When Eating Is Not Possible — Facing the Tube-Feeding Decision

For someone who can no longer swallow, should a tube be inserted? This decision must be made in a hurry, and it can feel irreversible once a tube is put in. Looking at the evidence gathered in this article, the situation varies depending on the type of tube and the timing. Nasogastric tubes are easy to remove, stroke guidelines have recommended a sequence of trying a nasogastric tube first for a set period, and among people who have had a percutaneous endoscopic gastrostomy tube inserted, some have had the tube removed and returned to eating by mouth. Furthermore, the data thought to support this decision do not actually answer that question precisely.

Tube feeding is the process of delivering nutrition into the stomach or intestines through a tube instead of eating by mouth. Dysphagia is a condition where swallowing does not work properly or goes awry. The primer "Swallowing, Speech and Language — The Crossing Where the Food Path Meets the Airway" listed preventing complications such as pneumonia and malnutrition as the first branch of dysphagia treatment. If necessary, tube feeding was also included as a single line within that branch. This article looks at the place where that single line becomes an actual decision.

This article does not tell you what to do. It does not divide the side of inserting a tube and the side of not inserting a tube into right and wrong. The article does one thing: it establishes what the evidence mobilized for that decision actually said and what it did not say.

The article goes through the largest randomized trial, the two types of tubes, what tubes fail to prevent, why tubes and poor outcomes appear together, returning swallowing, and progressive dementia in turn, ending with a few questions that shape the decision.

Almost all the cited research is foreign data. It was conducted in Europe including the UK, the US, Canada, and Switzerland, as well as Hong Kong and Japan, and each country's healthcare system, long-term care environment, legal framework, and practices are mixed into the numbers. A commentary written about patients with progressive dementia in US nursing homes points out that the use of tube nutrition is influenced by financial incentives, regulations, and the litigation environment (PMID 17498602). Therefore, the numbers below are values observed under specific systems, and this article makes no claims about South Korea's laws, systems, or practices.

Why This Decision Arises So Often

A nasogastric tube (NG tube) is a tube inserted through the nose all the way to the stomach. In data where 570 patients admitted with acute ischemic stroke were consecutively and completely enrolled, dysphagia was present in 118 patients (20.7%). Among them, half (50.9%) still had dysphagia at discharge, and 36 patients (30.5%) had severe enough dysphagia to require a nasogastric tube.

Patients with dysphagia had pneumonia more frequently than those without it (23.1% versus 1.1%). Three months later, the proportion living at home was lower (38.8% versus 76.5%), and deaths were higher (13.6% versus 1.6%). Even in a multivariable analysis considering multiple factors together, severe dysphagia requiring tube placement was strongly associated with death (PMID 26863627).

This decision is not rare. And people at the moment of facing the decision are already tipped toward poor outcomes. We must keep in mind this tilt—that the person receiving the tube is sicker to begin with—so as not to misread the numbers that follow.

First, the Terms — Two Tubes and One Scale

A nasogastric tube can be inserted right at the bedside and is easy to remove. Instead, it frequently dislodges and clogs.

Percutaneous endoscopic gastrostomy (PEG) is a procedure that uses an endoscope to create a direct pathway from outside the abdomen into the stomach and places a tube through that pathway. A tube placed this way is commonly called a gastrostomy tube, and in this article, PEG refers to both this procedure and its tube. It is more stable than a nasogastric tube, but it is a procedure, and a procedure carries procedural risks.

The Functional Oral Intake Scale (FOIS) is a score that classifies the level of eating by mouth from 1 to 7. Oral intake means eating by mouth. Levels 1–3 are states of receiving nutrition through a tube, levels 4–5 are states of eating by mouth with modified food consistency and thickness, and levels 6–7 are states of eating by mouth without restrictions (PMID 31866129). Several numbers in this article are read on this scale.

Regarding the timing of tube insertion, two paths clash. "Early," which inserts the tube now, and "avoidance," which waits at least a week without inserting it. These are the two paths that the FOOD trial actually randomized between.

The Greatest Trial — FOOD

A randomized trial is a study that compares two methods by dividing patients at random, like a draw of lots. Because neither the doctor nor the family chooses who goes to which side, both groups are filled with similar people from the start. This property is what most significantly distinguishes it from observational data that come later.

