Is Aspiration Pneumonia Really Caused by Aspiration?

Medical disease names usually describe a location. Pneumonia is inflammation of the lung, and a stroke is an event that occurred in the brain. The name aspiration pneumonia does not describe a location, but a cause. Aspiration pneumonia is a name that refers to pneumonia caused by something entering the airway. So the moment this name is spoken, the causal claim that what entered caused the pneumonia is already made.

Here, aspiration is, as defined in the primer "Swallowing, Speech and Language — The Crossing Where the Food Path Meets the Airway", food, water, or saliva entering the airway below the vocal cords. When it happens without signs like a cough, a throat-clearing, or a voice change, it is silent aspiration. That primer noted that the mere co-occurrence of dysphagia and pneumonia cannot establish causality, and briefly introduced a review on aspiration in older adults suggesting that while dysphagia and pneumonia are important factors, aspiration alone does not cause pneumonia without other conditions. This article examines that single line in depth.

If it is a disease caused by entering, you can simply prevent things from entering. Fasting to prevent eating anything by mouth, thickened liquids, a nasogastric tube inserted through the nose down to the stomach, and a gastrostomy tube inserted directly into the stomach through the abdominal wall are measures in that direction. This article asks two things. Does the mere act of entering result in pneumonia? And do measures that prevent entry actually reduce pneumonia?

The answers to these questions do not yet exist, and this article establishes why those answers do not yet exist. The fact that causality is unconfirmed does not mean we should abandon care aimed at preventing pneumonia. Guidelines from the European Stroke Organisation (ESO) treat swallowing screening to catch dysphagia and swallowing management as already established parts of care to prevent post-stroke pneumonia. However, the same guidelines note that most of their recommendations rely on expert consensus (PMID 42095755). We will also look at the fact that the data supporting that care mostly consists of correlations or low-certainty evidence.

Reading Rules — Almost All Studies in This Piece Are Observational Studies

Observational studies are research that gathers data by watching what happens to people without interfering. A randomized controlled trial is a study where people are divided by a draw like a lottery, the intervention is applied to only one side, and the two sides are compared. Most of the numbers you will see ahead come from observational studies.

The trap to be most careful about in observational studies is confounding. Confounding is when the relationship between two things looks distorted because of another factor that reaches both the side that looks like the cause and the result. This concept is built out in detail in "If People Who Take It Live Longer, How Were They Different to Begin With? — Confounding".

One type of confounding plays an especially large role in this topic. Confounding by indication is confounding where sicker people receive that intervention, so it looks as though the intervention changed the outcome. Sicker patients are put on fasting, and sicker patients get tubes inserted. Then the observation that "people who fasted got pneumonia more easily" could be the effect of fasting, or it could be the shadow of the fact that those people were sicker to begin with.

Often, the data alone cannot separate the two interpretations. That is why this article attaches a ⚠️ to every such passage. A sentence with a ⚠️ attached is a sentence to be read only up to the point of "this kind of association was observed."

Why We Need to Know — A Heavy Disease with Shaky Definitions

Aspiration pneumonia is a heavy illness. In data from the Beijing Urban Employee Medical Insurance, which covers roughly 12 million people, the 6-month all-cause mortality rate for people hospitalized with aspiration pneumonia was 35.2%. All-cause mortality counts deaths regardless of the cause. For community-acquired pneumonia, which is pneumonia caught and hospitalized in everyday life outside the hospital, it was 11.1% (PMID 37270854).

However, the two groups started from different places. The Charlson Comorbidity Index, which scores the burden of other concurrent illnesses, was more than double for the aspiration pneumonia group at 7.72 compared to 2.84 for the community-acquired pneumonia group (PMID 37270854). The gap in mortality includes both the weight of the pneumonia and the baseline condition of those patients to begin with. Narrowing it down to after a stroke, stroke-associated pneumonia occurs in about 12% of cases and is the most common infectious complication following a stroke (PMID 42095755).

Yet this heavy disease lacks agreed-upon diagnostic criteria. The diagnostic criteria are not well-defined, and there is no unique blood test marker (biomarker) that isolates this disease alone. The proportion of aspiration pneumonia among people hospitalized with community-acquired pneumonia is also estimated across a wide range of 5 to 24% (PMID 39612934).

