The pilot study reported here has provided much useful data for the design of a full-scale trial of acupuncture for chronic neck pain. A key finding is that 12.3% of eligible patients on the GP database consented to participate in the trial, a remarkably similar percentage to the 12% found by McCarney et al [20
] in a trial of acupuncture for migraine. Based on the experience of our study, this method will enable a large sample to be recruited to a trial relatively quickly, in comparison with prospective recruitment of incident cases. We also found that a sample size of 458 patients and a GP practice population base of 230,000 would be required in a full-scale trial if GP databases are to be used as a source of identification and recruitment of patients. The evidence from our pilot study suggests that this recruitment strategy is feasible.
We found a trend towards higher levels of satisfaction among those patients referred to acupuncture, compared to those receiving usual GP care alone. All of the patients receiving acupuncture treatment reported high levels of acceptability (with one exception). However, there were some concerns about the safety of acupuncture, specifically the negative reactions to treatment. These included a temporary worsening of symptoms, dizziness, and tiredness, with two patients withdrawing from treatment as a result. Although there were no serious adverse events, defined as "events requiring hospital admission, leading to permanent disability, or resulting in death" [25
], safety is clearly an important issue that should be carefully considered when developing the design of a full-scale trial. We recommend the provision of adequate monitoring of adverse events and clinical supervision for acupuncturists.
The pragmatic design of the trial was also considered appropriate given the widespread use of acupuncture and the fact that this design facilitates economic evaluations [26
]. Trials embedded in real world practice tend to have strong external validity, though often at the expense of weaker internal validity than explanatory trials. Due to the difficulties of blinding in acupuncture research generally, Birch [10
] has highlighted the added importance of ensuring that all stages of analysis are blinded, from data-entry to evaluation of the results. Although impractical for the purposes of this pilot, blinding should be implemented in a large-scale trial, to help avoid the potential for bias [10
The recruitment processes used in this pilot were successful as patients were identified and recruited relatively simply and quickly. One limitation of our screening procedures was that one patient with cancer inadvertently entered the trial. This indicated a potential flaw with the search strategy, and a future large-scale trial should consider including a cancer question as part of the screening questionnaire given to patients, if the intention remains to exclude patients with cancer.
This study did not set out to establish whether there were components of acupuncture that had specific efficacy. Instead our design was a pragmatic one, where we are working towards an evaluation of the impact of the overall package of acupuncture care. A different research question could have been used to determine the relative contributions of components of the treatment. To do this, a 'placebo' or sham acupuncture control would be used, to control for the components of treatment that are not specific to acupuncture, such as time and attention. Sham approaches are not suitable within pragmatic trials, since they are artificial controls that do not model usual practice [27
], making it difficult to meaningfully interpret their results [6
]. Sham acupuncture approaches are generally problematic since there is evidence that they can produce a physiological effect that may be therapeutic [28
Although the main outcome measure used in this pilot is a validated scale, such self-report measures are subjective. Therefore for a large-scale trial it might be useful to consider including an objective outcome as a secondary measure [30
]. Also, given that there is evidence to suggest that preference and belief might influence outcome [31
], it is recommended that a large-scale trial establishes preference and belief prior to randomisation, so that their potential influence on outcome can be explored in the analysis [31
Overall, in terms of generalisability of the trial, the broad inclusion criteria for recruiting patients made it more likely that the patients in the trial were fairly representative of those typically presenting with chronic neck pain. The acupuncture treatment protocol was also fairly broad, though sufficiently standardised to assist replication. Acupuncturists found it sufficiently flexible to allow them to use an individualised approach, reflecting traditional acupuncture as it is usually practiced. The generalisability of acupuncture treatment was further improved by using more than one acupuncturist in the trial.