (Grabia & Ernst 2003), but case series have reported rates of aggravation ranging from 2.68% to 24% (Thompson et al., 2004 and Endrizzi et al., 2005).

Eliciting symptoms with homeopathic reminders by giving them to healthy people and doing detailed analysis as the method called ‘proving’ was used by Hahnemann and others as a basis for seeing symptom pictures for certain remedies. They are now usually called ‘human pathogenetic trials’ or HPTs.

A systematic review of the modern literature identified 156 HPTs on 143 medicines, involving nearly 3000 volunteers, conducted worldwide between 1945 and 1995. The review concluded that there was a wide variation in methods and results, and that methodological quality was generally low (Dantas et al. 2007). Design flaws, particularly absence of proper randomization, blinding, placebo control and criteria for analysis of outcomes, were common. More symptoms were reported from HPTs of poor quality than from better ones. In addition considerable numbers of HPTs have been published outside the peer-reviewed literature, of substances as diverse as plutonium, chocolate, Coca-Cola and American bald eagle. These substances often seem to have been selected on symbolical or metaphorical grounds (Sankaran, 1998, Sherr, 1998 and Sherr, 2002). They often report large numbers of symptoms. The methodology is frequently not that of contemporary science, and we will not consider them further.

It has been objected that using RCT methods and excessive concern for statistical significance or the elimination of background noise risk ‘throwing the baby out with the bathwater’ (Sherr & Quirk 2007). And that modern provings unconsciously confound two different objectives: qualitative research and attributional clinical research. Qualitative research aims at maximizing variability to bring out the rich subtleties of the remedy administration and its experience, looking for striking, repeated (within individuals) and similar symptoms (between individuals). The validity of the symptom picture is later determined in clinical practice, not necessarily in other provers (Riley 2005). In attributional research the goal is to determine statistically if there are quantitatively different symptom counts between groups on the remedy and controls. A number of HPTs have been published subsequent to the cut-off date of the systematic review, with mixed results. Some of these have focused on the quality or ‘typicality’ of symptoms reported by volunteers, rather than their number, while preserving methodological rigour, and these do seem more likely to yield positive results (Signorini et al., 2005, Walach et al., 2008 and Möllinger et al., 2009) than those which have focused only on symptom count (Walach et al., 2001, Brien et al., 2003 and Teut et al., 2008).

Although the basic idea of homeopathy is similarity, its most controversial claim concerns the properties of ultramolecular dilutions. Avogadro’s constant, the number of particles (atoms or molecules) in a gram mole of a substance, is of the order of 10²³. In homeopathic terminology, 10²³ corresponds to a 23X or 12C dilution. Homeopathic preparations in dilution less than those contain material traces of the original substance; those in high (ultramolecular) dilution are unlikely to do so. We will divide our discussion into biological models, physicochemical research and theoretical work on the basis of homeopathic and ultramolecular effects.

The largest body of research on biological models in homeopathy is based on experimental intoxication. A critical review and meta-analysis focused on 135 experiments published in 105 articles exploring the protective effect of homeopathic dilutions against toxins (Linde et al. 1994). The studies were extremely diverse and included many different experimental models: more than 70% of the high-quality studies reported positive effects. A more recent meta-analysis evaluated 67 in vitro biological experiments in 75 research publications and found high-potency effects were reported in nearly 75% of all replicated studies; however, no positive result was stable enough to be reproduced by all investigators. The most frequently used in vitro model was basophils, used in 42% of experiments (Witt et al. 2007).

A review of replicated basic studies in homeopathy identified 21 different models that have been repeated. The earliest experiments are those of Kolisko on the effect of homeopathically prepared silver nitrate on wheat seedlings, dating from 1923. This model has been the subject of four independent repetitions, three of them positive. The earliest in vitro model, first reported from the Soviet Union in 1932, on the effect of mercuric chloride on starch diastase, was not reproducible in the most recent repetition (Persson, 1933, Boyd, 1954 and Witt et al., 2006).

Another relatively well-established plant model is the effect of the plant hormone gibberellic acid on plant growth. Again, repetitions have been mostly positive, although sometimes with qualitative differences between the results obtained by different groups. The problem of differing results between groups is mostly clearly seen with the wheat seedling/arsenic intoxication model (Brizzi et al. 2005). Multiple experiments, some of them with large datasets, have been conducted; most have shown statistically significant differences (Betti et al. 1997, 2005; Binder et al., 2005 and Scherr et al., 2009