At the center of the evidence for tube feeding lies the FOOD (Feed Or Ordinary Diet) trial. FOOD is a collection of three pragmatic randomized trials that enrolled 5,033 patients admitted for stroke. Between November 1996 and July 2003, 131 hospitals in 18 countries participated (PMID 16409880).

FOOD measured death and disability at 6 months. A poor outcome is a state where the patient is alive but left with enough disability to need another person's help in daily life. Thus, "death or a poor outcome" is the combined rate of people who died or, even if they survived, were left with a high degree of disability.

Trial 1 looked at oral nutritional supplementation. It divided 4,023 patients who could swallow into one group receiving only standard hospital food and another group receiving standard hospital food plus oral nutritional supplements (1.5 kcal/mL 360 mL, protein 20 g/day). Only 7.8% (314 patients) were judged to be malnourished at enrollment. The risk of death was lower by 0.7% in absolute terms in the supplement group (95% CI −1.4 to 2.7; p=0.5), and death or a poor outcome increased by 0.7% (−2.3 to 3.8; p=0.6).

A 95% confidence interval (CI) is the range where the true value is likely to lie. If this range spans across zero on both sides, one cannot conclude that there is a difference, nor that there is none. Both results of Trial 1 are like this. The authors concluded that this result does not support uniformly giving nutritional supplements to stroke patients who generally have good nutritional status upon admission (PMID 16409880).

Trial 2 sits right in the middle of this decision. It divided 859 dysphagia patients enrolled within 7 days of admission into early tube feeding and at least 1 week of avoidance. At 6 months, 182 out of 429 patients (42.4%) in the early group and 207 out of 430 patients (48.1%) in the avoidance group had died. The risk of death was lower by 5.8% in absolute terms in the early group, but the confidence interval spans from −0.8 to 12.5, crossing zero (p=0.09). This is a position where one can neither conclude that it decreased nor that there is no difference.

However, the primary outcome (the primary endpoint) that this trial had predefined was death or a poor outcome. Here, the reduction was limited to 1.2% (−4.2 to 6.6; p=0.7). People who were alive but remained with a poor outcome were 36.6% in the early group and 31.9% in the avoidance group (PMID 15733717, PMID 16409880).

The authors' sentence is the core of this trial. "Early tube feeding may reduce the case fatality rate, but it may come at the cost of increasing the proportion of survivors with a poor outcome." The case fatality rate is the proportion of people who die from a disease. The fact that death and disability—the two outcomes—did not move in the same direction is the most honest legacy left by this trial.

Trial 3 compared PEG and nasogastric tubes. It divided 321 dysphagia patients within 30 days of admission between the two tubes. 79 out of 162 patients (48.8%) in the PEG group and 76 out of 159 patients (47.8%) in the nasogastric tube group died, so the difference in mortality was not large at 1.0% (−10.0 to 11.9; p=0.9), but the confidence interval widened to more than 10 percentage points on both sides. However, the proportion reaching a good outcome was 11.1% for PEG and 18.9% for nasogastric tubes, and death or a poor outcome was 7.8% (0.0 to 15.5; p=0.05) higher in the PEG group (PMID 15733717).

The conclusion of FOOD was that during the first 2 to 3 weeks of the acute phase, the nasogastric tube side appears to yield better functional outcomes, and this data does not support an early PEG policy (PMID 16409880). In the same report, the authors also noted that patients whose dysphagia is prolonged or who cannot tolerate a nasogastric tube inevitably end up being fed via PEG (PMID 16409880). What this conclusion targeted is the policy of hastily inserting a PEG in the first 2 to 3 weeks.

There are also things FOOD did not ask. The subjects were acute stroke, the observation period was 6 months, and the assignment in Trial 3 was within 30 days of admission. Progressive dementia, cancer, neurodegenerative diseases, and long-term care are not questions this trial answered. And although the 5.8% reduction in death in Trial 2 has a confidence interval crossing zero, the authors themselves noted that it is very unlikely a policy of avoiding early tube feeding significantly improves survival (PMID 16409880).

