Objective Few studies have investigated the predictors of the specific and non-specific effects of acupuncture. The aim of this secondary analysis was to determine patient characteristics that may predict a better treatment response to acupuncture for insomnia.
Methods We pooled the data of three randomised, double-blind, placebo-controlled trials of acupuncture for insomnia to examine sociodemographic variables, clinical characteristics, baseline sleep-wake variables, and treatment expectancy in relation to acupuncture response. Subjects with an improvement in insomnia severity index (ISI) scores of ≥8 points from baseline to 1 week post-treatment were classified as responders. Factors were compared between responders and non-responders, and also by univariate and multivariate logistic regression analysis.
Results A total of 116 subjects who received traditional needle acupuncture were included, of which 37 (31.9%) were classified as responders. Acupuncture responders had a higher educational level (p<0.01) and higher baseline ISI score (p<0.05), compared to non-responders. In the multivariate logistic regression analysis, only the number of years spent in full-time education remained significant as a predictor of treatment response (OR 1.21, 95% CI 1.06 to 1.38, p<0.01).
Conclusions Consistent with previous studies, our data suggest that the response to acupuncture is difficult to predict. Although the predictive power of educational level is weak overall, our findings provide potentially valuable information that could be built upon in further research (including a larger sample size), and may help to inform patient selection for the treatment of chronic insomnia with acupuncture in the future.
Statistics from Altmetric.com
Acupuncture is commonly used as a complementary and alternative medical (CAM) therapy for insomnia.1 Previous systematic reviews have suggested that acupuncture is effective for insomnia, but with a small effect size and marked heterogeneity in treatment outcome,2 ,3 possibly due to variation in acupuncture protocols and sample characteristics. It is believed that acupuncture has a relatively small specific effect but a large non-specific effect;4 however, the underlying mechanisms remain poorly understood. Several physiological mechanisms, such as sympathetic inhibition and modulation of the opioid and melatonin systems, may be involved in the specific effects,5 ,6 while a combination of treatment expectancy, diagnostic procedures, patient–practitioner interaction, conditioning, intensive monitoring, anxiolysis, and reporting bias may account for the non-specific effects.7 ,8
Only a few studies have investigated the predictors of specific and non-specific effects of acupuncture. The number and frequency of treatment sessions, the number of acupuncture needles and sites of insertion, other variations in treatment protocol, and the characteristic sensations elicited by acupuncture, known as de qi, are believed to predict treatment outcome.9 Among the non-specific factors, patient expectation is probably the most frequently studied factor. A randomised controlled trial (RCT) of patients with knee osteoarthritis found that treatment expectancy, but not race, gender, or age, predicted acupuncture response.10 In a study of patients with low back pain, higher expectation predicted a greater response to both real and placebo acupuncture, while psychiatric comorbidity was not a significant predictor.11 In contrast, a secondary analysis of an RCT of acupuncture for chronic back pain found no significant association between pretreatment expectation or preference and treatment response.12 Another non-specific factor that may contribute to treatment response is regression to the mean. Patients with more severe illness usually show a greater response to acupuncture.12–14 By contrast, baseline characteristics that have been found to negatively predict acupuncture response include the use of analgesics,12 ,14 receipt of disability allowance,14 previous surgical treatment,14 and headache.14
All these studies have focused on patients with musculoskeletal pain, and none has addressed subjects with mental health problems. The limited number of publications available on this topic indicates that this is an under-studied field. Therefore, we conducted a post-hoc analysis of three randomised, double-blind, placebo-controlled trials of acupuncture for primary or residual insomnia in depressed patients. Our aim was to determine patient characteristics predictive of a better response to acupuncture treatment for insomnia.
