A study by an “airway” orthodontist looks interesting?

I realise that I have been writing a lot about orthodontics and the airway. The main reason is that several recent publications make claims that warrant close evaluation. This paper, produced by Derek Mahoney, a high-profile airway orthodontist in Australia, was published this week and gained some interest on social media. Nevertheless, I thought this was worth a closer look because it may attract further attention. This is my academic interpretation of this publication.

In this study, they examined the effect of sequencing slow rapid maxillary expansion (SRME) and adenotonsillectomy (TA) in paediatric obstructive sleep apnoea. A team from Australia and Hungary conducted the study. The Journal of Clinical Medicine published the paper.

The paper is open access so anyone can read it.  The Journal of Clinical Medicine is listed as a predatory journal on the Predatory Journals website. 

What did they ask?

They did this study to ask the following question. 

” What are the combined effects of adenotonsillectomy and maxillary arch expansion in pre-pubertal children with OSA, and is there an effect of the sequence of treatment? 

What did they do? 

This was a retrospective cohort study using a subsample of patients from a larger cohort that I have previously posted about. 

They selected 80 sets of patient records from an original sample of 3,671 children aged 7-9 years who underwent polysomnography at their first orthodontic consultation, representing 2% of the original sample.

The main inclusion criteria were the availability of complete PSG records at three times:

• Baseline (pre-treatment)
• After the first intervention
• After both interventions

They had to complete slow rapid maxillary expansion and adenotonsillectomy, regardless of the sequence. 

They had no history of prior orthodontic treatment or diagnosed sleep disordered breathing at baseline. 

Exclusion criteria were incomplete records and the absence of level one hospital-based PSGs. 

They then divided this sub-sample into two groups. This was “based on a treatment sequence that was determined by the patient and parental preference and logistical factors rather than randomisation”. As a result, they suggested this reflected the “real world” of decision making.

Group one was designated TA first, comprising patients who underwent tonsillectomy as the initial intervention, followed by SRME. 

The second group had SRME as the initial intervention. This was due to delays in accessing publicly funded ENT appointments or waiting periods for private insurance cover. 

They collected demographic details, cephalometric classifications, and BEARS questionnaire scores greater than 5, indicating an SDB risk. 

Accredited sleep laboraties did the polysomnography measurements.

The primary outcome for this study was the respiratory disturbance index (RDI). This is the average number of apnoeas, hypopnoeas, and respiratory effort-related arousals per hour of sleep. 

Statistical analysis was based on a repeated-measures analysis of variance over time. The authors did explain the method clearly, and I couldn’t really work out the statistical plan.

What did they find? 

The mean age of the participants was 8.3 years, and 43% were male. At the start of treatment, the mean RDI was 18.99, indicating moderate severity.

At the start of treatment, there were no differences between the groups in most measured variables; however, there was a marked difference in BMI. The TA first group had 87% of the sample classified as overweight or obese, whereas the SRME group had 45% of participants in this classification. This is important because it represents a marked imbalance between the groups. 

I have extracted some of the data into this table. This illustrates the RDI for the combined BMI groups. I also calculated the 95% confidence intervals for this data.

Time	TA first	SRME first
Start	18.73 (18.00, 19.455)	19.48 (18.81, 20.14)
After- first intervention	10.36 (9.71, 11.00)	9.64 (9.0, 10.22)
After second intervention	4.86 (4.42, 5.29)	4.19 (3.79, 4.58)

When you look at the 95% confidence intervals, you can see that they overlap for the interventions’ effects. This means that the effects may not be statistically significant. Furthermore, they do not overlap at the start of treatment.

I looked closely at their statistical analysis, and the tables outlining it were rather brief, which made it difficult to interpret the findings.

They also presented a table showing RDI reduction. However, I felt we could be more focused in this blog post by simply comparing the values between the two groups.

The conclusions were 

“Both TA and SRMA significantly improved the RDI in pre-pubertal children with OSA and maxillary constriction. The greatest benefit was seen when these were combined. The SRMA first, followed by TA, was somewhat more effective by about 1.5 events per hour overall.” 

What did I think?

This study used the Respiratory Distress Index (RDI). I was unfamiliar with the RDI and sought further information on this measure. The RDI is similar to the Apnoea-Hypoxia Index (AHI) but includes Respiratory Effort-Related Arousals (RERAs). This makes it useful for assessing milder forms of sleep-disordered breathing. It is highly sensitive to treatment and, importantly, more sensitive than the AHI.

The RDI values can be classified as

• Normal: RDI < 1 event per hour.
• Mild OSA: RDI between 1 and 5 events per hour.
• Moderate OSA: RDI between 5 and 10 events per hour.
• Severe OSA: RDI > 10 events per hour. 

Problems with the study

Importantly, we may need to consider the accuracy of diagnosing sleep disorders from a single night’s readings. A systematic review has examined this and concluded that this variation may lead to misdiagnosis in single-night studies. We need to consider this when we interpret the results of this study, which used a single-night recording.

Bias

There is also considerable risk of further selection bias in the study, given that the authors analysed only 80 sets of records from a total sample of 3971 (2%). This risk is further compounded by the inclusion criteria, which require completion of treatment and full records. As a result, the study is at risk of being biased towards favourable outcomes.

When I looked at the data they presented, it was clear that the groups were unbalanced for BMI at the start of treatment. It appears that the clinicians had a preference for TA in obese children, which could lead to selection bias in the study. Furthermore, the treatment allocation order was determined by delays in access to ENT care. It appears that if a patient’s ENT treatment was delayed, they received SRME. This design is subject to systematic confounding, and we must account for differences between the groups. For example, they may be children from low socioeconomic groups.

Control group and effect size

In addition, there was no untreated control group. This is crucial for this form of study because the CHAT study estimated that 50% of 5- to 9-year-old children with OSA who were randomised to watchful waiting achieved OSA remission, as defined by PSG findings, after 7 months.

When we look at a paper, it is important to consider the differences in effect sizes. Even a cursory examination shows that there is limited difference between the two intervention sequences. I also could not really understand the statistics they used because of a lack of data, particularly the p-values. They also presented the percentage change in RDI as an outcome. It is more important to look at the group differences.

It is also relevant to point out that they did the in 9 private orthodontic practices with an interest in airway treatment. As a result, the findings may not be relevant to the general population.

We must also consider whether Dr Mahoney has an undeclared conflict of interest, given that he is the CEO of a company that operates several orthodontic practices specialising in “airway-focused” care.

Final comments

I may be criticised for being so blunt about a study. However, I felt it was necessary to point out the deficiencies in this paper.  This study provides very low-level evidence. However, its data may be used for planning randomised trials in this important area.

Now that you have read this blog post.

Last year, I requested donations to support my blog, and thankfully, I received an overwhelming response. With your help, I was able to raise sufficient funds to keep my blog going for another year. 

Now, I’m seeking your support again to help me cover the blog’s running expenses for the upcoming year. The funds will cover web hosting, software upgrades, email lists and general running costs. 

I believe that if every reader contributes a small amount, I can cover the yearly running costs. Your help will be much appreciated.

To make a donation, just use the form below. You can use the set-amount buttons or the free-field button to enter other amounts.

Then the easiest way to pay is to tap the Apple Pay button. If you want to use another payment method, click the “next” button and enter your details. This will take you to credit cards, PayPal, and other payment methods.