The modified Mallampati classification: four classes, and a screen that misses about half of difficult intubations

What it says#

The score exists because of one idea: that the size of the base of the tongue, relative to the space it sits in, decides how much of the larynx a laryngoscopist will be able to see. Mallampati proposed it in 1983 and tested it in 1985 in 210 healthy adults. The examination he described is the one still used. The patient sits, opens the mouth fully and protrudes the tongue maximally, without phonation, and the examiner records which structures remain visible. Three classes came out of it: class I, complete visualization of the faucial pillars, soft palate and uvula; class II, the faucial pillars and soft palate visible with the uvula obscured by the tongue base; class III, visualization limited to the soft palate alone. Mallampati defined difficult intubation as inadequate exposure of the glottis at direct laryngoscopy and reported a significant correlation (p less than 0.001).

Two years later Samsoon and Young added the class that made it a four-point scale: a view in which the faucial pillars, uvula and soft palate are all obscured. Their paper was a retrospective look at patients whose tracheas had previously been impossible to intubate; failed intubation occurred in 7 of 1980 obstetric cases over 3 years and in 6 of 13,380 surgical patients, and in their obstetric series all 7 difficult intubations corresponded to class III or IV. In 1998 a class 0 was added for the rare case in which the epiglottis is visible during pharyngoscopy, a finding that correlated with consistently easy intubation.

The modified score in routine use today, as StatPearls states it: class 0, epiglottis partially or fully visible; class I, full view of soft palate, uvula and tonsillar pillars; class II, soft palate and uvula visible, pillars obscured; class III, soft palate and uvula base visible; class IV, visualization restricted to the hard palate. An extended version, performed with the craniocervical junction extended, has been reported to have greater sensitivity and specificity than the modified test.

The accuracy is the part that gets left out of the training video. The Cochrane review of airway physical examination tests, 133 studies and 844,206 participants, found 105 studies of the modified Mallampati test. For difficult tracheal intubation, its summary sensitivity was 0.51 (95% CI 0.40 to 0.61) with a specificity of 0.87 (0.82 to 0.91); that sensitivity was the highest of the bedside tests for this target, which says as much about the other tests as about this one. For difficult laryngoscopy, the modified test had the lowest specificity in the set, 0.80 (0.74 to 0.85). For difficult face mask ventilation it could only be estimated at a sensitivity of 0.17 (0.06 to 0.39) and specificity of 0.90 (0.81 to 0.95). The reviewers' conclusion is that screening tests are expected to have high sensitivities, that all the tests they investigated had relatively low sensitivities with high variability, and that the upper lip bite test had the most favorable properties of the group.

The older literature points the same way. Positive predictive values as low as 33.3% have been reported, and 22% for class III specifically, in part because views the Mallampati system calls difficult often turn out to be favorable under the Cormack-Lehane grading of the actual laryngoscopic view. StatPearls adds the context that difficulty is common enough to matter, roughly 1% to 5% of patients are difficult to mask ventilate and around 5% are difficult to intubate, and that in a large cohort 93% of difficult intubations and 94% of difficult mask ventilations were unanticipated.

Two findings travel beyond the operating room. In 242 pregnant patients the proportion classified class IV rose by 34% at 38 weeks compared with the same women at 12 weeks, attributed to fluid retention and soft tissue edema. And the score is an independent predictor of obstructive sleep apnea, with the odds of OSA doubling for every one-point increase in class and the apnea-hypopnea index rising by more than 5 events per hour on average.

Who it applies to#

Anyone evaluating a patient before sedation, anesthesia or intubation, in a dental office or anywhere else, who can get the patient to sit up, open wide and stick out the tongue. It is a 10-second observation with no equipment, which is exactly why it spread.

It does not apply cleanly to a patient who cannot cooperate, and it does not see the airway made difficult by disease rather than anatomy: infections and tumors of the head and neck, epiglottitis, Ludwig angina, goiter and facial trauma are named as conditions whose difficulty the score cannot predict. It is also not the same thing as the Cormack-Lehane grade, which describes what the laryngoscopist actually sees, and it is not a rule in dentistry. The ADA's October 2025 sedation guidelines, read in full for this entry, never mention it; the pre-procedural evaluation they describe runs on ASA physical status, consultation with the physician for ASA III and IV patients, and body mass index.

What changed#

Nothing about the classes has changed since 1998. What has changed is the claim made for them. The score entered practice as a predictor; the meta-analytic work of the last twenty years has demoted it to one risk factor among several, to be read next to neck circumference over 40 cm, limited neck mobility, mouth opening under three finger breadths, thyromental distance under 6 cm, limited mandibular protrusion and a documented history of a difficult airway. StatPearls states the current position plainly: the score alone lacks sufficient sensitivity to identify patients with challenging airways, and reliance on it in isolation is not advised.

Where the source is#

The link under the title opens the StatPearls chapter Mallampati Score, by James S. Jackson and Bryan Rondeau, on the NCBI Bookshelf (Bookshelf ID NBK585119), last updated July 7, 2025 and distributed under a CC BY-NC-ND 4.0 license; it was read in full for this entry and is the source of the class definitions, the examination technique and the historical sequence. The two original papers are cited above and were read as PubMed abstracts: Mallampati and colleagues, Canadian Anaesthetists' Society Journal 1985, and Samsoon and Young, Anaesthesia 1987; both journals are behind paywalls, so the class definitions here are as StatPearls states them, not transcribed from the originals. The accuracy figures come from the Cochrane review by Roth and colleagues, CD008874, whose abstract and authors' conclusions were read on PubMed and whose full text is free in PubMed Central. The statement that the ADA's sedation guidelines do not use the score is based on a full-text search of the ADA's October 2025 guidelines PDF, linked above.