I read with great interest this pilot of acupuncture in chronic cervical pain. The design of the pilot seems correct. For the next step (large-scale randomised trial to assess acupuncture versus standard GP care), some key points to consider are the method of randomisation and overall discuss a double blinded study (evoked by the authors in the discussion). In acupuncture, some methods have been described such as sham acupuncture interventions that keep the patient unaware of intervention received. If a sham procedure is not possible, the assessor must absolutely be blinded regarding the group allocation to minimize bias. Also, one risk is that patients randomized in the group acupuncture withdraw their consent (beliefs and cultural background). This would lead to important lost to follow up. A study design such as Zelen design may be effective in that case to limit lost to follow up. The patients would be randomised and their consent would be collected after randomisation. If they refuse the strategy they are randomised for (acupuncture), the adverse strategy (standard GP care) will be proposed. Regarding randomisation, it is mentioned that several GPs practices will participate in the large-scale trial. A centralized randomisation would probably be the better strategy of randomisation. Stratification on GPs practices may be interesting to limit the effect of each particular physician. The statistical analysis as mentioned was realized on the principle of intention to treat. I think it would be interesting (especially in the large-scale trial) to detail methods used to treat missing data in the chapter analysis.
Methodological suggestions
9 March 2007
I read with great interest this pilot of acupuncture in chronic cervical pain. The design of the pilot seems correct. For the next step (large-scale randomised trial to assess acupuncture versus standard GP care), some key points to consider are the method of randomisation and overall discuss a double blinded study (evoked by the authors in the discussion). In acupuncture, some methods have been described such as sham acupuncture interventions that keep the patient unaware of intervention received. If a sham procedure is not possible, the assessor must absolutely be blinded regarding the group allocation to minimize bias. Also, one risk is that patients randomized in the group acupuncture withdraw their consent (beliefs and cultural background). This would lead to important lost to follow up. A study design such as Zelen design may be effective in that case to limit lost to follow up. The patients would be randomised and their consent would be collected after randomisation. If they refuse the strategy they are randomised for (acupuncture), the adverse strategy (standard GP care) will be proposed. Regarding randomisation, it is mentioned that several GPs practices will participate in the large-scale trial. A centralized randomisation would probably be the better strategy of randomisation. Stratification on GPs practices may be interesting to limit the effect of each particular physician. The statistical analysis as mentioned was realized on the principle of intention to treat. I think it would be interesting (especially in the large-scale trial) to detail methods used to treat missing data in the chapter analysis.
Competing interests
None declared