The Decision to Change Viscosity — What Thickening Actually Changes

One of the most frequent prescriptions written in the hospital ward is not a medicine. It is a single line: "Make the water thick, and grind the meals." This single line carries no dosage like a medication, no side effect section, and no separate consent form received. Yet this single line changes a person's three meals a day and every beverage, and sometimes changes all of the time they have left.

First, let us clarify a few terms. Dysphagia is a condition where swallowing does not work properly or is misaligned. Aspiration is when things like food, water, or saliva enter the airway below the vocal cords. The primer "Swallowing, Speech and Language — The Crossing Where the Food Path Meets the Airway" first established these two terms.

Consistency adjustment is the work of changing the firmness of food and the thickness of liquids to make swallowing easier. Grinding or mincing food is on the food side, and mixing powder or gel into water to make it thick is on the liquid side. This powder or gel that makes liquids thick is called a thickener.

Viscosity is a property representing how sticky a liquid is so that it flows slowly. In this article, "thick" means high viscosity, and "thin" means low viscosity like plain water.

The result that making things thick immediately reduces aspiration comes out in the same direction across studies of various designs. This article is not a text telling you not to make things thick. What this article aims to establish is a different reading method. Even for the same measure, the report card varies depending on what outcome is measured and for how long. We will look in turn at how thickness and volume change swallowing, what determines the "levels" of thickness, why immediate effects and results months later can diverge, and what the price of that is.

Why We Need to Know — The Default for Dysphagia Diets

Texture-modified diets and thickened drinks are commonly used worldwide to reduce the risk of choking and aspiration. Dysphagia is estimated to affect about 8% of the global population (PMID 27913916).

In long-term care facilities, texture-modified diets were the most common meal for people with dysphagia. In data surveying 23,549 residents across 926 care units in 19 countries over a single day, 13.4% of residents were reported to have dysphagia. Among the nutrition methods they received, texture-modified diets were the most common at 42.5%, higher than regular diets (28.2%) (PMID 29030310).

The rate is higher in older adults hospitalized in the hospital. Among 218 hospitalized patients with dysphagia aged 70 and older (average 87.4 years), up to 58.3% needed liquids thickened to swallow safely, and 93.6% received texture-modified diets (PMID 39576562).

Therefore, texture modification is the default for dysphagia management. The more people receive an intervention every day, the greater the reason to closely examine its evidence.

① The Amount Swallowed at Once and the Thickness Act Differently on the Body

A bolus is a single mass of food or water swallowed at once, and in papers it is called a bolus. In the clinic, "making it thick" and "giving it in small amounts" are treated as a single package. However, the amount of the bolus and its thickness act on the swallowing body in different ways.

A classic study measuring this difference side by side is a simultaneous videofluoroscopy-manometry from 1990. Videofluoroscopy is an examination where you watch the passage of food mixed with barium, which shows up well on X-rays, using X-ray video. Ten healthy volunteers swallowed 2 to 20 mL of thin barium and thick barium paste, and the pressure inside the throat was measured at the same time.

The pharynx is the space inside the back of the throat extending downward from behind the nose and mouth, where both food and air pass. At the lower end of the pharynx, food splits off into the esophagus and air into the larynx in front. The upper esophageal sphincter (UES) is a ring of muscle that tightens the entrance to the esophagus and opens when you swallow.

• The main effect when the amount was increased was that the swallowing action started earlier. The tongue base moving forward, the palate rising, the larynx moving forward, and the UES opening all happened earlier. The UES opened for longer and wider. The authors called this getting ready to receive the bolus.
• The main effect when the viscosity was increased went in a different direction. The time it took for the bolus to pass through the mouth and pharynx was delayed, the contraction squeezing the pharynx from top to bottom (pharyngeal peristaltic wave) became longer, and the UES opened longer and larger.

The authors' conclusion is that the amount and viscosity of the bolus affect swallowing in different ways (PMID 2333995). Increasing the amount brings forward the readiness to receive it, and increasing the viscosity slows down the flow.

The body closes the vocal cords first if the flow is fast. Viscosity changes when the airway closes. When swallowing, the larynx closes in multiple layers, and the deepest layer is the vocal cords that make sound, namely the true vocal cords. Ten healthy volunteers were captured in 3D using 320-row CT as they swallowed 10 mL each of honey-thick and thin barium. The thin liquid reached the hypopharynx, which is the lower part of the pharynx, faster and stayed longer.

Among the various movements of the larynx closing, the only one that changed significantly—that is, by an amount too large to be considered a coincidence—was true vocal cord closure. In the thin liquid, the vocal cords started closing earlier and stayed closed longer. The authors read this as a response to fast flow (PMID 22665214).

