How accurate is the Modified Checklist for Autism in Toddlers for identifying autism in children up to 48 months of age?
In this crazy meta-analysis in which the authors intended one thing and did something else, it appears that some version of the M-CHAT is reasonably sensitive and specific for detecting ASD in children younger than 48 months.
Plan de l'etude:
These authors searched 3 databases to find published English-language studies that used the Modified Checklist for Autism in Toddlers (M-CHAT) or the M-CHAT, Revised with Follow-up (M-CHAT-R/F) to screen children younger than 48 months for autism spectrum disorder (ASD). They described no efforts to identify unpublished studies and do not report data addressing potential publication bias. To be included, the studies had to have detected at least 10 children with ASD and had to describe how their authors identified false-negative cases. Two authors independently assessed the methodologic quality of the included studies. Fifty studies were included, one of which described 2 separate samples, resulting in 51 separate analyses. The authors provide no information on the total number of children assessed other than 41% of the studies had fewer than 500 participants. The M-CHAT was administered in English in 26 of the studies. The studies used different versions of the M-CHAT (32 studies used the original version, 19 used the revised version) and 30 used the follow-up portion. Additionally, the authors found different methods of confirmation of ASD diagnoses and for identifying false-negative cases. Most (71%) of the studies were at high risk of bias. Despite all this, the authors decided to pool the data for all versions of the M-CHAT, not just the M-CHAT-R/F (pooling bad data doesn’t improve them). The overall sensitivity was 0.83 (95% CI 0.77 - 0.88) and the overall specificity was 0.94 (0.89 - 0.97). These translate to a positive likelihood ratio of 13.8 (7.0 - 29.3) and a negative likelihood ratio of 0.18 (0.12 - 0.25). Not surprisingly, the authors identified significant heterogeneity (and several sources) among the data. The most clinically important source was when the M-CHAT was used with children at low likelihood of ASD, it was even more specific (0.99 vs 0.70) than when used with high-likelihood children. Additionally, when the diagnosis was confirmed within 6 months of the initial screen, the M-CHAT was more sensitive (0.86 vs 0.57) than with later confirmation.
Henry C. Barry, MD, MS
Michigan State University
East Lansing, MI