In this guest blog by Kyle Fagala, he re-examines a recent systematic review on the controversial topic of expansion and paediatric Obstructive Sleep Apnea, using AI. The comparisons to the original systematic review were surprising, with differing outcomes to the authors results and conclusions. The post made me reconsider the possible future role AI may have within orthodontics.
This part 1 guest blog explores the systematic review by Yu et al 2026, a paper published in Sleep and Breathing in March 2026, Kyle has also carried out the same exercise to re-examining the AAO’s White Paper on the same topic of Sleep Disorder Breathing in Orthodontics Palomo 2026. part 2, and a comparison of both papers by AI, part 3.
I have followed Kyle’s popular facebook group, The Digital Orthodontist for a long time, and enjoy reading his posts.
Yu et al. (2026) recently published a systematic review in Sleep and Breathing on rapid maxillary expansion and pediatric OSA. Reading and critically evaluating all 14 included studies is not something most clinicians have the time or the specific research methodology training to do well on their own. But AI can help.
I purchased a Springer+ subscription, downloaded the full text of the systematic review and every one of the 14 clinical studies it included, and had Claude Opus 4.6 (Anthropic’s highest-level AI research assistant) systematically read, extract, and evaluate each study on a standardized framework: study design, sample size, diagnostic method, presence of controls, risk of bias, key findings, strengths, weaknesses, and an overall evidence quality rating.
The goal was simple: let the studies speak for themselves, as fairly and objectively as possible. I asked it not to favor either “side” of the debate. What follows is what it found.
Yu et al. pooled 14 studies and found a statistically significant reduction in AHI (the main measure of sleep apnea severity) at 6 months (mean reduction of 4.04 events/hour) and 12 months (mean reduction of 6.15 events/hour) following RME in children with PSG-diagnosed OSA.
Of the 14 included studies, only 3 were randomized controlled trials. The rest were before-after studies or non-randomized comparisons with no untreated control group. That matters, because pediatric OSA has a well-documented tendency toward spontaneous resolution as children grow through adolescence.
The three RCTs tell a very different story than the pooled numbers:
The consistent pattern: when a control group or stronger methodology is present, the effect of RME on AHI either disappears or is insufficient as standalone therapy.
To their credit, the authors were transparent about the limitations. Their GRADE ratings for most outcomes were very low to low certainty. The I-squared values (a measure of how much the studies agree with each other) were 93% and 97%, which is far above the threshold for “considerable” heterogeneity. Ten of 14 studies were rated as having serious risk of bias. And the authors’ own conclusion describes RME as “complementary, not definitive.” Those are important qualifications that tend to get lost when only the abstract gets shared.
This review does not prove that RME has zero effect on airway outcomes. What it shows is that the current evidence is not strong enough to confidently attribute AHI improvements to RME itself, rather than to natural resolution, growth, or other factors.
The best available data (the RCTs) suggests that RME may contribute as part of a multimodal approach alongside adenotonsillectomy and medical management, but not as a standalone treatment. If you’re expanding a child who has maxillary transverse deficiency and also has diagnosed OSA, you’re doing appropriate orthodontic treatment that may provide additional respiratory benefit. But the evidence does not currently support performing expansion specifically to treat or prevent sleep-disordered breathing in the absence of an orthodontic indication and a physician’s diagnosis.
Tool: Claude Opus 4.6 (Anthropic)
Access date: April 2, 2026
Purpose of use: The AI was used to
The standardized evidence review framework was created as part of this analysis and does not appear in the original systematic review.
Inputs provided by the author: Full text of Yu et al. (2026), full text of all 14 included clinical studies (purchased and uploaded by the author), and direction throughout the process to ensure thoroughness, consistency, and equal scrutiny of all claims regardless of their conclusions.
Human role: The author read and reviewed each draft for accuracy and fairness, confirmed the analysis was consistent with the source texts based on clinical familiarity with the literature, and directed the AI to apply equal scrutiny to all studies. The author did not select, modify, or exclude findings to steer the analysis in any direction.
Nature of the analysis: The AI extracted findings from the included studies and evaluated them against the systematic review’s pooled conclusions. In the course of this evaluation, the AI also made interpretive observations about patterns in the evidence, such as the divergence between RCT findings and before-after study findings. These observations represent analytical judgments informed by the evidence, not raw data extraction. The author reviewed these observations and confirmed they were consistent with the underlying studies.
Original paper: Yu et al: The effect of rapid maxillary expansion on children with obstructive sleep apnea: a systematic review and meta-analysis
Full AI review: https://www.dropbox.com/scl/fi/5y3ta9atztjxcokzmxlce/Yu-et.-al-An-AI-Assisted-Review.docx?rlkey=4mq6ljddpvaomzepttepc8acj&dl=0
Contents: Kyle Fagala 
Visit Kyle at:
