Essentials of Complementary and Alternative Medicine (June 1999)

DEFINITIONAL ISSUES
Many CAM practices are derived from the medical systems of non-Western cultures, or carry basic epistemological assumptions about the nature of man and reality 
that are quite different from conventional science and medicine. These issues must be considered to prevent misunderstanding that may arise regarding different 
values underlying these practices. Definitional issues can impact safety judgments in several areas, including decisions on the most preferred product; diagnostic and 
patient classifications; type of outcomes valued by the complementary medical system; the personal value placed by individuals on outcomes; and the explanatory 
model that drives the forgoing. (A full discussion of definitional issues and how they impact on safety in CAM is beyond the scope of this chapter, but they are 
discussed in each chapter in Part III of this book for particular CAM systems.)
INDIRECT RISKS
Contrary to popular belief, complementary therapies can be associated with serious health risks. In addition to the risks just described, there is the indirect risk that a 
complementary treatment with unproven therapeutic potential delays or replaces a more effective form of conventional therapy. This can occur if an alternative 
practitioner is overly optimistic about his or her diagnostic or therapeutic abilities or when a naive patient or headstrong parent of a sick child puts too much trust in 
the healing powers of nature. There are various case reports, both in the medical and judicial literature, to indicate that this is not just a theoretical concern (
48
, 
49
, 
50
, 
51
, 
52
, 
53
 and 
54
). For example, life-threatening and fatal ketoacidotic coma has been observed in patients with insulin-dependent diabetes following reduction or 
withdrawal of insulin treatment in favor of an ineffective alternative approach (
55
, 
56
 and 
57
). Of particular concern may be the phenomenon that non-Western 
patients can seek refuge in a traditional therapy of their homeland. According to Belgian researchers, for example, serious problems have arisen in Moroccan 
migrants with asthma or diabetes, who returned to their homeland for a holiday, where they swapped their Western medicines for local herbs (
58
, 
59
).
Scientific information about the extent of the indirect health risks of complementary medicine is still difficult to find. In a questionnaire study among Dutch family 
doctors about their personal experiences with negligent behavior by alternative practitioners in 1986, 120 respondents reported 10 cases of complications: 6 of the 
patients had received complementary treatment instead of conventional therapy, whereas the other 4 had discontinued conventional drug therapy on the advice of an 
alternative practitioner (
60
). In a systematic Swedish survey among 242 hospitals over the period 1984–1988, 233 hospitals reported a total of 123 detailed cases. In 
6 cases the patient had died and another 23 patients had needed intensive care to save their lives (
61
). Unfortunately, without a reliable denominator, it is impossible 
to assess the incidence of the reported cases. Moreover, one should measure doctor's delay not only among alternative practitioners, but also among conventional 
physicians to allow a fair comparison between both types of health care providers. More and better research in this area is warranted.
Another indirect risk of complementary medicines is that the effectiveness of a conventional medicine may be compromised by concurrent use of an alternative 
product. This point is well illustrated by two fairly recent reports on adverse interactions between traditional Indian medicines and conventional medicines:
The oleoresinous product gugulipid (derived from an ancient Indian medicinal plant,  Commiphora mukul) was found to reduce the bioavailability of certain 
synthetic agents (e.g., diazepam, propranolol) by one-third (
62
). For conventional drugs with a narrow therapeutic window or steep dose-response curve, a 
decrease of this magnitude is relevant.
Co-administration of phenytoin with an Ayurvedic syrup called Shankhapushpi (prepared from  Centella asiatica, Convolvulus pluricaulis, Nardostachys 
jatamansi, Nepeta elliptica, Nepeta hindostana, and Onosma bracteatum) was reported to result in reduced plasma levels of phenytoin and in loss of seizure 
control (
63
). Additional studies on the risk of such interactions between traditional medicines and conventional pharmaceuticals are needed.