The fact that what counts as aspiration pneumonia is shaky becomes evident when counting the same patients using two methods. In the electronic health records of 15,603 pneumonia hospitalizations aged 65 and older, aspiration pneumonia counted by discharge diagnosis codes was 15.57%. When re-estimating by selecting people with oropharyngeal Dysphagia—meaning swallowing goes awry at the mouth and throat stage—using a machine learning algorithm on the same records, it was 25.32% (PMID 40993813). These are the same patients, only the counting method changed. The latter value is an estimate that views pneumonia with Dysphagia as aspiration pneumonia.

Another name intertwines here. Aspiration pneumonitis is inflammation caused by chemical irritants such as stomach contents, while aspiration pneumonia is an infection. Although their treatment approaches differ, the two are difficult to distinguish clinically. When terminology is unclear, it is hard to compare studies against each other, which in turn makes it difficult to build evidence-based conclusions or clinical guidelines (PMID 28879323).

The Big Picture — Pneumonia Requires Several Factors to Align

This article views aspiration pneumonia as an event that requires three overlapping conditions to occur.

1. Entry — The contents of the mouth and throat or stomach cross past the vocal cords into the airway.
2. Content of what enters — How many bacteria that cause pneumonia, in other words, respiratory pathogens, are present in the material that passed through.
3. The defensive side — Whether the body's defenses (coughing, airway mucus clearance, lung immune cells, whole-body immunity) handle it.

The first two axes are included directly in a review article that re-proposes the definition of aspiration pneumonia. To diagnose aspiration pneumonia even when aspiration is not witnessed, expert opinion states that at least one risk factor for aspiration and at least one risk factor for bacteria colonizing the mouth must be present together. Diagnoses when aspiration is witnessed remain as they are (PMID 34087609).

① Entering Is Not Enough

The oldest clue came from people without pneumonia. A 1994 study used a radioactive material called indium-111 on 14 elderly community-acquired pneumonia patients and 10 age-matched controls to track silent aspiration during sleep. 71% of the pneumonia patients aspirated, and 10% of the controls aspirated as well (PMID 8025758). The authors read these results as evidence that silent aspiration matters in elderly pneumonia. The fact that 10% of the control group was aspirating without pneumonia is the counter-question that follows for us. It is an old study with a very small sample, but the direction aligns with later data.

More recent data broadens this counter-question. Microaspiration is when an amount so small it goes unnoticed enters the airway, and it is a common occurrence even in healthy people. The same research team had previously shown that in about half of healthy people, the fluid washed out from deep within the lungs (bronchoalveolar lavage fluid) contains many oral bacteria. Analyzing 49 people without lung disease, this state was linked to inflammation signals, an increase in Th17 lymphocytes, and a blunted TLR4 response in lung immune cells (alveolar macrophages) (PMID 27572644). This means the mere fact that something enters the airway does not allow us to say disease is present.

Clinical data points in the same direction. In a flexible endoscopic evaluation of swallowing—where we look inside the throat with a thin endoscope while watching swallowing happen—73 people who showed penetration or aspiration at least once were followed for an average of 3.87 years. Penetration is when what entered stays above the vocal cords, and the Penetration-Aspiration Scale (PAS) is a score that rates how deeply it went down, a term established in "How Do We Know Swallowing Has Failed". 41 people (56%) developed pneumonia and 28 (38.4%) died. Yet the PAS score had no significant association with either pneumonia or death (PMID 40088308).

The odds ratio (OR) is a value comparing the likelihood of an event occurring between two groups, and when it exceeds one, it means the event happened more readily when that factor was present. In the same data, what predicted pneumonia was a prior history of pneumonia (OR 1.374) and the Murray secretion scale, which grades secretions pooling in the throat (OR 1.121) (PMID 40088308). Whether one had suffered from pneumonia before and whether secretions were pooled screened the future better than the depth of aspiration.

In a study evaluating 148 nursing home seniors with videoendoscopy during meals and following them for 3 months, 8.1% developed pneumonia during the 3 months. The examination items included food aspiration as well, but what predicted pneumonia and a weight loss of 3% or more together was "saliva aspiration" or "silent saliva aspiration" (PMID 22329364). However, because these seniors were followed while receiving feeding therapy tailored to the test results, food aspiration may have already been managed (PMID 22329364). This is a clue like "People Who Are Already Using Umbrellas," which we will see later. Therefore, this result is read as the first signal that what enters can take precedence over how much enters.

