The increasing popularity of acupuncture in the west has led to growing demands for evidence of effectiveness and safety.^1,2 There is a growing literature on adverse events associated with acupuncture, including more serious events such as pneumothorax and hepatitis.^3,4 This literature has mostly consisted of case reports and retrospective surveys of practitioners. Better evidence has come recently from two prospective studies of adverse events reported by practitioners,^5,6 supporting the assertion that “acupuncture is safe in competent hands”.⁷ In these two surveys combined, over 60 000 consultations were monitored and 86 significant non-serious adverse events were reported—most commonly nausea, fainting, and dizziness.^5,6 However, schemes which rely on practitioners reporting suspected adverse reactions are known to suffer from widespread underreporting.⁸ A more active role for patients in identifying adverse events is increasingly being advocated on the basis that patients are likely to provide more accurate reporting as well as having a legitimate interest in their own safety.⁸

With 2 million acupuncture treatments being provided in the UK each year by acupuncturists outside the National Health Service (NHS),^9,10 safety has clearly become a public health issue.^3,4,11 In addition to questions about the safety of acupuncture treatment itself, concerns have been raised that “since acupuncture is used by some practitioners as a complete system of medicine, it may constitute a risk of delayed conventional diagnosis (or treatment)” because “a sinister underlying cause” might be missed.³ Furthermore, it has been suggested that “practitioners have the unfortunate habit of (changing prescribed medication) which could be associated with important risks”.³ It is important to examine the degree to which these speculated risks are demonstrated in practice.

In the UK there have been recent moves to establish statutory regulation of acupuncture.² Four self-regulated professional associations of acupuncture, including the British Acupuncture Council (BAcC), have come together under the aegis of the Department of Health and the Prince of Wales’ Foundation for Integrated Health and developed recommendations for the statutory regulation of acupuncture. Members of the BAcC have a minimum of 3 years of full time training or equivalent, have signed up to a code of practice with a requirement that only single use needles are used, and by and large practise outside the NHS. In the context of the recommendations for statutory self-regulation, this study sought to establish more clearly the evidence base for the safety of both acupuncture and acupuncturists.

Our primary aim was to establish prospectively the type and frequency of adverse events reported independently by patients over a 3 month period. The secondary aims were to identify the perceived adverse events associated with either advice about medication or about delayed conventional diagnosis or treatment.

METHODS

Recruitment

Over a 6 month period in 2002, all 1955 acupuncturists registered with the BAcC were invited to identify up to 60 consecutive patients aged over 18 and to give them a short questionnaire and consent form. Participating patients returned their baseline questionnaires and consent forms direct to the research centre and 3 months later they received postal questionnaires. We aimed to collect data from patients covering 30 000 consultations to facilitate comparisons with practitioners’ reports of adverse events.^5,6 The Northern & Yorkshire multicentre research ethics committee gave ethical approval for this study.

Data collection on adverse events

The 3 month questionnaires provided data on patients’ reporting on what they perceived to be adverse events, either caused directly by the acupuncture treatment or indirectly as a result of following the acupuncturist’s advice about medication or from delayed conventional diagnosis or treatment. For the purposes of this survey we did not define an adverse event but, instead, provided patients with a checklist of possible events. This and the overall questionnaire, while not formally validated, were developed from two practitioner surveys.^5,6 In contrast, “serious adverse events” were predefined as those resulting in admission to hospital or being permanently disabling or life threatening,^12,13 and patients ticked these categories if they believed them relevant. We also asked for details of adverse consequences resulting from both advice about medication and from delayed diagnosis or treatment. For patients who followed advice about medication we sent a short additional questionnaire asking about the type medication, whether the advice was to stop or reduce, and whether this advice had been discussed with their GP or specialist. With the patients’ permission, all serious events and all cases of delayed diagnosis and treatment were followed up with a short telephone interview, clarifying the details of the event with questions from a checklist.