We pooled the data of three RCTs that were previously conducted by our research team, all of which are registered at http://clinicaltrials.gov (#NCT00839592, #NCT00838994 and #NCT01707706). The rationale, design, and results of each RCT have been previously reported.15–17 In short, subjects were randomised to receive traditional acupuncture, minimal acupuncture or non-invasive placebo acupuncture in a 1:1:1 ratio, or traditional acupuncture versus placebo acupuncture in a 1:1 ratio. Two of the RCTs included subjects with residual insomnia associated with major depressive disorder,16 ,17 and the third RCT was on primary insomnia.15 Subjects were recruited from the community and at psychiatric outpatient clinics. This secondary analysis involved 116 subjects receiving traditional acupuncture. Detailed inclusion and exclusion criteria can be found at http://clinicaltrials.gov. Briefly, patients were of Chinese ethnicity, aged 18–70 years, and fulfilled the symptomatic criteria of insomnia and functional impairment for the diagnosis of primary insomnia according to the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision (DSM-IV-TR),18 having had insomnia of at least three nights per week for ≥3 months, an insomnia severity index (ISI) score ≥15, no specific sleep disorders (including circadian rhythm sleep-wake disorder, parasomnia, sleep apnoea or periodic limb movement disorder, as assessed by clinical interview or overnight polysomnography), and had not received acupuncture in the preceding 12 months. In the RCTs of residual insomnia associated with major depressive disorder, subjects also had to have a past history of major depression according to DSM-IV-TR and be in partial/full remission.
Study design and intervention
All study procedures were reviewed and approved by the local institutional review board (HKU/HA HKW IRB: UW05-353, UW07-304 and UW08-417). Subjects were screened by telephone, face-to-face interview, and laboratory-based overnight polysomnography. Eligible subjects completed a 7-day sleep-diary and a 3- or 7-day actigraphy recording 1 week before baseline. The acupuncture points used were the same in two RCTs,15 ,16 and included GV20 (Baihui), Yintang, bilateral ear Shenmen, Sishencong, and Anmian. In the third RCT, bilateral PC6 (Neiguan), HT7 (Shenmen), and SP6 (Sanyinjiao) were used in addition. These acupuncture points were selected based on our previous systematic reviews2 ,3 and expert opinion. De qi was achieved if possible; afterwards the needles were connected to an electrical stimulator (ITO ES160, Japan) to deliver electrical stimulation in the form of a continuous 0.4 ms square wave pulse at 4 Hz frequency and constant current up to the subjects' tolerance level. The needles were left for 30 min and then removed. Acupuncture treatment was performed by acupuncturists with at least 3 years of clinical experience. Subjects were treated three times per week for three consecutive weeks in a quiet treatment room.
Outcome measures included the ISI,19 which was used to assess the perceived severity of insomnia and accompanying functional impairments (scores ranging from 0 to 28), sleep diary and actigraphy, the 17-item Hamilton depression rating scale (HAMD; a clinician-administered assessment of depression and associated symptoms),20 hospital anxiety and depression scale (HADS; a self-rated 14-item scale designed to assess the severity of depressive and anxiety symptoms),21 and the credibility of the treatment rating scale (CTRS; a four-item scale that assesses a subjects' confidence that a treatment will alleviate their complaints, their confidence in recommending the treatment to friends with similar complaints, the perceived logic of the treatment, and likelihood that the treatment will alleviate their other complaints).22 The daily sleep diary23 captured bedtime and rising time, time in bed (TIB), sleep onset latency, waking up after sleep onset, number of awakenings, terminal wakefulness, and total sleep time (TST). Sleep efficiency (SE) was calculated using the following equation: SE=TST/TIB×100%. A wrist actigraph, a watch-like device incorporating an accelerometer-microprocessor link, was used to record individuals' physical movements and to provide a valid objective measure of sleep.24 Subjects were instructed to wear the actigraph on the non-dominant wrist every day for 3 days or 1 week. The recording length of each epoch was set at 1 min and the data were analysed using Action-W (Ambulatory Monitoring, Inc, Ardsley, New York, USA) or Actiware (Respironics Inc, Murrysville, Pennsylvania) software. The Chinese version of the various questionnaires and sleep diaries were used, as these have been shown to have adequate validity and reliability.25
