Cleft Lip & Palate Research

Direction 1

Stop asking whether presurgical orthopedics works — ask for whom

The tension. Four meta-analyses across fourteen years (Papadopoulos 2012, Hosseini 2016, Di Blasio 2024, Rabal-Soláns 2024) find no significant effect of presurgical orthopedics on arch dimensions, feeding, or growth. But Bühling (2025), the only prospective interventional study in this set, finds active plates outperform passive ones — 5.85 mm versus 3.57 mm of cleft width reduction — and names its own key limitation:

"the lack of numerical or categorical guidelines for the allocation of patients into the active or passive plate groups."

Bühling also reports that baseline cleft size did not predict how much change occurred. That is either a real and surprising finding or an artifact of a 20-patient sample, and it matters enormously which.

The study. Derive and externally validate a morphology-based rule for allocating infants to active, passive, or no plate. Use archived neonatal casts — every cleft center has decades of them in cupboards. Test the treatment-by-morphology interaction rather than the main effect.

Why it is feasible. Framed as heterogeneous treatment effect estimation, this sidesteps the ethical objection Machado raises about withholding treatment from newborns: you are comparing two accepted treatments, not treatment against nothing.

Direction 2

Test whether the protocol literature is secretly measuring access to orthodontic care

The tension. van Roey's most interesting paragraph is buried in his discussion. Dental arch relationships were worse under one-stage palatoplasty — but 67% of those patients came from Asia and South America, where cleft care is often not covered by insurance, against 60% of Oslo-protocol patients from Europe and Oceania. He then cites Southall: orthodontic treatment alone shifts the GOSLON score by 1.1 points, which is larger than the between-protocol differences the field has been arguing about for decades.

He raises the hypothesis. He does not test it.

The study. Meta-regression across his own 156 study groups, with country income, insurance coverage, and reported orthodontic exposure as moderators of arch relationship outcomes.

Why it is feasible. The data are already extracted and published in his supplementary files. This is weeks of work, not years, and no patients are involved. The conclusion would be outsized: it would tell the field how much of its central controversy is an artifact of who can afford a retainer.

Direction 3

Build the preoperative VPI prediction model that Liu (2026) tried and failed to build

The tension. Liu compared pharyngeal anatomy in isolated cleft palate against Pierre Robin sequence and concluded there was no difference in need ratio. But the cleft palate group had a median age of 23.7 months and the Pierre Robin group 3.8 months. The paper's own discussion concedes the age gap "likely explains" the differences found. As a comparison it is uninterpretable.

Meanwhile Tahmasebifard (2025) has the right method — MRI, 105 individuals with velopharyngeal insufficiency against 45 controls, closure height 3.7 mm lower in the VPI group — but is cross-sectional, underpowered for subgroups, and explicitly flags that "the timing and type of palatal surgery might also influence the height of VP closure. Future studies should evaluate the impact of these factors."

The study. A prospective, age-standardized preoperative imaging cohort. Measure need ratio, closure height, velar length, and cleft width at a fixed preoperative timepoint; follow to speech assessment at five years. Output: a calibrated risk score.

Why it matters. It would let surgeons match the palatoplasty technique to the anatomy rather than to habit, and it converts the Pierre-Robin-versus-isolated-cleft debate into a covariate.

Direction 4

Close the nutrition → surgical timing → complication loop

The tension. Four papers touch this chain and none connect it.

• Kinter (2026) — children with clefts sit near −1 SD weight-for-age; 10–50% have failure to thrive; data past 12 months is sparse. Cites a finding that length predicts fistula better than weight, then says "further studies are required to identify the metrics with the greatest clinical relevance."
• Williams (2025) — Pierre Robin sequence, poor growth, and Hispanic ethnicity independently predict delayed palate repair across 17 teams.
• Wu (2023) — in 980 palatoplasties, poor healing tracks cleft width, operative time, and surgeon seniority. Nutritional status was not among the recorded variables.
• Ormanidou (2026) — lower birthweight in infants with clefts, stratified by country income.