Data analysis

The representativeness of the practitioners and patients was assessed from the baseline data to facilitate an estimate of the extent to which we could generalise the results. The reported type and frequency of adverse events were tabulated, calculating the absolute risk in terms of an adverse event rate per 1000 patients over 3 months with 95% confidence intervals. Relative risks were calculated as odds ratios (OR) of at least one adverse event occurring for key characteristics of practice using logistic regression, adjusted first for age and sex and then for relevant confounders simultaneously, thereby establishing evidence on risk factors.

RESULTS

Practitioners

1114 practitioners (52%) responded and 638 (33%) agreed to participate. The reasons given by practitioners who declined to participate were that they were too busy (n = 146), not in practice (n = 45), seeing too few patients (n = 35), retiring or moving away (n = 28), on holiday (n = 21), on a maternity break (n = 18), disapproving of the survey (n = 15), or other reasons (n = 60). Participating practitioners were significantly more likely to be female (64% v 58%; difference 6%, 95% CI 1 to 10; p = 0.015) and to have been in practice for longer (table 1).

Patients

9408 patients agreed to participate in the study and 6348 (67.5%, 95% CI 66.5 to 68.4) completed questionnaires at 3 months. Table 2 shows that patients who completed the 3 month follow up questionnaire were significantly more likely to be female and older, and significantly less likely to be seeing an acupuncturist for the first time, having the NHS pay for their treatment, or being treated by a practitioner with less than 2 years experience.

The following analysis will focus on patients who returned adverse event questionnaires at 3 months. For these patients, the main baseline problems or complaints were musculoskeletal (n = 2404, 38%) and psychological (n = 680, 11%). Only 276 (4.3%) reported that they were having their treatment paid for by the NHS and a large majority (n = 6034, 95.1%) were paying for treatment themselves. Regarding their pathway to acupuncture, 2535 (39.9%) reported that “no one” had recommended acupuncture, 2082 (32.8%) came via a family member, friend or colleague, and 614 (9.7%) came through an NHS referral or recommendation. At the outset, 707 (11.2%) had not had acupuncture before and 897 (14.1%) reported no contact with their GP or hospital specialist either before or during the 3 month monitoring period. 376 patients (5.9%) were treated by practitioners who had been in practice for less than 2 years. Patients in the achieved sample reported an average of 4.8 acupuncture visits over the 3 months, a total of 30 196 consultations for the sample as a whole.

Direct adverse events

682 patients (10.7%) reported a total of 1044 adverse events caused directly by the acupuncture treatment. The most common event reported was “severe tiredness and exhaustion”, followed by prolonged or unacceptable pain at the site of needling (table 3). Patients also spontaneously mentioned a number of adverse events under the category “other”, most commonly bruising at the needling site. Three events were deemed to be “serious” by three patients (0.5 per 1000 patients over 3 months, 95% CI 0.2 to 1.4), of which one was followed by admission to hospital and two were perceived as potentially life threatening; a summary of the details obtained from telephone interviews is given in box 1. In terms of practitioner behaviour or equipment that might have been problematic, 109 (10%) adverse events were deemed (post hoc) to be potentially avoidable—most commonly: needles being left in at the end of treatment (n = 55); moxibustion burns to the skin (n = 28); perceived problems with electro-acupuncture such as too strong a current (n = 15); and being left alone too long (n = 10). The adverse event rate reported by the patients in relation to the 30 196 consultations covered by this survey was 350 per 10 000 consultations (95% CI 330 to 370). Only 40 patients said they would not be willing to have acupuncture again.