② What Goes In — The Mouth Is Half the Problem

The inside of the mouth is different for a person with dysphagia. When researchers compared 47 frail older adults with oropharyngeal dysphagia and 14 older adults without dysphagia, oral health was poor in both groups (90% had periodontitis, 72% had tooth decay). However, the rate at which respiratory pathogens had taken root in the mouth (oral colonization) split: 93% in the dysphagia side versus 67% in the control side. The authors therefore consider these older adults to be at risk for aspiration pneumonia (PMID 26416412).

There are also data showing that the composition of oral bacteria leads to outcomes. Researchers classified the tongue bacteria of 173 nursing home residents using genes (16S rRNA) and followed them for a median of 19 months. A hazard ratio (HR) is a value comparing the speed at which that event happens at a given time between two groups. The group with dominant Prevotella and Veillonella had a higher risk of pneumonia-related death than the group with dominant Neisseria and Fusobacterium (adjusted HR 13.88) (PMID 29053769). ⚠️ However, the 95% confidence interval, which is the range where the true value is likely to lie, is extremely wide at 1.64–117.21. This means the estimate is unstable because the number of events was small, so the middle value should not be read as is.

There is also a classic study that measured this axis head-on. A prospective study from Michigan, USA, in 1998 followed 189 older adults for up to 4 years, comparing different categories of risk factors all at once. The strongest predictors of pneumonia were feeding dependency, oral care dependency, the number of decayed teeth, tube feeding, two or more diagnoses, the number of medications taken, and smoking. Feeding dependency is a state where someone else must feed the meal, and in a model excluding tube-feeding patients, the odds ratio for this factor was an overwhelming 19.98 (PMID 9513300).

The authors' conclusion is the axis of this article. Dysphagia is an important risk for aspiration pneumonia, but it is generally not sufficient to cause pneumonia unless other risk factors are present together (PMID 9513300). The risk also lay outside of swallowing, in the mouth and in everyday dependency.

③ Who Blocks the Way — Stroke Also Lowers the Body's Defenses

The third axis is the understanding that has changed the most over the past 20 years. Stroke does not just break swallowing; it also suppresses the immunity of the entire body.

The cleanest clinical evidence is the PREDICT study. It prospectively followed 486 patients with acute ischemic stroke across 11 institutions in Germany and Spain, and the pneumonia incidence was 5.2%. This study used a lower amount of a protein called HLA-DR on the surface of monocytes, which are immune cells, as a marker that immunity is suppressed. In an analysis incorporating multiple factors together, dysphagia and low HLA-DR each independently predicted stroke-associated pneumonia (PMID 27733675). This means that even after adjusting for one factor, the contribution of the other remained separate.

Numbers show this independence. When patients were divided into four groups by their HLA-DR value, pneumonia was 0.9% in the highest group and 8.5% in the lowest group. When dysphagia overlapped with this, it rose to 5.9% and 18.8% respectively. Stroke-associated pneumonia did not occur in patients without dysphagia and with normal HLA-DR (PMID 27733675). Pneumonia was most frequent when swallowing and immunity collapsed together.

This immunosuppression follows the size of the brain injury. In a study that repeatedly measured the immune cells of 50 ischemic stroke patients and 40 controls, the infarct volume—the size of the brain tissue that died because blood flow was cut off—was the main determinant of lymphocyte reduction and monocyte dysfunction, and infarct volume also remained an independent early predictor of respiratory infection (PMID 19661470). This state is organized as a two-phase response where peripheral immunosuppression follows initial inflammation. One question is still open: whether this immunosuppression is an adaptation to protect the damaged brain, or a collateral consequence caused by a broken balance between the nervous system and immunity (PMID 42212161).

⚠️ Animal data must be read only as context. In mice where a stroke was created by blocking the middle cerebral artery, a major artery of the brain, just putting 200 CFU (a unit for the number of live bacteria) of Streptococcus pneumoniae into the nose caused severe pneumonia and bacteremia. To produce the same degree in mice that only received sham surgery, 200,000 CFU was needed, and this progression was prevented by beta-adrenergic receptor blockade, which blocks sympathetic nerve signals (PMID 16946159). The contrast of 1,000 times is striking, but it is not human data.

In other species, the picture wavers. In a rat model, sham-surgery animals also showed immune changes and bacterial proliferation in the lungs similar to stroke animals. The authors reported that the stress of surgery and anesthesia may be the cause, and concluded that this model is not suitable for stroke-induced infection research and the results should be interpreted with caution (PMID 24922549). Data looking at beta-blockers in humans were also negative. In a registry of 306 severe stroke patients who underwent thrombectomy, beta-blocker treatment did not reduce pneumonia (OR 0.78), urinary tract infection, sepsis, or death (PMID 29694433).