Classification of responders/non-responders
Subjects with ISI scores that improved by ≥8 points between baseline and 1 week post-treatment were classified as responders. The ISI score was selected because it assesses severities of both insomnia symptoms and related daytime impairments. This ISI cut-off as a definition of response has been used in a previous study of cognitive behavioural therapy for insomnia, which found a mean post-treatment reduction in ISI score of 8.3, equivalent to a within-group effect size of 2.0.26 This ISI cut-off is able to detect an independent assessor-rated moderate improvement with 60% sensitivity and 70% specificity, and is considered the minimum clinically significant difference.27
Factors examined were sociodemographic variables, clinical characteristics, current use of hypnotics, previous experience of acupuncture, CTRS scores, and baseline sleep-wake variables including ISI. To minimise multiple comparisons, we only analysed TST and SE for the sleep-diary and actigraph variables. The independent samples t-test or χ2 test was used to compare baseline variables between acupuncture responders and non-responders. A univariate logistic regression analysis was conducted with response or non-response as the binary variable, and the potential factors as explanatory variables. We did not include a Bonferroni correction as it was considered to be overly conservative and would have limited the number of variables entered into the regression model. ORs and their 95% CI were calculated. Finally, all significant variables were included in a multivariate logistic regression analysis. All data are presented as mean±SD unless otherwise stated.
Effect size and response rate
From baseline to 1 week post-treatment, the average reduction in ISI score for the 116 participants was 5.4±4.9 points (within group effect size=1.34), while the average increase in sleep-diary-derived TST was 26.5±74.4 min (within group effect size=0.32), and the increase in SE was 7.3±13.0% (within group effect size=0.45). Thirty-seven of the 116 subjects (31.9%) were classified as responders. The average reduction in ISI score among the responders was 10.9±3.1 points; for non-responders, it was 2.8±3.0 points. Univariate analyses showed that acupuncture responders had a higher educational level (p<0.01) and higher baseline ISI score (p<0.05) compared to non-responders (table 1). All other sociodemographic and clinical variables appeared to be unrelated to acupuncture response.
Table 2 summarises the univariate and multivariate logistic regression analyses. After controlling for all significant factors, only educational level remained significant (OR 1.21, 95% CI 1.06 to 1.38; p<0.01). Acupuncture responders had a significantly higher educational level than non-responders.
Our secondary analysis of the pooled data of three RCTs showed that a higher educational level and greater baseline ISI score were predictive of a better response to acupuncture treatment for insomnia. However, when both variables were taken into account, only educational level remained a significant predictor. Consistent with the findings of previous studies of acupuncture for painful conditions, our preliminary data suggest that the response to acupuncture among subjects with chronic insomnia is difficult to predict.
One possible explanation for our findings is that higher education is associated with a greater willingness to participate in acupuncture trials. Accordingly, a positive attitude toward participation in an experiment, regardless of treatment expectancy, may enhance treatment response. A previous survey of breast cancer patients showed that those with a higher level of education were more willing to participate in an acupuncture trial than those with lower education.28 Cited reasons for declining to participate included concern about experimentation, presence of placebo, and a lack of interest in acupuncture. Consequently, sceptical attitudes toward experimentation may hamper treatment response. Another potential explanation is that subjects with higher educational level might more accurately report the de qi sensation; hence the beneficial effect of acupuncture may have been enhanced.
Previous research has explored the predictors of successful pharmacotherapy and psychotherapy in psychiatric conditions such as obsessive-compulsive disorder and panic disorder;29 ,30 however, no consistent predictors have been identified. This inability to predict treatment response may be particularly relevant for acupuncture, which is an intervention that includes significant non-specific effects. Therefore, it is perhaps unsurprising that only a few significant predictors for acupuncture response were found in this study.