However, this tuning disappears when the disease deepens. In a study measuring 54 stroke patients swallowing 10 mL each of thick and thin liquids using swallow CT, the overall pattern came out the same as in healthy people. The thin liquid reached the vallecula (the concavity between the base of the tongue and the epiglottis), hypopharynx, and esophagus significantly earlier, and true vocal cord closure also started earlier (median 0.2 seconds vs. 0.3 seconds, p=0.020). But when divided by the severity of dysphagia, this regulation was significant only in mild cases, and did not appear in moderate and moderate-to-severe cases (PMID 41243023).

Putting the two observations side by side, the healthy pharynx closed the airway earlier in response to the fast flow of thin liquid, and this regulation was not seen in stroke patients with severe dysphagia. Whether thickening makes up for this regulation was not measured by these studies. Rather, in stroke patients, thickeners increased swallowing safety in a way the authors called a compensatory mechanism, without changing the timing of airway closure (PMID 26607158). Thickening is not the only option either. Posture during swallowing or the amount swallowed at once are used together, and there is also the path of weighing it against other risks, like "the approach permitting plain water only" to be seen later.

② Stages of Thickness — Names Have Been Unified, but Actual Thickness Is Not Yet All the Same

Until 2015, the biggest problem in this field was words rather than evidence. When 2,050 responses from 33 countries were compiled, there were 54 names for the firmness of food and 27 names for the consistency of liquids (PMID 27913916). The nectar consistency and honey consistency that appear in this article are such old names. Honey consistency is thicker than nectar consistency, close to a liquid very thick enough to be eaten with a spoon (PMID 30251253).

The International Dysphagia Diet Standardisation Initiative (IDDSI) is an international gathering created to sort out this confusion, and it has put forward an 8-level framework from 0 to 7. Each level is accompanied by a number, name, color, definition, and measurement method (PMID 27913916). The lower the number, the more watery and soft it is. Things to drink are placed in the lower levels, things to eat are placed in the higher levels, and regular meals belong to the highest levels. Therefore, the more food is ground or softened, the lower the level goes (PMID 34139151). Since the mixed use of multiple names itself is a risk to patient safety, the unification of names is an achievement in itself.

The problem lies in the fact that unifying names and unifying actual thickness—in other words, physical properties—are different things. A systematic literature review is a research synthesis that gathers and examines studies comprehensively using established procedures. In a systematic literature review that checked whether the IDDSI test method measures properly (validity) and produces the same value each time it is measured (reliability) using the Consensus-based Standards for the selection of health Measurement Instruments (COSMIN) evaluation criteria, there was only 1 study targeting elderly people with dysphagia. The conclusion was that, for now, the evidence is insufficient to recommend the IDDSI test method (PMID 33096747).

There are also parts that have been well verified. The flow test is a test in which a liquid is filled into a standard syringe, allowed to flow for a set time, and graded with the remaining amount (PMID 38613313). When a problem was revealed that the specific model of syringe required by this test could not be obtained equally around the world, the IDDSI funnel was created. This funnel showed complete agreement between the syringe reference test and level judgment in 67 out of 73 test conditions (92%) (PMID 38613313). This is an example of standardization working properly.

Outside of that, values fluctuate. First, if the measuring method is different, the value changes. The Bostwick consistometer is a device that spreads liquid in a horizontal groove and measures the distance it has traveled under its own weight for a set time. When 32 types of commercially available thickened beverages were measured respectively with this consistometer and the IDDSI flow test, the values of the two methods were very strongly linked (r=−0.93) but could not be used interchangeably. The consistency of values measured multiple times by the same person using the same method (intra-rater reliability) was excellent (ICC ≥ 0.99) (PMID 30370530). Each method measures consistently but does not yield the same values as one another.

Second, even when following the manufacturer's instructions, it differs from the label. Thickeners include starch-based ones made from starch and gum-based ones made from gums that become sticky when dissolved in water, such as xanthan gum. When 4 concentration labels were made and measured according to instructions with two products, starch-based and gum-based, the actual thickness depended heavily on the type of thickener, and the starch-based one was more watery than indicated at low concentrations (PMID 34348310).

Third, when people classify, it differs again. 68 medical and non-medical personnel studied the IDDSI training materials alone for 30 minutes and then classified 21 samples. Knowledge scores increased, but classification accuracy was generally low, and among them, thick liquids (Levels 1, 2, and 3) were the most inaccurate (PMID 32510986).