Trying to Block It 1 — Fasting and Dietary Restrictions

If the three-axis framework is correct, measures that touch only one axis should not reduce pneumonia significantly. The data have generally turned out that way.

The most counterintuitive data lie on the fasting side. When retrospective records of 160 patients with post-stroke dysphagia at two university hospitals were analyzed, 63% of the patients who developed stroke-associated pneumonia were already fasting (NPO) upon admission, and only 3.7% were eating without restrictions. The rate at which fasting patients contracted pneumonia was 2.5 times that of patients without dietary restrictions (P=0.008). Most pneumonia was diagnosed before patients ate anything by mouth, and many cases were within 3 days (PMID 41044338). Identified as risk factors for pneumonia in the same data were stroke severity, male sex, and age. The authors conclude that this data does not support fasting or modified-texture diets as pneumonia prevention measures, and instead write that early, comprehensive care is needed. They close the article with the possibility that an appropriate dysphagia diet that utilizes remaining swallowing capacity can help both pneumonia rates and swallowing rehabilitation (PMID 41044338).

⚠️ This is observational data with heavy confounding by indication. It is nearly certain that sicker patients were prescribed fasting, and that alone could account for a 2.5-fold difference. However, the temporal sequence showing pneumonia was diagnosed before eating by mouth is a different type of information from confounding. Pneumonia that arrives in the hospital before oral feeding begins is not explained by aspiration from that diet. This does not rule out aspiration around the time of the stroke or prior to admission. The authors also place these results alongside evidence emphasizing the roles of oral hygiene and immune changes (PMID 41044338).

What about the side that changes the texture and consistency of food, namely consistency adjustment? In a retrospective analysis of 654 cases of dysphagia following head and neck cancer surgery, the proportions diagnosed with or suspected of aspiration pneumonia after starting oral feeding were 9.3% in the group with no beverage restrictions, 15.9% in the group allowed only thickened liquids, and 17.6% in the group with beverage restrictions. The authors' conclusion is that restricting beverages may not reduce the risk of aspiration pneumonia (PMID 36775772).

⚠️ Confounding in this study is particularly heavy. The proportions of those with severe swallowing impairment (swallowing toxicity grades 3–4) were 2.8% in the unrestricted group and 53.3% in the restricted group. This means much sicker patients received restrictions. The authors performed both multivariable analysis and propensity score matching—pairing patients with similar conditions—and report that both beverage restrictions and allowing only thickened liquids worked unfavorably against aspiration pneumonia (PMID 36775772). Still, adjustment can only correct for recorded differences, so what these numbers say is limited to: "The direction that restricting reduces it does not appear in this data." Data directly comparing consistency adjustments through a randomized trial are reviewed in "The Decision to Change Viscosity — What Thickening Actually Changes".

A 2025 review on recurrence prevention summarizes the current position of this branch. It organizes that while no single treatment has shown clear efficacy in directly preventing aspiration pneumonia, measures such as a semi-recumbent position (for patients with altered consciousness or tube feeding), speech therapy, rehabilitation, and removal of dental plaque were effective in prevention (PMID 40499817). In the place where a single drug could not block it, multiple care measures remain as evidence. Which measure comes first is a research task for the future, the authors write (PMID 40499817).

Blocking It 2 — The Decision to Insert a Tube

Stopping eating makes a tube necessary, and a tube adds one more variable to weigh.

In an Australian hospital dataset analyzing past records of 536 acute stroke patients, the factors significantly associated with respiratory infection in the first week were fasting (OR 5.62) and nasogastric tube insertion (OR 3.91). Total dependence for mobility at admission (OR 6.48) and urinary incontinence (OR 3.21) were also associated (PMID 25402187). The authors report that this result supports the research showing aspiration pneumonia is a disease where multiple factors overlap, and write that a nasogastric tube may play a part in infection by encouraging bacteria to colonize the mouth and throat (PMID 25402187). ⚠️ This is retrospective data, and here too, tubes go to sicker patients.