Placing the observational data of the same FOOD patients side by side makes the texture of the evidence sharper. Among 3,012 registered patients, 37% of the 275 who were malnourished died within 6 months, while 20% of the 2,194 who had normal nutritional status died. The odds ratio (OR) is a value comparing the likelihood of something happening in two groups; if it is greater than 1, it means it happened more frequently in that group. The mortality odds ratio for the malnourished side was 2.32 (1.78–3.02), and although it weakened after adjusting for age, pre-stroke function, and severity, the association remained (OR 1.82; 1.34–2.47) (PMID 12750536). That nutritional status is associated with an outcome and that supplementing nutrition changes the outcome are different propositions. Trial 1 illustrates precisely that gap.

PEG or Nasogastric Tube — Why the Data Diverge

Looking at the FOOD trial 3 alone makes the PEG seem disadvantageous. Yet other data point in the opposite direction.

• A Cochrane review gathered 11 randomized trials involving 735 participants regardless of the underlying disease. Intervention failure, meaning tubes getting blocked, falling out, or failing to be maintained, was markedly lower with PEG (RR 0.18; 0.05–0.59, 8 trials, 408 participants, low certainty of evidence). An RR (relative risk) of 0.18 means that intervention failure on the PEG side was 0.18 times that of the nasogastric tube side. Aspiration pneumonia is pneumonia that develops after food or saliva enters the airway. There was no significant difference in death (RR 0.86; 0.58–1.28) and aspiration pneumonia (RR 0.70; 0.46–1.06). Serum albumin, a blood test value reflecting nutritional status, was higher in the PEG group (MD 6.03; 2.31–9.74), and fewer people in the PEG group complained of discomfort, changes in body image, and restrictions on social activities (PMID 25997528).
• A meta-analysis is an analysis that pools the results of multiple studies and calculates them together. A 2026 meta-analysis (10 studies, 1,416 participants) focusing only on post-stroke Dysphagia reported that death was 30% lower (RR 0.70; 0.50–0.98) and pneumonia was 47% lower (RR 0.53; 0.42–0.67, I²=0%) with PEG. I²=0% means that the pneumonia results barely varied across the studies. However, the survival benefit disappeared in a sensitivity analysis recalculated leaving only randomized trials, and the authors themselves noted that this could be due to the two groups having been different to begin with (confounding) (PMID 41496648).
• Data from Japanese recovery-phase rehabilitation wards, meaning wards where patients receive rehabilitation after passing the acute phase (47 people with PEG versus 49 people with nasogastric tubes), showed that aspiration pneumonia was 4.67 times (1.30–16.67; P=0.02) more frequent in the nasogastric tube group. This is a value adjusted through propensity score matching, which ensures that people with similar likelihoods of receiving a PEG are compared. There was no difference in the improvement of activities of daily living (PMID 34514181).

Rather than being contradictory, these data are answering different questions. FOOD trial 3 is the result of random allocation in the first 2 to 3 weeks of the acute phase of stroke. The Cochrane review gathered trials regardless of the underlying disease and timing, and the remaining two mixed observational studies or are observational data. In observational data, the people who receive a PEG and those who maintain a nasogastric tube are fundamentally different people. Neither "PEG is better" nor "nasogastric tube is better" holds true without accounting for timing and study design.

Even With a Tube, Aspiration Pneumonia Does Not Disappear

The expectation that getting a tube makes aspiration pneumonia disappear is the point most frequently misunderstood. Aspiration is when things like food, water, or saliva enter the airway below the vocal cords, and how much aspiration causes pneumonia is examined in "Is Aspiration Pneumonia Really Caused by Aspiration?".

In a prospective study tracking 100 acute stroke patients nourished via nasogastric tubes, 44% developed pneumonia, and most were diagnosed on days 2 to 3 after onset. Patients with impaired consciousness or severe facial paralysis were at especially high risk. The authors' concluding sentence was "nasogastric tubes offer limited protection against aspiration pneumonia in acute stroke dysphagia patients" (PMID 15145999).

The same goes for PEGs. In a meta-analysis of 22 studies covering 996,567 stroke patients, looking exclusively at patients who received a PEG, in-hospital mortality was 19.8%, aspiration pneumonia was 13.6%, and total pneumonia was 58% (PMID 39190711). In the same analysis, the risk of aspiration pneumonia in the PEG group came out to 11.72 times (3.75–36.62) that of the control group. However, this number cannot be read as the effect of PEG. The control group in this comparison was a separate cohort of patients, making it close to essentially comparing "people sick enough to need a tube" with "people likely not that sick."