Despite the availability of effective pharmacological31 ,32 and psychobehavioural33 ,34 treatments for insomnia, it is fairly common for people to seek CAM therapies to help them sleep.1 ,35 The growing popularity of CAM therapies, such as yoga, acupressure and herbal medicine, may reflect the relative unavailability of psychobehavioural treatments and the potential adverse events associated with pharmacotherapy.35 ,36 However, it remains unclear which patients are more likely to respond to a particular CAM therapy. Patients may shift from one CAM treatment to another when the response is unsatisfactory, which may delay the receipt of effective treatment. To our knowledge, there have been no previous studies examining the predictors of response to acupuncture in people with insomnia. Our study serves as a first step in identifying those patients whose insomnia is more likely to be effectively treated by acupuncture.
We recently completed a similar secondary analysis on the response to placebo acupuncture in subjects with chronic insomnia, in which we found that higher ISI scores, longer sleep-diary-derived TSTs, fewer discrepancies between sleep-diary and actigraphy-derived TSTs, and higher expectations of acupuncture were predictors of a greater placebo response.37 As real acupuncture exerts both specific and non-specific effects, while placebo acupuncture produces only non-specific effects, it is perhaps unsurprising that the predictors are different.
In the present study, we observed that treatment expectancy did not appear to be a predictor of acupuncture effectiveness in subjects with insomnia. This finding is in contrast to the literature on chronic pain. Except for one report,12 all other studies10 ,11 ,38 ,39 to date have shown that treatment expectancy is relevant to acupuncture response. The difference in results suggests that the relationship between treatment expectancy and outcome may vary across conditions that have different pathophysiological mechanisms. Further research is needed to understand the effects of pretreatment expectations on acupuncture effectiveness.
In terms of Traditional Chinese Medicine (TCM) theory, the de qi sensation is an essential part of acupuncture and is related to efficacy.9 It has been postulated that the characteristics of de qi, including intensity, duration and propagation, may have an impact on the acupuncture response.40 In our RCTs, although the de qi sensation was elicited whenever possible, we did not formally assess its occurrence. In the future, it would be helpful to use a validated assessment tool to quantify the de qi sensation in order to study its effects on treatment response.
There are several limitations to our study. Firstly, the acupuncture protocol was standardised instead of individualised, which differs from usual clinical practice from a TCM perspective. Nevertheless, the acupuncture points used in our RCTs were all commonly used acupuncture points for insomnia, suggesting that our protocols were probably similar to the clinical practice of most TCM practitioners. Secondly, despite the pooling of three RCTs, the sample size remained small and may have lacked the necessary statistical power to reveal other potential predictors. Thirdly, although the ISI cut-off of ≥8 has been used to define treatment response in a previous study,26 it has not been specifically validated in the Chinese population. Finally, we were unable to assess some other potentially important predictors of acupuncture response, such as patient–practitioner interaction.
In conclusion, the number of years spent in full-time education was shown to be an independent predictor of the response to acupuncture in subjects with chronic insomnia. Although the overall predictive power was weak, these preliminary findings may help to identify which patients with insomnia are more likely to be helped by acupuncture, which could be useful for pre-treatment patient selection. Future studies with larger sample sizes are ultimately required to gain a better understanding of this topic in order to have any significant clinical impact.
Contributors W-FY and KF-C conceived the study. YMBY acquired the data. W-FY and YMBY analysed and interpreted the data. W-FY and KF-C drafted the paper and LL revised it critically for important intellectual content. All authors approved the final version accepted for publication.
Funding One of the primary studies (ClinicalTrials.gov identifier: #NCT01707706) was funded by the Health and Health Services Research Fund, Food and Health Bureau, Hong Kong SAR (Project no. 08091101); LL's research team is supported by Vivian Taam Wang Professorship in Integrative Medicine.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Ethics approval was obtained from the Institutional Review Board of the University of Hong Kong.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.