Even at the same level, bodily responses can be different. Thick liquids have the property of becoming more watery the faster they are made to flow (PMID 39739331). Therefore, viscosity is recorded together with how fast it was made to flow when measured, in other words, the shear rate. Shear viscosity is the viscosity measured at a designated shear rate, and the unit is mPa·s.

The same IDDSI level does not mean the same physiological response. Levels 1, 2, and 3 were made with xanthan gum and carboxymethylcellulose, respectively, and swallowed by 10 healthy adults. At a shear rate of 50 s⁻¹, the viscosities of the two products were almost identical (74.3/161.2/399.6 versus 78.0/176.5/429.2 mPa·s). However, at the faster 300 s⁻¹, the carboxymethylcellulose side was about double (19.5/34.4/84.8 versus 41.3/80.8/160.2).

When actually swallowed, the flow resistance of the carboxymethylcellulose side was also greater. The pressure caught in the throat when pushing the bolus (intra-bolus pressure) and the pressure remaining while the UES relaxes (UES residual pressure) were higher, and the time the UES was relaxed and the time the bolus stayed were shorter. The authors' conclusion is that these physiological responses are linked to shear viscosity, not the IDDSI level (PMID 39739331).

To summarize, what IDDSI has standardized are names and test procedures. Whether those procedures yield the same physical properties even when people, tools, and thickeners differ, and whether those physical properties produce the same responses in the body, have still only been partially verified.

③ In the Testing Room, Aspiration Actually Decreases

The result that thickening immediately reduces aspiration points in the same direction across varying study designs. This immediate effect is mainly measured using imaging tests like videofluoroscopy. This is because aspiration does not always reveal itself with a cough. Silent aspiration is aspiration that occurs without outward signs such as coughing, throat-clearing, or voice changes.

Imaging also looks at the steps before aspiration. Penetration is when food or water enters the larynx but stays above the vocal cords. The Penetration-Aspiration Scale (PAS) is a score that grades how deeply the entered material went down and whether it was pushed back out across eight stages, and the higher it is, the worse it is.

A meta-analysis is a method that combines the results of multiple studies to calculate a single value. An odds ratio (OR) is a value comparing the likelihood of an event occurring in two groups, and if it is less than 1, it means the event happened less in the thickened group.

• Double-blind prospective trial (100 people). Double-blind is a design where neither the patients nor the researchers knew which condition was which. Patients with dysphagia were given thin barium, starch-thickened barium, and gum-thickened barium in random order and underwent fluoroscopic examination. The number of people who aspirated was 20, 15, and 11, respectively (p<0.05, thin liquid vs. gum), and the mean PAS dropped from 2.11 to 1.76 to 1.42 (PMID 24161369).
• 122 chronic stroke patients. The proportion of safe swallowing on videofluoroscopy increased from 30.25% in liquid to 92.64% in spoon-thick with a modified starch thickener, and from 29.12% to 89.91% with a xanthan gum thickener (p<0.001) (PMID 26607158). Spoon-thick is a consistency thick enough to be scooped with a spoon.
• 128 people across four groups with different causes. When viscosity was adjusted to 250–1000 mPa·s using a starch and gum mixed thickener, safe swallowing was 74.19–96.67% in the elderly, Parkinson's disease, and stroke groups, while only the head and neck cancer group was 58.06% (PMID 32585942).
• Meta-analysis (2025). Across 16 papers covering adults with swallowing disorders caused by neurological diseases, thickened liquids reduced aspiration events (randomized trials OR 0.49; 0.28–0.88) and unsafe swallowing (0.27; 0.14–0.51), and lowered PAS (mean difference −1.99) (PMID 40693645).
• Systematic review update (2026). Starting from 63,247 cases and organizing 70 papers according to the IDDSI framework, all 25 papers reporting safety outcomes showed that thicker liquids improved swallowing safety compared to thinner liquids (PMID 42503522).

Therefore, the result that thickening immediately reduces aspiration points in the same direction across multiple designs. However, the authors of meta-analyses pooling these data report that the evidence is still weak because it relies largely on non-randomized studies (PMID 40693645), and a Cochrane review covering dementia patients notes that both immediate and long-term effects are uncertain (PMID 30251253). The debate begins from then on.

④ Instead, what remains in the throat increases

The same manipulation increases something else at the same time. Residue is food or liquid that remains in the mouth or pharynx after a swallow.

A 2015 systematic literature review (36 final studies out of 10,147) summarizing the effects of consistency adjustment on swallowing drew out two trends. Thicker liquids reduce the risk of penetration and aspiration, but increase the risk of residue remaining in the pharynx after swallowing. The authors stated that the literature was not sufficient to draw specific viscosity boundaries tailored to these results (PMID 25343878).