ADVERSE EFFECTS: TYPES A, B, C, AND D REACTIONS
Limited experience with complementary medicine can help to identify adverse effects that can develop rapidly after the start of therapy in a high proportion of users. In 
the clinical pharmacological literature, such acute effects are known as  type A reactions (
64
). A classic example is the induction of anticholinergic symptoms, such as 
palpitations, dryness of the mouth, and dilation of the pupils, by herbal medicines rich in belladonna alkaloids. As 
Table 1
 illustrates (
65
), type A reactions are 
pharmacologically predictable and dose-dependent, which implies that they can be anticipated and that they could be prevented by dose reduction. Traditional 

experience can bring these dose-dependencies to light and it can also help to detect ways of processing to reduce the likelihood of acute problems.
Table 1. Effects of Atropine in Relation to Dosage
However, not all adverse reactions occur immediately after a therapy has been initiated. The importance of delayed reactions has been underlined by a recent 
retrospective study covering clinical safety trials with 27 different pharmaceuticals. Nine of these 27 drug compounds were associated with serious drug-related 
adverse events that first occurred during the second half of a 12-month testing period. For three of the compounds, these late discoveries were so serious that they 
eventually affected the final dose selected, the product labeling, or the target population (
66
). When reactions develop during chronic therapy in a pharmacologically 
predictable way, they are called  type C reactions (
64
). An herbal example is muscular weakness due to hypokalemia in long-term users of herbal anthranoid laxatives 
(
67
). Type C reactions can be anticipated, but only after they have been identified, and such an identification may be more difficult than with type A reactions.
It is also difficult for alternative practitioners and their clients to recognize  type B reactions. These reactions are not associated with the principal pharmacological 
properties of a product and they do not improve when the dose is reduced—the product has to be withdrawn completely. Type B reactions are often immunologically 
mediated but some have a non- immunological basis (e.g., genetic cause). Although type B reactions occur in only a minority of the users, they can be so severe that 
withdrawal of the responsible agent from general use is warranted (
64
, 
68
). Examples of type B reactions to complementary products are the hepatotoxic reactions 
that have been attributed in recent years to the wall germander ( Teucrium chamaedrys) (
69
), skullcap (Scutellaria or Teucrium sp.) (
70
, 
71
), and chaparral (Larrea 
tridentata) (
72
, 
73
). In chaparral, the hepatotoxic potential only became apparent after an estimated 500 million capsules had been used without concern over a 
20-year period (
74
). At present, 5% of presumed viral hepatitis cases are not confirmed on serological testing. To what extent complementary medicines play a role in 
such cases is currently unknown. Doctors should definitely keep this possibility in mind, however, when they examine patients with unexplained hepatic disease (
75
).
Finally, type D reactions may be readily overlooked. This category consists of certain delayed effects, such as teratogenicity and carcinogenicity (
64
). It has been 
shown, for example, that the amines of certain Nigerian medicinal plants can be converted to  N-nitroso carcinogens under simulated gastric conditions (
76
). A clinical 
example is the presence of aristolochic acids in various medicinal species in the genus of  Aristolochia. These acids are potent rodent carcinogens (
77
), and human 
cases of Aristolochia-associated malignancy have recently been described (
78
, 
79
). The use of complementary medicines during pregnancy and lactation is also of 
concern. When a herb has oxytocic properties (the capacity to cause contraction of the uterus), the risks of its unrestricted use during pregnancy will be readily 
discovered. However, when a sick baby is born, who will attribute the disease to maternal consumption of a complementary product many months before the baby's 
delivery? Herbal remedies containing pyrrolizidine alkaloids may have been used since prehistoric times, but the first case report about neonatal hepatotoxicity 
following the use of such a remedy during pregnancy did not appear until the late 1980s. There is a need for more and better information about the embryotoxic and 
fetotoxic risks of complementary medicines, not in the least because the use of herbal medicines during pregnancy is sometimes encouraged by uncritical publications 
(
80
).
LIMITATIONS OF TRADITIONAL EXPERIENCE
Alternative practitioners and their customers are likely to detect some types of adverse reactions to herbal medicines less readily than other types (e.g., type A). 
