Methods - Published and unpublished reports were identified through computerised and manual searches of bibliographical databases, reference lists from primary articles, and letters to authors, agencies, foundations, and content experts. Selection criteria were applied to blinded articles, and selected articles were scored for methodological quality. Effect sizes were calculated from raw pain scores and combined by using meta-analytic techniques when appropriate.

Results - Twenty four randomised clinical trials met the selection criteria and were categorised by type of intervention: nine used manual treatments; 12 physical medicine methods; four drug treatment; and three education of patients (four trials investigated more than one form of intervention). The intervention strategies were summarised separately. Pooling of studies was considered only within each category. Five of the nine trials that used manual treatment in combination with other treatments were combined. One to four weeks after treatment the pooled effect size was -0.6 (95% confidence interval -0.9 to -0.4), equivalent to an improvement of 16 (6.9 to 23.1) points on a 100 point scale. Sensitivity analyses on study quality, chronicity, and data imputation did not alter this estimate. For other interventions, studies could not be combined to arrive at pooled estimates of effect.

Conclusions - There is little information available from clinical trials to support many of the treatments for mechanical neck pain. In general, conservative interventions have not been studied in enough detail to assess efficacy or effectiveness adequately.

Division of Graduate Studies and Research,
Canadian Memorial Chiropractic College,
Toronto,
Ontario,
Canada

Peter D Aker, associate professor

Chedoke-McMaster Hospitals and School of Rehabilitation Sciences,
McMaster University,
Hamilton,
Ontario,
Canada

Anita R Gross, assistant professor

Department of Clinical Epidemiology and Biostatistics,
McMaster University,
Hamilton,
Ontario

Charles H Goldsmith, professor

Department of Medicine,
University of Saskatchewan,
Saskatoon,
Saskatchewan,
Canada

Paul Peloso, assistant professor

Correspondence to: Dr C H Goldsmith,
Centre for Evaluation of Medicines,
St Joseph's Hospital,
50 Charlton Avenue East,
Hamilton,
Ontario L8N 4A6,
Canada.

Design - Poisson regression analysis of incidence rates in relation to the variables of interest.

Setting - The 403 county districts of England and Wales during 1979-85.

Subjects - Children aged under 15 years.

Results - There were significant trends in the incidence of lymphoblastic leukaemia at ages 0-4 and 5-9 years with the proportion of children in a district who had recently entered the district. While there was no consistent relation between the proportion of recent incomers in the total population of a district and its incidence rate, the combination of higher migration with greater diversity of origins or distance moved was associated with higher incidence in both age groups. Incidence increased significantly at age 0-4 with the level of employment in a district and at age 5-9 with the proportion of households with access to a car. No significant trends were found with commuting.

Conclusions - The results for level of child migration and diversity of total migration provide evidence of an effect of population mixing on the incidence of childhood leukaemia which is not restricted to areas experiencing the most extreme levels of mixing.

Childhood Cancer Research Group,
University of Oxford,
Oxford OX2 6HJ

C A Stiller, epidemiologist

School of Geography,
University of Leeds,
Leeds LS2 9JT

P J Boyle, lecturer

Correspondence to: Mr Stiller.