A meta-analysis 10 years later attaches numbers to that trend. It is the result of the very analysis that reduced aspiration and PAS earlier. Penetration events were not significantly reduced (OR 0.40; 0.13–1.22, p=0.11), and thickened liquids were associated with increased pharyngeal residue (odds ratio 1.57; 1.20–2.06) and increased oral residue (odds ratio 2.87; 1.88–4.40). The authors' conclusion is cautious. It is that the evidence is still weak because it relies mostly on non-randomized studies (PMID 40693645).

The European Society for Swallowing Disorders white paper lists this trade-off as five items. Raising bolus viscosity (a) increases swallowing safety, (b) increases residue in the mouth and pharynx, which can lead to airway invasion after a swallow, (c) has effects on physiology that are controversial, (d) reduces taste, and (e) correlates with an increased risk of dehydration and lower quality of life. However, it added the caveat to (e) that the severity of dysphagia may be a confounding factor, namely a third factor making the two move together. The white paper's conclusion is that raising viscosity is a valid management strategy to reduce the risk of airway invasion, and that new thickeners are needed to avoid side effects of residue, taste, and compliance (PMID 27016216).

Two caveats are attached. First, the type of thickener determines residue. In a study of 122 chronic stroke patients, modified starch increased oral and pharyngeal residue at nectar and spoon-thick consistencies, but xanthan gum did not (PMID 26607158).

Second, residue does not increase in a healthy pharynx. When 38 healthy adults drank five levels from thin liquids to extremely thick liquids using a standard procedure, the odds of penetration decreased significantly as it got thicker. On the other hand, the sequence of movements, closure of the laryngeal inlet, contraction of the pharynx, and residue after a swallow did not change (PMID 31021676). In a study observing four patient groups with a thickener mixed with starch and gum, pharyngeal residue also did not increase (PMID 32585942). Whether residue increases is not determined by viscosity alone, and turned out differently depending on which thickener is used and who swallows.

Therefore, clinical practice uses the two levers together. In a 2026 systematic literature review, when using thickened liquids, making the volume swallowed at one time small helped reduce residue (PMID 42503522). This is where the discovery in section ① that volume and viscosity are distinct levers returns to clinical practice.

⑤ The Lab's Winner and Pneumonia Three Months Later — Two Paired Trials

The most important evidence in this area comes from two randomized trials from a large program. A randomized trial is a study where people are divided by a draw, like a lottery, to set conditions and compare them. The two trials involved patients with the same conditions; one measured the immediate effect in the lab, and the other measured pneumonia 3 months later.

Trial 1 — Immediate Effect (711 people). The subjects were 711 patients with dementia or Parkinson's disease who were confirmed on videofluoroscopy to aspirate thin liquids. Each patient was subjected to all three methods in a randomized order and compared within the same test. The three methods were the chin-tuck posture, nectar consistency, and honey consistency, and the chin-tuck posture is a posture of swallowing with the chin tucked toward the chest. Cases where thin-liquid aspiration disappeared on the spot were most frequent with honey consistency, followed by nectar, and then chin-tuck. Two things were further observed. In patients with the most severe dementia, the effectiveness of all three methods was the lowest, and what patients preferred most was, conversely, the chin-tuck (PMID 18230864). It is also pointed out that the order in which methods were applied during the test may have influenced the response (PMID 30251253).

Trial 2 — 3-Month Pneumonia (515 people). The same program followed 515 people for 3 months across 47 hospitals and 79 subacute care facilities. The target conditions were the same as in Trial 1, and subjects were randomly assigned to chin-tuck (259 people), nectar (133 people), and honey (123 people). The cumulative incidence is the proportion of people in whom the event occurred during a set period. The hazard ratio (HR) is a value comparing the speed at which the event occurs at a given point between two groups.

• Pneumonia occurred in 52 people, and the overall 3-month cumulative incidence was 11%.
• There was no difference between the chin-tuck group at 0.098 versus the thickened liquid group at 0.116 (HR 0.84; 0.49–1.45, P=0.53).
• Within the thickened liquid group, it was nectar 0.084 versus honey 0.150 (HR 0.50; 0.23–1.09, P=0.083).
• In the thickened liquid group, dehydration (6% vs 2%), urinary tract infections (6% vs 3%), and fever (4% vs 2%) were more frequent (PMID 18378947).

These results are easy to exaggerate, so we look at the limitations first.