Recognition can be particularly difficult when the signs and symptoms are not unusual in the population and could thus also be ascribed to various causes. In other 
words, although long-standing experience may tell much about striking and predictable acute toxicity, it is a less reliable tool for the detection of reactions that occur 
uncommonly, develop very gradually, need a prolonged latency period, or that are inconspicuous (
80
). This phenomenon of unobtrusive problems remaining 
undetected can be denoted as the  Aje-Mutin trap. Aje-imutin is the native name used by the Nigerian Yoruba people for an African relative of the ink-cap mushroom. 
The literal translation of the term is “eat-without-drinking-alcohol” (
81
), which shows that the Yoruba have learned that ingestion of  Coprinus mushrooms can induce a 
disulfiram-like sensitivity to alcohol. Yet the same Yoruba employ herbal enemas to treat diarrhea and dysentery, apparently without realizing that this can exacerbate 
the dehydration produced by the diarrhea, thereby reducing (instead of increasing) the patient's chance of recovery (
82
). Another example in Africa of inconspicuous 
traditional toxicity is the risk that eye medicines damage the eye by a direct action of toxic substances introduced into the conjunctival sac, by the introduction of 
microorganisms leading to infection, by physical trauma resulting from the application, or indirectly by delaying the patient's presentation to a clinic for therapy. 
Epidemiological research has shown that 25% of the corneal ulcers and childhood blindness in rural Africa is associated with the instillation of traditional eye 
medicines (
83
, 
84
 and 
85
).
The risk that rare adverse reactions to complementary medicines remain unnoticed can also be illustrated by the statistical “rule of three,” which dictates that the 
number of studied subjects must be three times as high as the frequency of an adverse reaction to have a 95% chance that the reaction will actually occur in the 
studied population. When an adverse reaction to a medicine occurs with a clinically relevant frequency of 1 in 1000, a practitioner treating 1000 patients with this 
medicine still has a 37% chance that he or she will not observe the reaction at all. To be 95% certain that the practitioner will see the reaction, he or she would have 
to treat at least 3000 patients (
Table 2
). The practitioner may need to see more than one reaction, however, before he or she can make a mental connection with the 
medicine. To have a 95% chance that the clinician observes the reaction three times, he or she would have to treat 6500 patients—that is, one patient every working 
day for almost 25 years (
55
). These calculations make clear that personal experience is not a reliable basis for the exclusion of uncommon reactions to 
complementary medicines.
Table 2. Number of Persons who Need to be Exposed to a Drug to Have a 95% Chance of Detecting an Adverse Drug Reaction Occurring with a Particular 
Frequency at Least Once, Twice, or Three Times
Nontraditional Hazards
There is another reason why safety claims cannot always be based on long-standing traditional experience: not all complementary medicines have firm roots in 

traditional practices, and this issue seems underestimated. When traditional source plants are extracted in a nontraditional way (e.g., by resorting to a nonpolar 
solvent such as hexane), one can question whether this nontraditional extract is as safe as the traditional one. Until recently, the ostrich fern ( Matteuccia 
struthiopteris) was generally considered to be a nontoxic, edible plant with a history of use as a spring vegetable that dates to the 1700s. However, recent 
observations of serious gastrointestinal toxicity following the consumption of lightly sauteed or blanched ostrich fern shoots suggest that this vegetable is safe only 
when it is thoroughly cooked before use (
86
). A similar example is the recent outbreak of bronchiolitis obliterans in Taiwan, which was associated with the ingestion of 
Sauropus androgynus. This herb is normally cooked before being eaten as a vegetable, but in this case the numerous victims had all consumed uncooked leaf juice 
as an unproven method of weight control (
87
).
It is also possible that an ingredient may have no medicinal tradition at all, and its route of administration or dose level may be quite different from that used in a 
traditional setting. The question could be raised, for example, to what extent the excellent oral safety record of certain traditional herbs is applicable to herbal 
cigarettes, available in Western health food stores. Certain respiratory risks attributed to tobacco smoking may extend to the smoking of nontobacco herbal products, 
particularly marijuana (
88
, 
89
, 
90
, 
91
, 
92
 and 
93
).
