Two meta-analyses of distant healing studies

Reviews of empirical work on the efficacy of distant healing have found that interceding on behalf of patients through prayer or by adopting various practices that incorporate an intention to heal can have some positive effect upon their wellbeing (e.g., Benor, 1990; Dossey, 1993), but reviewers have also raised concerns about study quality and the diversity of healing approaches adopted — ranging from techniques that usually involve close physical proximity between practitioner and patient, such as therapeutic touch and Reiki healing, through to techniques that work at a distance, such as psychic healing or intercessionary prayer to a higher being — and this makes the findings difficult to interpret since in some cases the beneficial effects could be attributable to placebo effects or to the consequences of general lifestyle changes that are involved in holistic approaches to medicine. Some of these concerns can be addressed by conducting double-blind randomised controlled clinical trials. These entail the random allocation of participants (or patients) to either a treatment or control condition so as to control for selection bias (or alternatively participants are matched on the basis of other variables that are thought to affect the prognosis of their health condition, such as age, gender, co-morbidity, and so on), with patients and attending physicians remaining blind to the allocation so as to control for placebo improvements. Some of the research that meet these criteria has been summarised by Astin, Harkness and Ernst (2000), who identified 23 studies, collectively involving 2774 participants, that produced the predicted improvement in condition with a combined effect size of .40 (p < .001). Despite remaining concerns about the heterogeneity of the database and methodological limitations with some studies, the authors were able to conclude that the evidence was sufficiently strong to warrant further study. However, Astin et al.‘s review is still susceptible to counter explanations in terms of their inability to create an appropriate control condition, since there can be no guarantee that control patients are not beneficiaries of healing intentions from friends, family or their own religious groups, or that relationships between healing intention and wellbeing are not obscured by reliance on relatively crude health outcomes. We planned to address this by focusing on healing studies that involved biological systems other than ‗whole‘ humans (i.e. to include animal andplant work but also work involving human biological matter such as blood samples or cell cultures), which allow for more circumscribed outcome measures and are less susceptible to placebo and expectancy effects. Secondly, since Astin‘s reviews have been published, doubts have been cast concerning the legitimacy of work conducted by Daniel P. Wirth such that it would be unsafe to base conclusions on data that he has provided. There have also been a number of replication attempts since Astin et al.‘s review. We therefore planned to conduct an updated review that included more recent studies and excluded Wirth‘s work. For phase 1, 65 non-whole human studies from 50 papers were eligible for review. The combined effect size weighted by sample size yielded a highly significant r of .381. The fail-safe N gave the value 8332, which is much larger than the critical number (calculated as 335), suggesting that publication bias is unlikely to be responsible for significance. However the effect sizes in the database are significantly heterogeneous, and 25 outliers need to be cropped in order to reduce to non-significance at p > .01, giving a weighted mean effect size for the cropped studies of r = .283, which is still significant. Study outcomes were not correlated with blind ratings of methodological quality, but the average quality rating for these studies is very low (mean quality rating [MQR] = 4.4/10). Studies were blocked by target system type: the 31 in vitro studies had the highest average quality rating (5.35) and a weighted mean effect size, r = .311; 18 studies of non-human animals (e.g., rats, mice, bush babies) had an MQR of 4.07 and a significant r = .218; 16 studies of plants or seeds (MQR = 3.22), gave a significant weighted mean effect size, r = .413. For phase 2, 61 whole human studies across 59 papers that were eligible for review. When combined, these studies yielded a small effect size of r = .145 that was significant. The fail-safe N was 3560 against a critical number of 315. This database is also significantly heterogeneous (χ2 = 512.01), but the mean weighted effect sized for the cropped sample actually increases to r = .202. Holistic judgements of study quality were negatively correlated with study outcome, suggesting that the observed effect might – at least in part – be attributable to methodological shortcomings. Whole human studies were blocked by target system: 9 Reiki or Johrei studies (MQR = 5.50), which gave a weighted mean effect size, r = .285; 11 studies incorporated intercessionary prayer (MQR = 5.33), giving the smallest weighted mean effect size, r = .078; therapeutic touch was implemented in 19 studies (MQR = 5.25), giving a weighted mean effect size, r = .346; and 22 unspecified /other studies (MQR = 5.94) gave a weighted mean effect size, r = .167. Findings with the non-whole human database suggest that subjects in the active condition were observed to have a significant improvement in wellbeing relative to control subjects under circumstances that do not seem to be susceptible to placebo and expectancy effects. Findings with the whole human database gave a smaller mean effect size but this was still significant and suggests that the effect is not dependent upon the inclusion of Wirth‘s suspect studies and is robust enough to accommodate some high profile failures to replicate. Both databases show problems with heterogeneity and with study quality and recommendations are made for necessary standards for future replication attempts