1. The difference between honey and nectar is not statistically significant (P=0.083). The P-value is the probability that a difference this large would occur by chance when there is actually no difference. This value exceeded the commonly used threshold. It is not proven that "honey is worse."
2. There was no control group receiving no intervention. This trial does not answer whether any of the three methods is better than doing nothing.
3. The follow-up was only 3 months, the healthcare providers caring for the patients knew which group they were in, and the confidence intervals were wide. A confidence interval is the range where the true value is likely to lie.
4. The overall pneumonia incidence was much lower than expected. This is a limitation written directly by the authors, and the power to detect a difference is reduced accordingly.
5. The authors' concluding sentence is, "No definitive conclusion can be drawn that any of the tested methods is better."

Still, an observation remains. The condition that best eliminated aspiration in the lab did not lead the way in 3-month pneumonia.

A Cochrane Review is a systematic review by an international cooperative network that gathers and examines studies using established procedures. When this review re-analyzed only the dementia patients from the two trials separately, the same picture emerged. Honey consistency was better at eliminating aspiration on videofluoroscopy, but during the 3-month follow-up, it had more adverse effects (PMID 30251253).

The same review considered both trials to have a high risk of bias—meaning the risk that results were skewed to one side due to the design. The certainty of the evidence was low. The certainty of the evidence is a grade rating how much we can trust the pooled results, classified as high, moderate, low, or very low. The review recommends considering immediate benefits and long-term health impacts together, and notes that neither trial addressed quality of life (PMID 30251253).

Why the Test Lab and Three Months Later Can Differ

The hypotheses supported by these data are three. None are proven by these data.

① Residue can enter the airway after swallowing. This is the pathway specified by the white paper. Raising viscosity increases residue, and that residue can lead to airway invasion after the swallow (PMID 27016216). The test lab observes that exact moment of swallowing, while pneumonia is built in the time afterward.

② Harmful pathways other than pneumonia open up. In the same trial, dehydration, urinary tract infections, and fever were more frequent in the thickened liquid group (PMID 18378947).

③ The course for a person who aspirates thick substances may be worse. There is an older study reviewing the records of 26 stroke patients who aspirate and 33 dysphagia controls without aspiration. The patients were tested at an average of 2 months after stroke, and followed up for an average of 16 months after stroke. Aspirating any consistency yielded an odds ratio for pneumonia of 7.6, and this value hovered at the borderline of the significance threshold with p=0.05. For those who aspirated thick liquids or harder, compared to those with no aspiration or aspiration of only thin liquids, the odds ratio for pneumonia was 5.6 and the odds ratio for mortality was 9.2. The pneumonia side fell short of the conventional significance level at p=0.06, while the mortality side was p=0.01. On the other hand, dehydration had no relation to aspiration status (PMID 8131429). Because people without aspiration were mixed into the comparison group, this is not a direct comparison between aspirating thick versus thin substances, and the possibility that people who even aspirate thick substances were more severe cannot be ruled out with these data either.

What Pooled Analyses and Guidelines Say

In patients with confirmed aspiration, there was no significant difference in pneumonia risk between those who drank thin liquids with safety strategies and those who drank only thickened liquids. Among 2,465 papers, 7 papers (650 people) met the criteria, and this is the result of a meta-analysis combining 6 of them (OR 0.82; 95% CI 0.05–13.42; p=0.89). The confidence interval is so wide that it essentially rules out nothing. Above all, all 7 papers excluded patients with two or more pneumonia risk factors. As the authors state, this result generalizes only to patients with a low risk of pneumonia (PMID 28730887).

Guidelines are more cautious. A national clinical guideline addressing consistency modification in adult oropharyngeal dysphagia, namely dysphagia in the mouth and pharynx stages, issued a weak recommendation against thickened liquids in its 2018 edition based on 2 randomized trials. Because the quality of evidence is low, the recommendation is also a weak one. There were no significant differences in mortality, pneumonia, aspiration, dehydration, weight loss, or adherence. In patient preference, the proportion dissatisfied with honey-thick consistency was significantly higher than with thin liquids/chin tuck (risk ratio RR 1.18; 1.01–1.37). For meal performance and quality of life, not a single piece of evidence was found (PMID 28939270).

In the second update in 2022, 1 new randomized trial was added to make a total of 3, but the recommendation did not change. The concluding sentence of the guideline is as follows: "There is no persuasive evidence that thickened liquids or consistency-modified diets prevent mortality or pneumonia, or improve quality of life, nutritional status, or oral intake" (PMID 35623866).

The same guideline judged liquids and foods separately. It issued a weak recommendation against thickened liquids, but for the food question, there was not a single piece of evidence meeting the criteria. Therefore, instead of a graded recommendation, weighing risks, benefits, and patient preferences pointed toward using consistency-modified foods as "good clinical practice" (PMID 28939270). The second update notes that there is no recommendation for food (PMID 35623866). The lack of evidence does not mean lack of effect. The observation of a 2026 systematic review also has a different grain. Food consistency rarely caused penetration and aspiration, but in the patient group, it was associated with increased residue (PMID 42503522). The results from liquids and the results from foods must be read separately without mixing them.