In table 4 we present absolute and relative risks associated with acupuncture treatment. We also present the OR estimates of relative risk after adjustment for age, sex, and other covariates (number of treatments, whether patient had had acupuncture before, whether treatment was paid for by the NHS, whether patient had had contact with their GP or hospital specialist before or during the 3 month period, and practitioner’s duration of practice) from a logistic regression analysis. Patients not funded by the NHS were less likely to report adverse events (adjusted OR 0.59), as were those patients who had not had contact with their GP or hospital specialist before or during the 3 month period (OR 0.66). However, patients were more likely to report adverse events if they had not had acupuncture before (OR 1.48) or if they were consulting a practitioner with less than 2 years experience (OR 1.44).

Reports on advice about medication

199 of the 6348 patients (3%) reported that at some point over the 3 month period they were given advice from their acupuncturist to reduce or stop taking medication. On sending out an additional questionnaire to this group (but only to the 177 who had agreed to be contacted again), 99 responded. With additional information from four patients who were followed up by telephone, we were able to analyse further the data on advice about medication for 103 patients. For 36 patients in this subgroup (35%) the advice was to stop rather than reduce their medication. In addition, 60 patients (58%) also discussed this advice with their GP or hospital specialist. Six patients (1/1000 (95% CI 0 to 2) over 3 months) experienced adverse consequences as a result of advice about medication; none was serious. The adverse event data for the 103 patients are shown in table 5.

Reports of delayed diagnosis and treatment

Over the 3 month period 660 patients (11%) reported receiving advice from their acupuncturist to consult a GP or hospital specialist. Two patients (0.3/1000 over 3 months, 95% CI 0 to 1) reported delayed conventional treatment as a result of consulting an acupuncturist. For one patient the adverse consequence was the cost of ineffective acupuncture for her jaw pain and for the other a prolonged urinary tract infection (see box 2).

DISCUSSION

In this study we have established from patients the type and frequency of adverse events that they associated with their acupuncture treatment. Of the 6348 patients participating, 682 reported at least one adverse event over 3 months, a rate of 107 per 1000 patients (95% CI 100 to 115). The most common events reported were severe tiredness and exhaustion, pain at the site of needling, and headache. Only three of the reported adverse events could be classified as serious. Our secondary objective to establish whether patients are at risk because they are consulting acupuncturists was not supported by the evidence from this study. We received reports of adverse but non-serious consequences related to advice about conventional (prescribed) medication from only six patients and delayed conventional treatment from only two.

This survey has considerably strengthened the evidence for the safety of acupuncture in routine practice. Patients had little difficulty in responding to questions about their experiences of adverse events associated with the acupuncture they received during the preceding 3 months. The implications are relevant to policy developments in the field and the current moves in the UK towards statutory regulation of acupuncturists.

The patient data from this survey are different from those which would have been provided by their practitioners and, arguably, are more robust on the basis of evidence from a systematic review of prospective acupuncture safety studies.¹⁴ For example, the adverse event rate reported by the patients in this study (350 per 10 000 treatments (95% CI 330 to 370)) is considerably higher than the rate reported by acupuncturists in an earlier survey of practitioners from the same professional body (13 per 10 000 (95% CI 9 to 17)).⁵ While some of this disparity might be explained by different definitions of adverse events, other factors likely to play a part include under-reporting by practitioners, perhaps due to sensitivity to criticism, and over-reporting by patients due to the use of checklists with tick boxes. The events listed in our survey tool were usually given the descriptor “severe”. It is possible that patients would have distinguished between severe events and others given the choice, and may have over-reported severe events. This is supported by the finding that very few (only 40) patients regarded their experience of adverse events to be such that it would deter them from trying acupuncture again in the future.

There were 109 (10%) adverse events that might be considered potentially avoidable on the basis that they could only happen if acupuncture procedures were executed poorly. For most types of avoidable events, patients reported higher proportions than did acupuncturists in practitioner surveys.^5,6,14 It is unlikely that the “skin infections” were caused by poor practice as the mandatory requirement that practitioners use sterile disposable needles should preclude cross infection. Such skin reactions are more likely to be due either to a local allergic response to chrome or nickel in the needles or a systemic response to the treatment itself.