A tube does not appear to increase aspiration by a single sip. 147 people receiving nutrition through a nasogastric tube swallowed 1 mL of liquid each, with the tube inserted and with it removed, during a videofluoroscopic swallowing study using X-ray imaging. There was no significant difference in the penetration-aspiration score (2.45 versus 2.57, P=.50) (PMID 29619559). However, what this study measured was only the depth reached during a single swallow of 1 mL of liquid. If tubes are associated with infection, the pathway may be the oral bacteria mentioned by the previous authors, and there is also a review stating that receiving nutrition via a tube increases oral bacteria and dental plaque (PMID 40499817). This is back to the story of Axis ②. The decision to provide nutrition through a tube itself is covered in "When Eating Is Not Possible — Facing the Tube-Feeding Decision".

Trying to Block 3 — Pressing Down Only on Bacteria

If the bacteria inside what went down are the problem, it seems like giving antibiotics in advance would work. This idea was tested in large trials, and pneumonia was not clearly reduced.

• STROKE-INF — A cluster-randomized trial that randomly divided 48 stroke wards in the UK by ward, with 1,217 people analyzed. Preventive antibiotics were given for 7 days to patients with Dysphagia after acute stroke. Pneumonia defined by preset criteria (an algorithm) was 13% in the antibiotic group and 10% in the control group (adjusted OR 1.21, 0.71–2.08), and pneumonia diagnosed by a doctor was 16% versus 15%, showing no difference. What decreased was not pneumonia, but infections, mainly urinary tract infections (OR 0.55). The authors concluded that preventive antibiotics cannot be recommended (PMID 26343840).
• PRECIOUS — A 2×2×2 factorial trial that randomly assigned 1,471 patients with moderate-to-severe acute stroke aged 66 and older to metoclopramide, ceftriaxone, and paracetamol separately. At 90 days, the adjusted common odds ratio for disability (mRS) with ceftriaxone was 0.99 (0.77–1.27), showing no signal of benefit (PMID 38074444).

A meta-analysis is an analysis that pools the results of multiple trials and recalculates them, and meta-analyses of antibiotics diverge on pneumonia. In an analysis pooling 7 trials with 4,261 people, overall infections decreased (OR 0.57), but pneumonia did not significantly decrease (OR 0.91, 0.73–1.13) (PMID 27841284). In an analysis pooling 6 trials with 4,125 people, early pneumonia decreased, with a risk ratio (RR) comparing the proportions of people experiencing the event of 0.64 (0.42–0.96) (PMID 29049353). Where the two analyses agree is on the outcome side. Neither mortality nor functional outcomes changed. ESO guidelines do not recommend prophylactic antibiotics because there is no benefit for pneumonia incidence and clinical outcomes. The same guidelines recommend starting empirical antibiotic treatment promptly after pneumonia is diagnosed (PMID 42095755).

Pressing down only on bacteria did not consistently reduce pneumonia, and neither mortality nor function changed. This matches the prediction of the framework that touching only one axis is not enough.

Attempting to Block 4 — Cleaning the Mouth and Raising the Body

This is the most promising yet most conflicting branch of evidence.

On the positive side, a randomized trial from a Chinese hospital divided 84 post-stroke patients with Dysphagia into groups and used an assessor-blinded design where the people measuring the results did not know which group patients were in. The group that received a bundled intervention of mouth cleaning, mouth exercises, massage, and safe swallowing education had a stroke-associated pneumonia rate of 10.26% on day 7, compared to 32.43% for the control group that received routine oral care (P<0.05) (PMID 41345604).

On the negative side, a Japanese multicenter prospective cohort compared hospitals implementing a standardized oral function management protocol with medical-dental collaboration against hospitals providing routine oral care. When 313 pairs were matched by propensity score among 1,616 acute stroke patients, there was no significant difference in the incidence of aspiration pneumonia (p=0.7639). What did improve were oral hygiene, tongue movement, scores rating the level of oral intake (FOIS), and length of stay (PMID 41482746). In a study linking survey data on swallowing care practices from 113 hospitals in England and Wales to the national stroke registry (SSNAP), neither the presence of a written oral care protocol, nasogastric tube insertion practices, nor the type of swallowing screening protocol was associated with the incidence of stroke-associated pneumonia. The authors note that larger sample studies are needed (PMID 34915473).

What the two negative results compared were structured protocols versus routine care, and hospitals with written protocols versus those without. The control hospitals in the Japanese study also provided routine oral care (PMID 41482746). Therefore, these results are an answer to whether "more structured care reduced pneumonia further," and cannot be read as a comparison of mouth care itself.