⑥ What Is Paid When Dropping to a Lower Level

The very first thing that shrinks is the water people drink. In a study reviewing the records of 69 hospitalized patients prescribed thickened liquids after a stroke, the average daily intake was 781 mL. The acute phase (519 mL) was significantly lower than the rehabilitation phase (1,274 mL), and the average daily intake fell short of the recommended water intake for hospitalized adults (PMID 24007386).

However, it is inaccurate to blame this entirely on viscosity. In a study measuring 39 stroke patients divided into a thin liquid group and a thickened liquid group for 72 hours, the thin liquid group drank significantly more (1,405 versus 907 mL). However, about 1,000 mL more was provided to them (2,575 versus 1,589 mL). And in the thickened liquid group, deficits in cognitive functions like memory and problem-solving predicted intake. The authors concluded that, regardless of viscosity, stroke patients' water intake was insufficient on average (PMID 25565596).

Dehydration is common. Among 218 hospitalized patients aged 70 and older with dysphagia, 78.9% based on the BUN/creatinine ratio and 81.2% based on osmolality—two indicators calculated from blood tests—were dehydrated (PMID 39576562). However, saying that thickened liquid treatment itself causes dehydration goes beyond the evidence.

There are two parallel literature reviews that tackle this question head-on. The proportion of dehydrated individuals among patients with dysphagia has been reported to be 19 to 100%. There were 2 high-quality studies (total 724 people) in which thickened liquid treatment showed a positive effect on hydration status. At the same time, in 9 out of 10 studies evaluating intake, thickened liquid intake fell short of basic water requirements. The review advises monitoring drinking amounts strictly while using thickened liquids (PMID 35745228).

Among 208 older adults admitted to a convalescent rehabilitation hospital, poor appetite was more common in those eating texture-modified diets (IDDSI 3–5). Even in a regression analysis that considered other factors together, texture-modified diets were independently associated with poor appetite (OR 3.443, P=.011) (PMID 34139151). Sarcopenia is a condition where muscle mass and strength decrease together. Among 443 people aged 65 and older undergoing rehabilitation after a stroke, malnutrition was 55.3% and sarcopenia was 62.1%, and both were more common the lower the IDDSI level eaten (trend P<0.001) (PMID 35418607).

Because both studies are cross-sectional studies measured at a single point in time, it is impossible to say which came first. The possibility that people with more severe dysphagia ate lower levels and also had poorer nutrition—in other words, the possibility that severity drove both together (confounding)—cannot be ruled out. However, in the appetite study, the association between texture-modified diets and poor appetite remained even after considering other factors together, and the authors suggest checking whether appetite improves when the appearance, taste, and nutrition of food are improved (PMID 34139151). Therefore, treating texture modification as a 「side-effect-free safety measure」 does not align with these data.

There are also the words of the patients themselves. When 14 adults prescribed thickened liquids after a stroke were interviewed in depth, three themes emerged: 「uncertainty」, 「unpleasant experiences」, and 「trade-offs」. Because there was a lack of experience of listening to explanations about why it is prescribed and deciding together, uncertainty was large, and this dislike may have affected adherence to the prescription, water intake, and quality of life. Even so, some understood that it was necessary for recovery (PMID 28621030).

There are data in the opposite direction as well. The conclusion that texture modification itself ruins nutrition goes beyond the evidence. In a randomized controlled trial of 40 patients with oropharyngeal dysphagia aged 65 and older, one group ate a diet for 12 weeks that accompanied texture-modified food with nectar- or pudding-consistency drinks and even controlled the amount swallowed at one time. Unlike the standard-treatment control group with the same calories, this group showed increases in energy intake (29 → 40 kcal/kg, p=0.009), protein intake (1.3 → 1.8 g/kg, p=0.03), body weight (56 → 60 kg, p<0.001), and grip strength (18 → 21 kg, p=0.004). The control group showed no change (PMID 30643223). It is a small trial, but the direction is important. The problem may lie not in texture modification itself, but in lowering only the texture while failing to design the rest.

⑦ Should Plain Water Be Allowed?

On the opposite side of "Should we thicken it?" is "Should water be allowed just as it is?". The Free Water Protocol is a method that allows only plain water under strict conditions to people who aspirate thin liquids. The evidence is divided.