The study. Does anthropometric status at the time of palatoplasty predict fistula and velopharyngeal insufficiency, independent of cleft width and surgeon experience? And the harder question: is delaying surgery to improve growth actually protective, or does it simply trade a healing benefit for a speech cost? Wu's dataset is nearly the right shape; it needs weight and length added.

The spin-off. The ethnicity finding deserves its own paper. Williams says only that "future research should investigate social determinants of health that may predict timing of palate repair." Nobody has.

Direction 5

Test whether CLEFT-Q means the same thing in every country using it

The tension. This is the most methodologically distinctive question in the set, and it has teeth because ICHOM has adopted CLEFT-Q as a global standard.

• Michael (2022), in Nigeria — after palatoplasty her patient exceeded normative CLEFT-Q values for speech function, speech distress, and psychological function, but school and social function stayed below the 95% confidence interval. Her conclusion: "the cleft Q may need further validation in our setting."
• Kabuyaya (2024) then uses the French CLEFT-Q in the DRC without addressing this, and concedes the study "did not take into account other potential factors… such as socioeconomic conditions."
• Ruiz-Guillén (2021) finds gender moderates improvement specifically in the social function domain.
• Hasanuddin (2025) documents, across 13 countries, that clefts are attributed to curses, eclipses, and ancestral spirits, with consequences running to stigmatization and infanticide.

The study. Formal measurement invariance and differential item functioning testing of the CLEFT-Q psychosocial subscales, with Hasanuddin's cultural-belief taxonomy as the grouping variable.

Why it matters. If the instrument is not invariant, every cross-country CLEFT-Q comparison — including the normative values everyone benchmarks against — rests on shaky ground. That is an important result whichever way it comes out.

Direction 6

Replace "cleft width" with automated 3D shape phenotyping

The tension. Du (2022) ends by asking for exactly this:

"more studies are required where data mining and machine learning techniques can be implemented to quantitatively determine the characteristics of different subtypes of the malformation… the computational analysis of the 3D model where topological features such as area and length of the palatal bony structure can be automatically extracted."

Bühling (2025) has already shown that digital measurement on 3D scans agrees with manual measurement on plaster casts, so the method is validated.

Almost every quantitative claim in this literature compresses a complex three-dimensional deformity into one or two numbers — cleft width, cleft palate index, alveolar cleft width. That is very likely a major source of the heterogeneity that Rabal-Soláns and Machado both blame for their inconclusive results.

The study. Statistical shape modelling of neonatal cleft maxillae from digitized archival casts, then use the shape coordinates rather than the scalars as predictors in Directions 1 and 4.

Why it is feasible. This is the enabling method for two other directions, and the input data is already sitting in storage at every major center.

Direction 7

Two quicker, cleaner wins

7a. A reporting-bias-corrected estimate of VPI and fistula rates. van Roey's funnel plot shows that studies reporting high rates of velopharyngeal insufficiency are underrepresented, and he states plainly that "in practice, the incidences of VPI are likely to be higher." He also notes that definitions of oronasal fistula are inconsistent or entirely absent across studies. So: apply selection models to his extracted data to estimate the true rates, and run a Delphi process to fix a definition. He even proposes the long-term mechanism — registries let complication rates be reported without tracing back to individual institutions, "which may reduce reluctance to report." That is an empirically testable claim about how researchers behave.

7b. An enhanced-recovery protocol with healing, not pain, as the primary outcome. Grabar (2023) found 61% of cleft surgeons never use a standardized perioperative protocol, but 68% would adopt one. Suleiman (2024) already supplies the evidence-based analgesia regimen and flags dexamethasone and pre-incisional infiltration as unresolved. Wu (2023) supplies the missing link: agitation and crying drive wound dehiscence and fistula. Nobody has run a cleft enhanced-recovery trial with fistula as the endpoint rather than a pain score — which is the outcome surgeons actually care about, and therefore the reason they would adopt the protocol.