We searched over 40 electronic databases, including MEDLINE^®, Ovid MEDLINE^® In-Process & Other Non-Indexed Citations, EBM Reviews – Cochrane Central Register of Controlled Trials, EMBASE, CINAHL^®, PsycINFO^®, International Pharmaceutical Abstracts, Pascal, Global Health, Science Citation Index Expanded and Social Sciences Citation Index (via Web of Science^®), Social Sciences Abstracts, and Psychology and Behavioral Sciences Collection. Trials registers (Current Controlled Trials, ClinicalTrials.gov, the Australian Clinical Trials Registry, and the National Research Register in the United Kingdom) were searched for additional trials. Search terms such as methamphetamine, variant spellings of methamphetamine, amphetamine-related disorders, and crystal meth, were adapted for each database and appropriate subject headings and keywords were used. In addition, an extensive search for grey literature was conducted. Hand searching was conducted in relevant scientific journals, scientific meetings, and the reference lists of relevant reviews and included studies were reviewed. We restricted the search results to English-language studies. The literature search is considered up to date as of May 15, 2006. Full search strategies and lists of resources searched are available [see Additional file 1].

Two reviewers independently screened the titles and, when available, the abstracts. Based on general inclusion criteria, studies were classified as "potentially relevant", "irrelevant", and "unclear". The full text of studies described as "potentially relevant" and "unclear" was obtained and two reviewers independently applied the specific inclusion criteria. Studies were included if they compared children ≤ 18 years of age who did and did not use MA (the comparison group could be other drug users or children who do not use drugs). The following study designs were included: case-control, cohort, and cross-sectional. Studies were excluded if they did not have a comparison group, if the outcomes were not measured quantitatively or if they were uncontrolled before and after studies, case-series, or case studies. Disagreements were resolved through discussion or through third party adjudication, as necessary.

Two reviewers independently assessed methodological quality; discrepancies were resolved through consensus or by third party adjudication as required. Observational studies were assessed using the Downs and Black checklist19. This tool comprises six sections that assess reporting, external validity, internal validity (bias), internal validity (confounding), and power.

Data were extracted by one reviewer and checked for accuracy and completeness by a second reviewer. A standard data extraction form was developed and the data were subsequently entered into an electronic database. For each included study, we extracted information about the population (demographics and sources), type of study (study design, prospective or retrospective data collection), definitions and details of risk factors, and the numeric results.

After reviewing the studies that met our inclusion criteria, we made a post hoc decision to group the studies and conduct the data analysis by the nature of the comparison group: 1) youth who did not use illicit drugs (referred to as "low-risk") and 2) youth who abused illicit drugs other than MA or were recruited from juvenile detention centers (referred to as "high-risk"). For dichotomous risk factors (e.g., sex) we calculated pooled odds ratios (OR) with 95% confidence intervals (95% CI), where appropriate. For continuous risk factors (e.g., years of education) we calculated weighted mean difference (WMD) and 95% CI using the DerSimonian and Laird random effects model20. A random effects model was used because it allows for combining heterogeneous results where the heterogeneity cannot readily be explained21. The resulting estimate is more conservative because the resulting confidence intervals are wider. Statistical heterogeneity was assessed using the I² statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance. For this review a value greater than 50% was considered as substantial heterogeneity22. In instances where only one study reported the specific risk factor or where pooling was not appropriate, the effect estimate (OR or WMD) and 95% CI were reported for each risk factor.

The database, grey literature searches and hand searching yielded 2,376 potentially relevant studies. In total, 106 unique studies were reviewed and 13 met our inclusion criteria. Of these, two studies assessed risk factors in the same population and are treated as one study for the purposes of this report23. Therefore, our review includes 13 publications but only 12 unique studies. Study retrieval and selection is outlined in Figure 1.

Studies were excluded from the review for the following reasons: not relevant to the topic (n = 34), incorrect study population (n = 33), inappropriate study design (n = 12), not primary research (n = 13), and inadequate data (n = 1).

The characteristics of the 12 included studies and populations are presented in Tables 1 [see Additional file 2] and 2 [see Additional file 3]. Most studies relied on self-reported MA use. Three studies administered urine tests to determine MA use242526 and one study diagnosed children with MA dependence27. The remaining studies relied on self-reported MA use. Seven studies were conducted in North America 27282930313233 and the remaining five in Asia2324263435. The majority of the studies were published recently; the median year of publication was 2004. The median sample size was 604 and ranged from 60 to 78,715. Three studies used a case-control design and the remaining nine were cross-sectional.

In five studies, youth who reported using MA were compared to youth who did not use illicit drugs, hereafter referred to as "low-risk" youth. These low-risk youth were sampled from school populations. For the remaining seven studies, youth who reported using MA were compared to youth who abused illicit drugs other than MA, and/or youth who were sampled from juvenile detention centers, hereafter referred to as "high-risk" youth. Youth were recruited from a variety of locations, including schools, detention centers, juvenile homes, or treatment facilities. One study included youth up to the age of 24 and one study did not report the age of participants. The studies assessed a variety of risk factors, including demographic variables, mental health status, and risky behaviors.

The median Downs and Black score was 15 of a possible score of 29 and ranged from 12 to 28 [Table 1] suggesting a risk of bias for those studies that received lower quality scores. Six of the twelve studies reported adjusted results, that is, the risk factor of interest was adjusted for other potentially confounding risk factors232628313234. Six studies disclosed their funding source and the most common source was a government agency2326293032.

Five studies compared risk factors for MA among low-risk youth, as defined above (Table 2) 2324262829. There was considerable heterogeneity among the studies, which precluded the calculation of pooled estimates of effects in most cases. Pooled estimates are presented where appropriate.

Seven studies examined risk factors for MA among high-risk youth (Table 3)27303132333435. Because of substantial heterogeneity among the studies, pooling of estimates was generally not appropriate.

There were only 12 studies that met our inclusion criteria. These studies were fairly heterogeneous, which precluded pooling of results for most risk factors. Furthermore, many of the risk factors were assessed in only one study and the sample size was small. This is reflected in the wide confidence intervals and imprecise effect estimates. Most studies were either cross-sectional or retrospective making it impossible to assess a causal relationship between the risk factors and MA use. As with any systematic review, there is the possibility of publication and selection bias. However, we feel the risk for publication bias was minimized by our exhaustive search process. In addition to electronic databases, the reference lists of the included studies were searched, relevant conference proceedings and key journals were hand searched, and a thorough grey literature search was conducted.

Youth who engage in risk-taking behaviors, live in an unstable home environment, have a psychiatric condition, and have peers that use or sell MA have a higher propensity for MA use. In order to identify youth at risk for MA use, health care workers and counsellors need to conduct a holistic assessment that includes psychiatric, lifestyle, and family history.