In a meta-analysis (30 studies) gathering safety incidents related to dysphagia in older adults by feeding method, the incident rate was 13.8% for eating by mouth, 23.9% for nasogastric tubes, and 26.5% for PEGs. Looking at aspiration pneumonia alone, it was 12.0% for mouth, 20.6% for nasogastric tube, and 12.4% for PEG (PMID 39985956). Each method also comes with its own unique incidents. Choking was reported when eating by mouth, gastroesophageal reflux with nasogastric tubes, and tube blockage, dislodgement, leakage, vomiting, nausea, insertion site pain, gastrointestinal bleeding, and peritonitis with PEGs.

The fact that the aforementioned Cochrane review found no difference in aspiration pneumonia between PEGs and nasogastric tubes (RR 0.70; 0.46–1.06) also falls into the same picture (PMID 25997528). Even if you change the type of tube, aspiration pneumonia remains. And the same fact underlies observational data comparisons. People who receive tubes are sicker patients.

Whether the tube itself worsens swallowing has also been checked separately. When 100 stroke patients were examined endoscopically, 5 people had nasogastric tubes incorrectly placed in the pharynx at the back of the throat, and all 5 of those people had worsened swallowing. On the other hand, when repeatedly comparing 25 people with and without a properly placed nasogastric tube, there was no change in the amount entering toward the airway. However, in two cases, what was swallowed adhered to the tube and entered inside the laryngeal entrance (laryngeal vestibule). Based on this, the authors concluded that there are no in-principle obstacles to starting oral feeding in a limited and monitored manner even with a nasogastric tube in place (PMID 18651972).

Tube Feeding at Discharge and 90 Days — and Why It Is Hard to Read as Causation

In a prospective study tracking 201 patients in a stroke intensive care unit, disability was defined as an mRS of 3 or higher, which is a state where a person needs at least some help in daily life. The mRS is a score that measures how much a person can manage on their own in daily life after a stroke. For patients receiving nutrition through a tube at discharge, the odds ratio for having lingering disability at 90 days was 14.97 (2.68–83.65; p=0.002), and the odds ratio for death was 9.79 (2.21–43.4; p=0.003). Conversely, having an FOIS score of 6–7 at discharge—meaning eating by mouth without restrictions—was linked to a lower chance of disability (OR 0.17). The authors' conclusion was that "dysphagia or tube feeding at discharge is a marker of poor prognosis," and they added that it is important to evaluate swallowing early and keep a close watch on patients with tube feeding (PMID 31866129).

In this conclusion, the word "marker" is decisive. A marker is a signal that lets you know in advance that a bad outcome is coming, and it does not carry the meaning that it caused that outcome. The confidence interval is also wide, stretching from 2.68 to 83.65. Because the value comes from 201 people, the true value could be anywhere within this wide range.

Stroke registry data from Ontario, Canada, show this problem more directly. One group consisted of patients who had a tube placed directly into the stomach or small intestine through the belly after an acute stroke (gastrostomy or jejunostomy), which is called a direct enteral tube (DET). The other group consisted of patients who only used a nasogastric tube. Among 1,448 patients who received a DET and survived until discharge, matched patients were compared with those who used only a nasogastric tube by matching them using propensity scores.

The DET group had higher rates across the board: severe disability at discharge at 89.6% versus 78.4%, discharge to a long-term care facility at 38.0% versus 16.1%, aspiration pneumonia at 14.4% versus 5.1%, and 2-year mortality at 41.1% versus 35.9%. However, 30-day mortality after discharge was actually lower in the DET group (9.7% versus 15.3%; HR 0.61; 0.49–0.75). The hazard ratio (HR) is a value comparing the speed at which an event occurs between two groups, and a value less than 1 means the event happened less often in that group. And once they matched the groups while also adjusting for the length of hospital stay and the degree of disability at discharge, this difference vanished (HR 0.90; 0.70–1.17). The authors wrote that these results can be used to discuss goals of care regarding long-term tube feeding (PMID 29367443). The fact that an advantage disappears entirely with a single adjustment—this is the nature of observational data.