What about early mobilization? In an observational retrospective study of 407 severe stroke patients, the group that got out of bed and moved had less pneumonia than the group that stayed in bed (45.5% versus 62.5%, OR 0.50), fewer pressure ulcers, and no increase in falls (PMID 32912565). ⚠️ This is observational data again. Patients who could be sat up were likely less severe to begin with.

Checking and managing swallowing appeared alongside a reduction in pneumonia. A systematic literature review is a research synthesis that gathers and evaluates studies comprehensively using a set procedure. In a systematic review gathering 12 studies involving 87,824 people, structured swallowing screening, early swallowing screening, and specialist evaluation by a speech-language pathologist were associated with a reduced risk of stroke-associated pneumonia. However, because differences between studies were large, the results could not be pooled and calculated together (PMID 30199856).

Certainty of evidence is a grade rating how much we can trust the pooled results, categorized as high, moderate, low, or very low. In a systematic review gathering 22 randomized and quasi-randomized trials involving 1,700 people in acute and critical care settings, the risk of pneumonia was lower in the group that received swallowing therapy (RR 0.71, 0.56–0.89, 8 studies, low certainty). Yet in the same analysis, aspiration itself was not significantly reduced (RR 0.79, 0.44–1.45, 4 studies). The authors conclude that the evidence supporting the efficacy of acute-phase swallowing therapy is still limited (PMID 32514597).

Pneumonia decreased, but it remains unconfirmed whether that was because aspiration decreased. Because only four trials measured aspiration, we cannot conclude definitively that aspiration did not decrease.

Why There Is Still No Answer — Severity Takes Up the Largest Share of the Pneumonia Model

The Shapley value is a calculation that divides a result produced by multiple factors into individual shares for each factor, and it originated from cooperative game theory. In 617 stroke patients at Monash Medical Centre in Australia, pneumonia occurred in 6.6%, and its contribution to the pneumonia prediction model was stroke severity 72.8%, comorbidity index 16.2%, age 7.2%, and swallowing screening 3.8% (PMID 30761063).

⚠️ We must also read the caveats the authors added themselves. Because swallowing screening was actually performed in this cohort, the possibility that the effect of Dysphagia on pneumonia was diminished cannot be ruled out (PMID 30761063). Therefore, this result is closer to "the remaining share of a factor that is already managed looks small" than to "Dysphagia is not important." If you measure the effect of rain on people who are already carrying umbrellas, the rain looks small.

Is Pneumonia a Waystation on the Road to Death — Mediation and Interaction

Mediation is when a cause produces an outcome through an intermediate step, and interaction is when the risk when two factors are present together is greater than the sum of their individual shares. If a severe stroke causes pneumonia and that pneumonia causes death, it is mediation, and if the two meet and amplify each other's risk, it is interaction.

Data on 339,139 people from the Sentinel Stroke National Audit Programme (SSNAP) in England and Wales directly address this distinction. Stroke-associated pneumonia occurred in 9.4%, and in-hospital death was 15.6% (45% of patients with pneumonia, 12% of patients without pneumonia). The total risk ratio by which a severe stroke raised the risk of in-hospital death compared to a mild or moderate stroke was 4.72. In a four-way decomposition dividing this total effect into four shares, 43% of it came from the additive interaction between severity and pneumonia, and in very severe strokes, it was 50%. There was no evidence of a mediating effect passing through pneumonia (PMID 37170807).

Pneumonia acted less like a waystation on the road to death and more like a factor that amplifies each other when present together with a severe stroke. This means the causal structure is not a single chain. The operational conclusion of the authors points toward prevention. Preventing pneumonia in patients with severe stroke can reduce deaths by up to 43% (PMID 37170807). The structure called interaction does not diminish the practical benefit of prevention.

When the Measurer Shakes, and Guidelines Do Not Hide That Fact

What we saw earlier meets here. Diagnostic criteria are not agreed upon (PMID 39612934), the prevalence moves from 15.57% to 25.32% depending on how you count (PMID 40993813), and pneumonitis and pneumonia get mixed up (PMID 28879323). Contending over causes while the yardstick for counting outcomes shakes is where we stand now.

The European Stroke Organisation (ESO) 2026 stroke-associated pneumonia guidelines set up 15 clinical questions, of which only 3 were supported mostly by low-quality evidence, and the guidelines themselves state that most guideline sentences lean on expert consensus. The same guidelines view dysphagia screening and swallowing management as parts already established in care preventing post-stroke pneumonia, and list posture, early mobilization, and personalized nutrition as prevention strategies (PMID 42095755).