PRODUCT-RELATED DETERMINANTS OF ADVERSE EFFECTS
Principal determinants of the toxic potential of complementary medicines are their composition and way of use. Various examples of potentially hazardous ingredients 
of complementary medicines are reviewed in detail later in this section. Unfortunately, checking out a product label is not always sufficient to exclude harmfulness. In 
many cases the adverse effects are associated with a hidden constituent or with a higher strength than that mentioned on the product label. For example, selenium 
toxicity has been repeatedly caused by health food tablets, which contained many times more selenium than was stated on their label (
94
, 
95
 and 
96
). In other words, 
the current adverse effects to complementary medicines can be caused by a lack of stringent quality assurance rather than to the toxicological spectrum of their 
declared ingredients.
TOOLS FOR SAFETY ASSESSMENT IN COMPLEMENTARY MEDICINE
Clinical Accuracy
Modern science has derived a number of methods for attempting to assess safety of conventional medical practices. Despite this safety net, it is not infrequent to find 
significant risks being uncovered after years of use (e.g., UV light treatments, anti-arrythmics, calcium channel blockers, prostate cancer surgery). For practical 
purposes, the methods needed to detect adverse events are the same for both indirect and long-term direct effects. Direct, short-term toxicological effects are much 
easier (depending upon severity) but in some cases can be exaggerated or underdetected without proper investigative design. First, there is often wide clinical 
disagreement about whether a particular adverse effect was caused by a therapy. Even pharmacologists, for example, disagree 36% of the time as to whether a 
particular adverse effect was caused by a drug. The rate of disagreement goes up the more serious the attribution, with a 50% disagreement on whether an admission 
to the hospital was due to an adverse drug reaction and a 71% disagreement on whether a death was due to an adverse drug reaction (
97
).
Methods of Measurement
Many methods of surveying for adverse effects —broad checklists, patient interviews, or questionnaires—often do not yield relevant associations. For example, 
symptoms attributable to adverse drug reactions using symptom checklists results in 81% of individuals checking “Yes,” with a mean of two symptoms per individual 
and 7% reporting six or more symptoms (
98
). Most symptoms collected this way will be false positive. The true rate of adverse events, even in extensively used 
therapies, may not be detected without rigorously designed, hypothesis driven, prospective trials with randomization and a control group.
Rare Events and Public Health Impact
Because adverse events occur rarely for many complementary medical practices (
34
), relatively large numbers (i.e., 3 × 1 over the inverse ratio of events) and special 
designs are needed to be 95% confident that even one adverse event would be detected. Adverse drug reporting can significantly underestimate these effects 
because only 5 to 10% of such events are ever reported. However, non-random cohort or case-control evaluations of adverse events often overinflate estimates of 
events with odd ratios of even two or three subsequently being found to be false (
99
). Because of their widespread use, complementary and alternative interventions 
may have significant public health implications. Accurately assessing the adverse effects of a single CAM therapy would require postmarketing surveillance of 
upwards of 3000 individuals. In addition, the rate of rare idiosyncratic and allergic reactions needs similar, large-scale postmarketing assessment (
100
).
Types of Evidence in Adverse Effects Assessment
Table 3
 lists various types of evidence often used for reporting on safety in medicine, its usefulness, and its major limitations. First, preclinical evidence can often give 
indications of areas where there is potential toxicity and guide hypothesis testing and mechanism studies. Such information may identify compounds of potentially 
high risk in humans that cannot be automatically inferred to be harmful in the context of their clinical use. For example, phenobarbital increases the rate of 
hepatocellular carcinoma in rats, but when prepared as a homeopathic dilution can result in decreased rates of such cancers (
101
).
Table 3. Tools for Assessment of Adverse Effects
Historical evidence is often used to indicate that natural products are self-evidently safe. It is useful for giving general indications about acute toxicity, but cannot be 
used to indicate potential chronic effects or indirect risks. Therefore, this type of evidence is not useful for making firm conclusions about the value of these products 
for chronic disease in practice (
102
).