Here is data from the positive side. In a trial that randomly assigned rehabilitation patients with confirmed thin-liquid aspiration after a stroke (14 analyzed), the diagnosis of urinary tract infection was significantly higher in the group that drank only thickened liquids (p=0.024), and pneumonia was 0 in both groups (PMID 26886370).

Here is data from the negative side. In a study observing 76 patients with dysphagia for 5 days, pulmonary complications in the group allowed to drink water were 14.3% (6/42) and 0/34 in the control group. The authors picked out patients with neurodegenerative diseases who cannot move well as the highest-risk group. And they recommended that acute and severe patients stick to thickened liquids, while relatively mobile subacute patients should choose for themselves after receiving a full explanation of the risks (PMID 21356121).

Common Misconceptions

"Making it thick makes it safe." Safety at that moment actually improves. It did so in all 25 papers that reported safety outcomes (PMID 42503522). However, that "safety" is the safety of penetration and aspiration indices. In a meta-analysis gathering swallowing disorders caused by neurological diseases, thickened liquids were associated with increased pharyngeal residue (odds ratio 1.57) and oral residue (odds ratio 2.87) (PMID 40693645), and in 3-month pneumonia, the condition that was best at the moment did not come out ahead (PMID 18378947).

"If the IDDSI levels are the same, they are the same." Even at the same level, viscosity at fast flow showed a twofold difference depending on the thickener ingredient, and physiological responses followed actual shear viscosity rather than the level (PMID 39739331). Even when made according to the manufacturer's instructions, the label and the actual thickness mismatched (PMID 34348310).

"Making it thickest makes it safest." Aspiration clearance at the moment was best with honey consistency (PMID 18230864), but the 3-month cumulative incidence of pneumonia was 0.150 for honey versus 0.084 for nectar. However, this difference was not significant at P=0.083 (PMID 18378947). Patient dissatisfaction was significantly higher with honey (PMID 28939270).

"Texture-modified diets ruin nutrition." Cross-sectional studies show associations between lower IDDSI levels and decreased appetite, malnutrition, and sarcopenia (PMID 34139151, PMID 35418607). Although chronological order cannot be stated and confounding by severity is not ruled out, the association with decreased appetite remained after considering other factors. Meanwhile, in a randomized trial of a texture-modified diet that even designed the swallow volume per bite, intake, body weight, and grip strength actually improved (PMID 30643223). Texture modification can neither be concluded as harmful nor as burden-free.

"Adult data can be used for children as is." There is a case study reviewing the records of 50 children at a single institution in whom aspiration was confirmed. This institution used its own protocol of reducing thickeners by 10% every two weeks while watching symptoms, and 78% (39/50) completely returned to thin liquids (average 0.9 years for successful children) (PMID 29121147). This is a study without a control group, and 2 out of 6 who failed to wean and 2 out of 39 who succeeded experienced pneumonia (PMID 29121147). Because children and adults are different populations, data from one side is not transferred to the other as is.

Honest Unresolved Questions

• The most suitable viscosity has not been established. There is not enough literature to draw viscosity boundaries tailored to clinical outcomes (PMID 25343878), and responses vary by the cause of dysphagia. In the same viscosity range, the safe swallowing rate for the head and neck cancer group was lower than that of the other three groups (PMID 32585942).
• Validation of testing methods is still lacking. Evidence is insufficient to recommend the IDDSI test method (PMID 33096747), values between methods are not interchangeable (PMID 30370530), and when humans classify, the slightly thick liquid range is precisely the most inaccurate (PMID 32510986).
• There is no persuasive evidence that it changes final outcomes. Despite two guideline updates, there are only three randomized trials, and a weak recommendation against it is maintained (PMID 35623866, PMID 28939270).
• Quality of life was not even measured. Both large-scale trials did not address quality of life (PMID 30251253), and in guideline reviews, there was zero evidence regarding meal performance and quality of life (PMID 28939270).
• The clinical benefit of amylase resistance has not been verified. Amylase is an enzyme in saliva that breaks down starch. Not a single study has confirmed the message "thickened products resistant to amylase are safer" through a direct comparison (PMID 35635111).
• The evidence for adults on when to stop is thin. While there are reports of systematic weaning procedures for children (PMID 29121147), in the data gathered for this article, we could not find reports of the same procedure being used for adults.