Data from 143,331 older adults with dementia in Ontario tell the same story. The proportion of people who received a tube during hospitalization was 0.9% (1,312 people). These individuals had a longer hospital stay (an average of 65.6 days versus 14.8 days), were admitted to the intensive care unit more often (42.5% versus 10.2%), and had higher in-hospital mortality (22.4% versus 10.3%) and mortality within one year of discharge (50.0% versus 28.4%). Among people receiving home care or living in long-term care facilities, the likelihood of receiving a tube was higher when swallowing problems (OR 2.22) and functional disability (OR 2.75) were present, and lower when there was a do-not-resuscitate order (OR 0.38). The authors concluded that tube insertion during hospitalization was not linked to improved survival or post-discharge outcomes (PMID 39976967).

There is a name that must be attached when reading data like this. Confounding is an optical illusion where two groups being compared are different from the beginning, making that difference look as if it were the effect of a treatment (the primer, "If People Who Take It Live Longer, How Were They Different to Begin With? — Confounding"). Confounding by indication is confounding that occurs because the very reason a treatment came to be given is linked to a bad outcome. Tubes go to sicker people. People with poor consciousness, large strokes, and dim prospects for recovery are the ones who receive tubes. Therefore, the observation that 「the outcome was poor for people who received a tube」 may mean not that the tube caused the poor outcome, but that the tube was used in a situation that foreshadowed a poor outcome. The real-world example of this is how the 30-day mortality advantage vanished purely through adjustment in the Ontario analysis above.

The Axis of Time — How Much Returns, and When

Ultimately, what makes the decision about tube feeding difficult is time. A significant number of people who cannot swallow right now will swallow later. The problem is who and when.

Looking at those who survive, the vast majority eat by mouth again. A cohort is a study that designates a group of patients and follows them over time. In a cohort of 427 acute stroke survivors, 55% had dysphagia at admission, 37% at discharge, and 5% needed tube nutrition at 6 months. Whether people who died in the meantime entered this denominator is not stated in the abstract. Among those with initial dysphagia, 32% resolved their dysphagia within 2 weeks, 44% started eating by mouth before discharge with dysphagia remaining, and 23% required alternative nutrition pathways like tubes throughout hospitalization. And 90% of those eating by mouth at discharge were still eating by mouth 6 months later. The authors concluded that while more than half of survivors experienced dysphagia, the majority were able to return to eating by mouth by 6 months (PMID 34280690).

When you leave only the severe cases, the picture changes. Among 723 stroke patients who still could not eat by mouth on day 10 of onset, out of 586 who answered a questionnaire 3 months later, 141 (24.2%) had become able to eat by mouth (PMID 22947895). Among 48 ischemic stroke patients who remained at FOIS 1–4 even after 3 weeks, 50% died within 3 months. Of the survivors, 62.5% ate by mouth again, but 75% answered that swallowing symptoms remained (PMID 41232430).

Looking further out, some people do remove the tube. Among 77 stroke patients who received a PEG tube due to severe dysphagia and were admitted to a rehabilitation hospital, 66 were followed for an average of 2 years; 64% were alive and 45% had removed the tube and returned to eating meals by mouth (PMID 16136260).

There are also tools to gauge recovery in advance. PRESS (Predictive Swallowing Score) is a score created and validated in Switzerland with ischemic stroke patients whose swallowing was initially severely poor. It predicts recovery using age, stroke severity, lesion location, initial aspiration risk, and initial impaired oral intake. Among participants combining the development and validation cohorts, 64% (59–71) had not regained the level of eating properly by mouth within 7 days, and 30% (24–37) had not regained it within 30 days. Discriminative ability (C) is a value indicating how well a score separates those who will recover from those who will not, with larger values meaning better separation. The discriminative ability of the validation cohort was 7. days C=0.84 and 30 days C=0.77 (PMID 30742198).

However, in an external validation with 149 people from 2 US institutions, the scores predicted by the model did not match reality (goodness-of-fit test p<0.00001), and discriminative ability also dropped to C=0.75. In particular, it overpredicted that dysphagia would persist in severe strokes (PMID 41687691). Predictive tools exist, but they do not work the same way everywhere.