It is also an old problem. A systematic review for the 2001 US AHRQ evidence report noted that introducing dysphagia diagnosis and treatment programs appeared alongside a significant drop in pneumonia rates. At the same time, it stated that available evidence is limited, and comparisons lean on historical controls using past patients as comparisons (PMID 11720404). Twenty-five years have passed, and the verdict is still deferred.

Common Misconceptions

「Once aspiration is confirmed, pneumonia is just a matter of time.」 Roughly half of healthy people have many oral bacteria deep in their lungs (PMID 27572644), 10% of age-matched controls also aspirated during sleep (PMID 8025758), and in 73 people with confirmed penetration or aspiration, the PAS score did not predict pneumonia (PMID 40088308). However, 56% of the same cohort developed pneumonia (PMID 40088308), and aspiration of saliva predicted pneumonia (PMID 22329364). Aspiration does not equal pneumonia, but a person with confirmed aspiration is at high risk.

「If you do not feed them, they are safe.」 In data from 160 people with Dysphagia, the risk of pneumonia was actually 2.5 times higher in fasting patients, and most were diagnosed before eating by mouth (PMID 41044338). ⚠️ There is major confounding in that sicker patients are the ones who receive fasting.

「If you insert a tube, they are safe.」 Nasogastric tube insertion (OR 3.91) and fasting (OR 5.62) were associated with respiratory infection in the first week (PMID 25402187). ⚠️ Here too, tubes go to sicker patients.

「You just need to give antibiotics in advance.」 In STROKE-INF, pneumonia was not reduced (PMID 26343840), in PRECIOUS the 90-day functional outcome odds ratio for ceftriaxone was 0.99 (PMID 38074444), and ESO guidelines do not recommend prophylactic antibiotics, but do recommend prompt initiation of antibiotic treatment once pneumonia is diagnosed (PMID 42095755).

「Aspiration pneumonia is a disease of Dysphagia.」 In a study following 189 older adults for up to 4 years, the strongest predictor was feeding dependence (OR 19.98) (PMID 9513300), and in stroke, Dysphagia and immunosuppression each independently predict pneumonia (PMID 27733675).

「Therefore, swallowing assessments and oral care are useless.」 The data point the other way. In a review pooling 12 studies and 87,824 people, early swallowing screening and speech-language pathologist evaluation were associated with a reduced risk of pneumonia (PMID 30199856), ESO guidelines view swallowing screening and swallowing management as an established part of pneumonia prevention care (PMID 42095755), and measures such as posture, speech therapy, rehabilitation, and plaque removal have also been reported to be effective (PMID 40499817). The argument of this article is not that we should omit evaluations and care, but rather that pneumonia is not determined by stopping aspiration alone.

Honest Unresolved Questions

• The yardstick for counting outcomes has not been finalized. There are no agreed diagnostic criteria or unique markers (PMID 39612934), and depending on how you count, the prevalence moves from 15.57% to 25.32% (PMID 40993813).
• The direction of the effect for fasting and eating restrictions remains open. Because fasting paradox data are observational (PMID 41044338), there is currently no basis to confirm that "restrictions are harmful" nor to say that "restrictions prevent pneumonia."
• The effect sizes of oral care programs diverge. Small randomized trials reported large differences (PMID 41345604), whereas a large propensity score-matched cohort and a nationwide practice survey found no difference (PMID 41482746, PMID 34915473).
• There is still no evidence from humans that reversing immunosuppression reduces infections. Registry data examining patients who were on beta-blockers were also negative (PMID 29694433), and whether this immunosuppression is an adaptation or maladaptation remains open (PMID 42212161).
• Guidelines lean on consensus rather than evidence. Only 3 out of 15 ESO guideline questions were supported, and even those relied mostly on low-quality evidence (PMID 42095755).

To Go Deeper

• When reading papers — First distinguish whether the measured result is aspiration (PAS, etc.) or pneumonia. The meta-analysis of acute-phase swallowing therapy that reduced pneumonia but could not confirm whether it reduced aspiration is an example of this (PMID 32514597).
• When reading risk factor papers — Separate whether the factor was given to the patient (severity, infarct volume) or done by the medical staff (fasting, tube, antibiotics). If something done by the medical staff is caught as a risk factor, you should almost always suspect indication confounding first.
• When reading prevention studies — Look at which axis was touched. Most trials touched only one among what enters (positioning·posture), the content of what entered (oral hygiene·antibiotics), or the blocking side (immunity·movement·nutrition), and most of them reported small or uncertain effects.
• If you leave only one — A necessary condition is a condition without which an event does not occur, and a sufficient condition is a condition by which an event occurs on its own. Aspiration is close to a necessary condition for aspiration pneumonia, but not a sufficient condition. A disease name points to a single cause, but the data points to multiple ones.

