Figure 1 documents the number of articles identified for review and inclusion at each stage of the project. The lengthy list of search terms used in this review led to a vast number of articles being identified for screening and full-text review. After application of the exclusion criteria, there were 465 articles reporting quantitative data for chewing, oral processing or swallowing in humans across more than one food or liquid consistency. Of these, 70 articles [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88] reported non-overlapping data for measures of oropharyngeal swallowing function or physiology, collected using stimuli that were described according to IDDSI terminology or unambiguously mappable to the IDDSI Framework. Six additional articles were excluded due to overlap [89,90,91,92,93,94]. Given the large number of qualifying articles in this subset, we decided not to synthesize results from additional articles in the full-text review set, for which it would be necessary to infer the mapping of stimuli to IDDSI levels. This decision had the consequence of limiting the data synthesis set to articles published beginning in 2017, aligned with release of the IDDSI Framework [8]. For articles reporting use of Varibar™, a commercially available barium product, stimuli were mapped to liquid levels on the IDDSI Framework based on information from the IDDSI website [95]: Varibar Thin = level 0 thin, Varibar Nectar = level 2 mildly thick, Varibar Thin Honey = level 3 moderately thick, Varibar Honey = level 4 extremely thick and Varibar Pudding = level 4 extremely thick. A summary table describing key characteristics of the 70 included articles can be found in supplementary material file 3.

PRISMA Diagram showing the results of the search strategy and review

The data synthesis set included 42 studies using VFSS, 15 using FEES, eight using HRPM, three using EMG and one study each using lingual pressure, nasal airflow and ultrasound. Although 10 studies reporting either acoustic or accelerometry data for swallowing across consistencies were included in the full-text review set [96,97,98,99,100,101,102,103,104,105], none of these described their stimuli using IDDSI terminology or testing methods. The use of IDDSI terminology was also rare amongst articles reporting EMG data. Table 1 (liquids) and Table 2 (foods) shows the extent to which data for different IDDSI levels were reported by study instrument type. Here it can be appreciated that the strong majority of studies included level 0 thin liquids in their protocols (regardless of instrument type). Inclusion of level 1 slightly thick, level 2 mildly thick and level 3 moderately thick liquid stimuli was more common in studies using VFSS, while a higher proportion of FEES studies included level 4 extremely thick liquids/pureed foods and level 7 regular food stimuli.

Among the 70 articles included for data synthesis, 25 reported comparative data for measures of swallowing safety across more than one IDDSI level, collected across a total of 2,596 participants. This set of 25 studies excludes eight additional articles where a floor effect was seen, with Penetration-Aspiration-Scale (PAS) [106] scores of 1 or 2 reported for all consistencies tested, suggesting safe swallowing without any meaningful opportunity to display a difference across consistencies [43, 53, 56, 60, 63, 80, 81, 86]. Three articles reported measures of swallowing safety in healthy adults [26, 39, 49]. On cup sip tasks, Garand and colleagues [39] reported that 10% of participants showed penetration on thin liquid stimuli; this dropped to 6% with mildly thick liquids, while aspiration was reported to occur in only a single participant on both of these stimuli. For 5 ml tasks with liquids, penetration was seen in 7% (thin), 6% (mildly thick), 2% (moderately thick) and 1% (extremely thick) of participants, and also in 1% on a level 7 regular food bolus (a bite of cookie). A single healthy participant was reported to aspirate on the 5 ml thin and mildly thick liquid tasks. The meta-analysis by Borders and Steele [26], included unique swallowing safety data for a sample of 78 healthy adults (source data [78]). When disaggregated, the healthy source data showed a 9% participant-level frequency of penetration on sips of thin liquid; this broke down to frequencies of 1%, 1% and 7% for PAS scores of 3, 4 and 5, respectively, with no aspiration (i.e., PAS scores of 6, 7 or 8). The frequency of penetration dropped to 1% and 4% on sips of slightly thick and mildly thick liquids, respectively, with no penetration events on 5 ml teaspoons of moderately or extremely thick liquids [26, 78]. The 30 healthy control participants in the study by Lee and colleagues [49] were reported to have mean ± standard deviation PAS scores of 1 ± 0 (i.e., no airway invasion) on both thin and moderately thick liquids.

Thirteen studies of the 25 studies in the safety set reported PAS scores for patients with dysphagia on slightly or mildly thick liquids compared to thin liquids [26, 28, 32, 33, 54, 57, 58, 65, 68, 71, 82,