Case reports in the literature are the most frequent method of illustrating and emphasizing potential toxicity from complementary treatments. Such reports on adverse 
effects have the same limitations as anecdotal reports of beneficial effects. Case reports allow for descriptions of possible adverse associations but cannot be used to 
make judgments about frequency or cause. These reports tell us that adverse events can occur, not that they must occur nor how frequently. As with anecdotal reports 
about benefit, they are likely to lead to an overinterpretation of the significance of such events. Lists of adverse reports have been collected for a number of practices 
(especially herbal practices, acupuncture, and megavitamin therapy) (
40
, 
103
). Most authors agree that many of the more established complementary, alternative, and 
traditional medical interventions do not produce many serious adverse reactions when used in the traditional and/or indicated manner. Postmarketing surveillance can 
provide information about prevalence, incidence, and associations related to these events. As discussed previously, however, postmarketing surveillance may require 
large numbers (at least 3000) to identify with any confidence the true adverse event rate. Surveys risk inflating adverse event rates because specifically searching for 
events among a population can lead to search intensity bias, even in case-control studies (
98
). All studies must be conducted using objective outcome measures in a 
way that provides equal opportunity for finding and extracting information from exposed and unexposed groups. Failure to use blind evaluations can lead to diagnostic 
suspicion bias (finding what you are looking for) which also can inflate and exaggerate the rate of these events (
104
).
Adverse effects-reporting registries, such as MedWatch used by the FDA in the United States, can provide an indication of the popularity of new therapies or about 

changes in opinion on the value of new therapies in practice. In addition, they can be used to identify new, unexpected problems, such as a sudden increase in 
reactions from treatment exposures. These registries, however, are notoriously subject to falsification and under-reporting. Without rigorous verification methods built 
in, they cannot be relied upon for accurate prevalence data.
Information on adverse reactions also comes from poison control centers. These data are valuable for the likelihood of adverse reactions that occur from misuse and 
abuse of therapies (e.g., overdosing, suicide attempts, fraudulent practices, accidents) (
105
). These sources of data cannot provide information about safety under 
conditions of appropriate therapeutic use, nor is the accuracy of such information high given the high rates of disagreement about true adverse drug effects previously 
discussed.
Phase I and Phase II controlled trials can begin to indicate adverse effects accurately. However, the frequent use of open-ended checklists of adverse effects 
increases the risk of missing the real effects secondary to the multiple outcome assessments. In addition, the small size of these trials and their usual short-term 
duration reduce the chance of detecting effects that are not frequent or that may be delayed.
Phase III, or true efficacy, trials allow us to make conclusions about cause and are likely to be accurate if properly conducted. However, unless the adverse effects 
themselves are hypothesis generated, real adverse effects may be obscured by the likelihood of obtaining false-positive associations from multiple outcome 
measures. In addition, many complementary and alternative practices may find adequate sham controls problematic. For example, sham acupuncture usually shows 
increased effects over no acupuncture but less effects than “real” acupuncture, indicating that both specific and nonspecific effects occur (
43
). Likewise, delivery of 
sham acupuncture cannot be done blind in a way that allows a complete approximation of how the therapy is delivered in clinical practice. Such trials may necessitate 
a pragmatic orientation that can never positively identify adverse effects from the specific therapy. If such effects are low to begin with, identifying these effects 
unequivocally may be impractical and unnecessary. The most valuable type of evidence would be hypothesis-generated toxicity studies done in randomized 
placebo-controlled fashion. This has the highest likelihood of revealing accurate adverse effects. However, because of its complexity, it is rarely done. Even 
hypothesized-generated, RCT toxicity studies, however, will not eliminate outcome substitution bias. This occurs when the more easily measured objective outcomes 
become the focus of the study, although less easily measured subjective outcomes are the most relevant for the patients involved. This is like the man looking for his 
lost car keys under a lamp where there is more light, though he lost them in the dark somewhere else.
Finally, none of these types of evidence adequately addresses the issue of model validity and the complexities that arise in attempting to assess optimal therapy 
(treatment of choice). This information comes best from direct randomized comparative trials that directly compare therapies and trials incorporating patient 
preferences into the investigation.