Source Papers

Design and cohorts are noted together. Figures already in the main text are not repeated. ⚠️ marks data where indication confounding is noted in the text.

Weight and Definition — PMID 37270854 (2023) Beijing Medical Insurance approx. 12 million cohort · PMID 42095755 (2026) ESO stroke-related pneumonia guideline · PMID 39612934 (2024) Epidemiology and pathogenesis review, lack of diagnostic criteria · PMID 40993813 (2025) Electronic health records 15,603 patients re-read · PMID 28879323 (2017) Nomenclature issues between pneumonitis and pneumonia · PMID 34087609 (2021) Redefinition proposed, requiring aspiration risk factors and oral colonization risk factors simultaneously.

Axis ① Getting In Is Not Enough — PMID 8025758 (1994) Indium-111 sleep aspiration, including age-matched controls · PMID 27572644 (2016) Multi-omics of bronchoalveolar lavage fluid in 49 people without lung disease · PMID 40088308 (2025) FEES penetration/aspiration confirmed in 73 people, mean 3.87-year follow-up · PMID 22329364 (2012) Nursing home 148 people videofluoroscopy and 3-month follow-up.

Axis ② Inside the Mouth — PMID 26416412 (2015) Dysphagia elderly 47 people vs. controls 14 people oral colonization · PMID 29053769 (2018) Nursing home 173 people tongue microbiome 16S, ⚠️confidence interval extremely wide · PMID 9513300 (1998) Elderly 189 people up to 4 years prospective, the conclusion that became the pivot of this piece.

Axis ③ The Blocking Side — PMID 27733675 (2017) PREDICT, acute ischemic stroke 486 people prospective · PMID 19661470 (2009) Stroke 50 people vs. controls 40 people serial immune cell measurement · PMID 42212161 (2026) Stroke-induced immunosuppression mechanism review, adaptive/ maladaptive unresolved · PMID 16946159 (2006) Mouse model · PMID 24922549 (2014) Rat model where sham-surgery animals showed similar changes, advising cautious interpretation · PMID 29694433 (2018) Human beta-blocker registry data 306 people, null.

Verification — Fasting and Consistency — PMID 41044338 (2025) Acute stroke dysphagia 160 people retrospective, fasting paradox and temporal sequence ⚠️ · PMID 36775772 (2023) Post-head-and-neck-cancer surgery 654 cases ⚠️large severity imbalance · PMID 40499817 (2025) Recurrence prevention review.

Verification — Tubes — PMID 25402187 (2014) Acute stroke 536 people retrospective multivariate ⚠️ · PMID 29619559 (2018) Comparison of penetration-aspiration scores with and without nasogastric tube in 147 people.

Verification — Antibiotics — PMID 26343840 (2015) STROKE-INF, 48 units 1,217 people cluster randomized · PMID 38074444 (2024) PRECIOUS, 1,471 people 2×2×2 factorial · PMID 27841284 (2016) Meta-analysis 7 studies 4,261 people · PMID 29049353 (2017) Meta-analysis 6 studies 4,125 people (pneumonia outcome diverges from the previous analysis).

Verification — Mouth, Mobility, and Swallowing — PMID 41345604 (2025) Stroke dysphagia 84 people assessor-blinded randomized · PMID 41482746 (2026) Japanese multicenter propensity score 313 pairs matched, null · PMID 34915473 (2022) England and Wales 113 hospitals practice linked to SSNAP, null · PMID 32912565 (2020) Severe stroke 407 people early mobilization retrospective ⚠️ · PMID 30199856 (2018) Screening and assessment systematic review 12 studies 87,824 people · PMID 32514597 (2020) Acute-phase swallowing therapy randomized trials 22 studies, pneumonia decreased (low certainty) and a reduction in aspiration was not confirmed.

Anatomy of Confounding — PMID 30761063 (2019) Stroke 617 people Shapley value decomposition, ⚠️including caveats added by the authors themselves · PMID 37170807 (2023) SSNAP 339,139 people 4-way decomposition · PMID 11720404 (2001) Systematic analysis for AHRQ evidence report.