This survey provides some valuable lessons for practitioners and educationalists. Firstly, patients appear to be most at risk of tiredness and exhaustion after treatment, leading to concerns about car accidents on driving away afterwards. Practitioners need to alert patients to this risk before treatment and to monitor levels of wooziness or light headedness on leaving the treatment room and before driving home. Ironically, for the one patient in this survey who reported a car accident, the association with acupuncture is extremely unlikely as it occurred 2 days after treatment. Given that 5% of adverse events involved needles being left in by mistake, practitioners need to take extra care in counting needles in and out. With 3% of adverse events involving moxibustion burns, the need for practitioners to take extra care in this area is also indicated by, for example, adhering to current practice guidelines requiring acupuncturists always to remain in the treatment room while using moxibustion.

Concerns in the literature have been raised about the safety of patients treated by non-physician practitioners who provide acupuncture outside national health care institutions.^3,4,11 The findings of this study suggest that these issues are not a major cause for concern for patients consulting this group of professionally registered acupuncturists. Patients potentially at greater risk—namely, those consulting outside the NHS and those who had no contact with their GP or hospital specialist before or during the course of treatment—reported significantly lower rates of adverse events.

From the data on advice about prescribed medication and delayed diagnosis and treatment, the evidence does not lead to concerns about patient safety. Advice from GPs or specialists about taking conventional medication was to reduce rather than to stop in more than half the cases; the drugs most commonly involved were antidepressants, corticosteroids, non-steroidal anti-inflammatory drugs, and hormonal drugs. The adverse consequences experienced by the six patients were not serious. Particular concern would be raised if patients were stopping their asthma medication, but our evidence shows that none of the 11 patients who reported taking prednisolone or steroid inhalers were advised to discontinue. Reports of delayed treatment were also infrequent and involved only two patients, both of whom experienced non-serious consequences.

While the evidence from patients might be subject to less response bias than that of practitioners, a number of potential limitations to this study remain. The subjectivity of patients’ reports must be considered, given the possibility of over- or under-reporting and the potential for misattribution. The 3 month period for monitoring may have been too short for adverse events that are rare or take time to manifest. However, a period of more than 3 months may have increased recall bias, with some patients remembering adverse events less well if they happened early on in the monitoring period.

There is the possibility that non-serious adverse events are clustered within acupuncturists and within treatments rather than distributed randomly. Such clustering would increase the width of the confidence intervals but would not alter the point estimates. In a trade-off to maximise practitioner and patient participation, our design precluded the possibility of linking reported events to individual practitioners or treatments. This meant that we were not able to test for clustering effects.

We found evidence of a difference in the patient characteristics of responders and non-responders to the survey at 3 months. We also found some evidence that the characteristics of the acupuncturists who agreed to take part in the survey and allow access to their patients may differ from those who refused to be involved in the survey. These factors may limit the generalisability of our findings. However, the large sample size and reasonable response rate (67%) to the survey should mitigate against this possibility.

Future research could usefully explore in more depth why almost all patients (94%) who experienced an adverse event continued to be willing to have acupuncture. For some patients the adverse event was an aggravation of existing symptoms, something that has been called a “healing crisis” when followed by an overall improvement of symptoms.^9,15 In addition, qualitative research could usefully explore how patients weigh up the perceived cost of an adverse event with the perceived benefit of acupuncture. Given that patients may also be interested in choosing between acupuncture and conventional treatment, future research could also compare risk profiles between interventions for common conditions such as chronic pain.

CONCLUSION

Patient reports of serious adverse events associated with acupuncture were rare. Adverse events related to advice about medication and delayed conventional treatment were also infrequent. There was no evidence that patients seeking treatment outside the NHS were at greater risk. While acupuncture treatment is associated with a range of adverse effects, these are not sufficient to prevent most patients seeking further acupuncture. This prospective study provides strong evidence that acupuncture, when practised by regulated practitioners, is a safe intervention.