A time-limited trial is an approach where treatment is tried for a set period, and then the decision whether to continue is re-evaluated based on the results. Stroke guidelines have recommended considering a nasogastric tube if dysphagia persists for 7 days or more, and a PEG if it does not recover within 30 days (PMID 30742198). This is a sequence of trying a nasogastric tube time-limited trial before a PEG (PMID 27965430). However, among 34,623 ischemic stroke hospitalizations that received a PEG between 2001 and 2011 in the US Nationwide Inpatient Sample, 53% received a PEG between days 1 and 7 of hospitalization. The median time to PEG varied widely from 3 days to over 3 weeks depending on the hospital. Furthermore, those aged 85 and older, who could benefit the most from a nasogastric tube time-limited trial, were paradoxically more likely to receive early PEG (aOR 1.68; 1.50–1.87) (PMID 27965430).

Here too, who is sent down which pathway gets mixed into the result numbers. In single-institution data where the Department of Rehabilitation Medicine judged the rehabilitation transfer of 426 stroke dysphagia patients, 86.3% (38/44) of patients transferred to rehabilitation with a nasogastric tube reached a diet of eating exclusively by mouth, whereas 38.6% (29/75) of patients transferred with a PEG did (PMID 38112750). Because the Department of Rehabilitation Medicine decided who would be sent with a nasogastric tube based on their recovery outlook, this difference reflects not only the type of tube, but also the initial choice of who was sent with a nasogastric tube to begin with.

The Question Is Different in Progressive Dementia

Stroke is an event with room for recovery. Progressive dementia is a stage where dementia has deepened to the point where help is needed for almost every daily task, and having difficulty eating at this point is part of the disease's progression itself. The texture of the evidence is also different.

A Cochrane Review did not find a single randomized trial on this question. Seven observational controlled studies were all that existed, and among the six that looked at mortality, there was no evidence that tube feeding increased survival. Not a single study looked at quality of life, and there was no evidence of benefit in nutritional status or pressure ulcers either. The authors also noted that data on harmful effects were lacking (PMID 19370678).

A position statement by the American Geriatrics Society summarizes it as follows. When older adults with progressive dementia develop difficulty eating, tubes are not recommended. Careful hand feeding is a method where someone stays by the side, watches the swallowing, and feeds by mouth, and the statement notes that this has been confirmed to be as good as tube feeding in terms of mortality, aspiration pneumonia, functional status, and comfort. Furthermore, the statement notes that tube feeding is associated with agitation, an increase in physical and chemical restraints (tying the body or using medications to sedate), healthcare utilization due to tube-related complications, and the occurrence of new pressure ulcers. The same statement defines tube feeding as a treatment that a surrogate decision-maker can accept or refuse according to the patient's advance directive or previously expressed wishes (PMID 25039796).

Subsequent data point in the same direction.

• Among 764 patients with progressive dementia in 2 Hong Kong geriatric hospitals (464 with nasogastric tubes, 300 with hand feeding), the 1-year survival rate showed no difference at 36% versus 37% (P=.71). Among the 577 who survived until discharge, pneumonia was lower in the hand-feeding group (48% vs. 60%; P=.004), and after adjusting for other factors, the nasogastric tube emerged as a risk factor for pneumonia (aHR 1.41; 1.08–1.85). Behavioral feeding difficulties refer to difficulty eating not because of swallowing strength, but due to behaviors such as keeping the mouth closed or holding food in the mouth. However, in a subgroup analysis dividing patients by problem type, this association was significant only in the group with both dysphagia and behavioral feeding difficulties (P=.01), and was not significant in the group with dysphagia alone (P=.30) or the group with behavioral feeding difficulties alone (P=.24) (PMID 35489380).
• In a prospective propensity score-matched study from two geriatric hospitals (283 patients, mean age 89.8 years; 176 matched pairs), 1-year survival also showed no difference (AHR 0.85; 0.59–1.23; P=.38), and pneumonia was higher in the nasogastric tube group (AHR 1.60; 1.06–2.42; P=.03) (PMID 40796169).
• In a meta-analysis pooling 12 studies and 1,805 tube-fed patients versus 3,861 non-tube-fed patients, tube feeding was associated with increased mortality (OR 1.79; 1.04–3.07). Pneumonia and pressure ulcers showed no association in the initial analysis, but in a sensitivity analysis, PEG showed a high risk of pneumonia (OR 3.56; 2.32–5.44) and pressure ulcers (OR 2.25; 1.92–2.63). There was no association with survival duration and nutritional status (PMID 32736992).
• In US nursing home data, tube insertion was 53.6 per 1,000 residents, and 68.1% were inserted in acute care hospitals. 1-year mortality after insertion was 64.1%, the median survival duration was 56 days, and 19.3% had to have their tubes replaced or repositioned within 1 year (PMID 19426943).