To Go Deeper

• When reading a paper — first separate whether the study measured physiological indicators on the spot (PAS, residue) or clinical events months later (pneumonia, death). A paired pair of trials is a representative example showing that difference (PMID 18230864, PMID 18378947).
• When looking at viscosity notations — papers that only list the IDDSI level and those that together report mPa·s values differ in precision. The observation that physiological responses followed actual shear viscosity rather than levels is the reason why (PMID 39739331).
• When looking at outcome measures and subjects — studies that report only safety (penetration, aspiration) and do not look at efficiency, namely whether it went down without leaving anything behind (residue), show only one side of the trade-off (PMID 25343878). Also check what cause of dysphagia it is, and how many pneumonia risk factors the people had. The earlier meta-analysis excluded patients with two or more risk factors in all seven papers (PMID 28730887).
• If leaving just one — changing viscosity is not about eliminating risk, but about trading one risk for another, and which one is greater depends on what is measured and for how long.

Source Papers

Both design and cohort are noted. Figures already present in the text are not repeated. ⚠️ indicates data in the text flagged for reverse causality or confounding.

Scale — PMID 27913916 (2017) Rationale and development of the IDDSI framework · PMID 29030310 (2018) Cross-section of 23,549 individuals across 926 care units in 19 countries · PMID 39576562 (2025) 218 hospitalized patients aged 70 and older with dysphagia.

Volume and Viscosity Are Different Levers — PMID 2333995 (1990) Simultaneous videofluoroscopy-manometry in 10 healthy individuals, the source paper for this section · PMID 22665214 (2013) 320-row CT in 10 healthy individuals, true vocal fold closure responds to flow velocity · PMID 41243023 (2026) Swallowing CT in 54 stroke patients, that regulation disappears depending on severity.

IDDSI — The Inside and Outside of Standardization — PMID 33096747 (2020) COSMIN evaluation of test validity and reliability, only one study in older adults · PMID 38613313 (2024) Agreement between the IDDSI funnel and syringe reference tests · PMID 30370530 (2019) 32 commercially available beverages, Bostwick and flow tests are not interchangeable · PMID 34348310 (2022) Measured values of four concentrations prepared per manufacturer instructions · PMID 32510986 (2020) Classification accuracy of 68 medical and non-medical personnel · PMID 39739331 (2025) Study where physiological responses diverge at the same IDDSI level but different shear viscosities.

Immediate Effects — PMID 24161369 (2014) Double-blind prospective study of 100 individuals · PMID 26607158 (2016) 122 chronic stroke patients, safety and residue by thickener type · PMID 32585942 (2020) 128 phenotypes, lower only in the head and neck cancer group · PMID 40693645 (2025) Meta-analysis of 16 studies on neurogenic dysphagia · PMID 42503522 (2026) Updated systematic review mapping 70 studies to the IDDSI framework.

The Trade-off Called Residue — PMID 25343878 (2015) Systematic review of 36 studies on the physiological effects of consistency adjustment, viscosity boundaries undetermined · PMID 27016216 (2016) European Society for Swallowing Disorders white paper · PMID 31021676 (2019) Residues did not increase in 38 healthy adults.

The Paired Trials — PMID 18230864 (2008) Immediate effects in 711 individuals, randomized crossover of three interventions · PMID 18378947 (2008) 3-month pneumonia in 515 individuals, parallel randomized trial · PMID 30251253 (2018) Cochrane review reanalyzing dementia patients from both trials.

Higher-Tier Evidence — PMID 8131429 (1994) 26 stroke aspirators and 33 controls followed for an average of 16 months, distinguishing what is aspirated · PMID 28730887 (2017) Meta-analysis of 650 patients with confirmed aspiration, ⚠️generalizable only to low-risk pneumonia groups · PMID 28939270 (2018) Clinical guideline for consistency adjustment · PMID 35623866 (2022) Second update of that guideline.

The Price Paid at Lower Tiers — PMID 24007386 (2014) Retrospective audit of 69 hospitalized patients prescribed thickened liquids · PMID 25565596 (2015) 39 stroke patients for 72 hours, ⚠️presence of confounding differences in administered amounts · PMID 35745228 (2022) Parallel review on dehydration, reporting 2 positive studies alongside 9 studies on insufficient intake · PMID 34139151 (2021) Appetite in 208 convalescent rehabilitation patients ⚠️cross-sectional · PMID 35418607 (2022) Malnutrition and sarcopenia in 443 post-stroke patients ⚠️cross-sectional · PMID 28621030 (2018) In-depth interviews with 14 patients · PMID 30643223 (2019) Randomized controlled trial of consistency-adjusted meals including bolus volume design in 40 patients.

Water — PMID 26886370 (2016) Free water protocol randomized allocation, 14 analyzed · PMID 21356121 (2011) Pulmonary complications in 14.3% of the water-access group versus 0/34 in the control group.

Remaining Gaps — PMID 35635111 (2022) Lack of studies directly comparing the clinical benefits of amylase resistance · PMID 29121147 (2018) Systematic weaning procedure in 50 children.