There are also data pointing in the opposite direction. This is a study reviewing 185 patients in 9 psychiatric hospitals in Okayama Prefecture, Japan, who had difficulty eating by mouth and whose attending physicians judged they could not live without long-term artificial nutrition. Dementia was the most common, and psychiatric disorders such as schizophrenia were also included. The median survival duration was 711 days for those who received tube feeding and 61 days for those who did not (nasogastric tube 611 days, PEG over 1,000 days). The authors concluded that enteral nutrition increases survival even in patients with dementia (PMID 29178502). However, the comparator in this study was "cases without artificial nutrition," whereas the previous progressive dementia data compared tubes with "careful hand feeding." Because the comparison groups differ, the conclusions may diverge.

Therefore, what the evidence says in this area is precisely this much. Evidence confirming that tube feeding increases survival in progressive dementia has not been found, and evidence confirming that it reduces pneumonia has not been found either. That is not the same sentence as 「tubes must not be inserted」. In reality, how much tube feeding is used is not determined by evidence alone. A commentary written about patients with progressive dementia in US nursing homes points out that financial incentives, avoidance of regulatory sanctions, reputation, and lawsuits, the urge to 「do something」, and basic misunderstandings about malnutrition and aspiration pneumonia drive up its use (PMID 17498602).

The Structure of Decisions — Is It Reversible, What Do You Gain and What Do You Lose?

If we translate the data up to this point into the language of decisions, a few axes remain. These axes do not give an answer, but show the shape of the questions we must ask.

① Is this decision reversible? Tubes can also be removed. Of stroke patients who received a feeding tube and came to a rehabilitation hospital, 45% had the tube removed and returned to oral meals at an average follow-up of 2 years (PMID 16136260). In a cohort of 427 acute stroke survivors, 5% needed a tube at 6 months (PMID 34280690). On the other hand, the procedure itself carries risks that may be difficult to reverse. Among stroke patients, those who received a direct enteral tube such as a PEG tube were discharged to long-term care facilities at a higher rate than those who only used a nasogastric tube (PMID 29367443). In progressive dementia, tube feeding is associated with increased use of restraints, according to an American Geriatrics Society statement (PMID 25039796). People who had a tube inserted in U.S. nursing homes were hospitalized an average of 1.0 times over the following year (PMID 19426943). Whether this is due to the tube itself or to the condition that led to the tube carries the exact same confounding by indication problem from the previous section.

② Are there means to buy time? There are attempts to slow the flow where a nasogastric tube keeps slipping out and transitions to PEG. Evidence for maintenance measures such as a nasal bridle, which loops inside the nose to secure the tube, or mitts that wrap around the hands is sparse and methodologically weak, and evidence on how well patients accept them is lacking (PMID 28793390).

③ What do you gain and what do you lose? Among the outcomes measured by a Cochrane review, the areas where PEG was clearly better were intervention failure (RR 0.18) and quality of life items (discomfort, body image, social activities). There was no difference in mortality and aspiration pneumonia (PMID 25997528). Conversely, tubes come with tube accidents (PMID 39985956), and in progressive dementia, even an increase in restraint use is added to the list (PMID 25039796).

④ Are withholding and withdrawal the same? Withholding is not starting a treatment in the first place, and withdrawal is stopping a treatment that was already underway. In the American fields of medicine, ethics, and law, there is a general consensus that the withholding and withdrawing of life-sustaining treatment at the end of life are morally the same. However, the same literature adds that despite this consensus, withdrawal is emotionally difficult, and the withdrawal of artificial nutrition and hydration is especially so. It also notes that recent literature challenges the evidence that tubes cause great harm in progressive dementia, and mentions time-limited trials as a practical method in clinical practice (PMID 27348795). This is a description of the U.S. context and is not a statement about the Korean system.

⑤ Does giving more information change the decision? In a randomized trial of 256 pairs of patients with progressive dementia and surrogate decision-makers across 24 nursing homes, a decision aid that organized choices and pros and cons increased knowledge (16.8 vs. 15.1), lowered decision conflict—meaning the degree of uncertainty about which side to choose—(1.65 vs. 1.90), and increased discussions with medical staff (46% vs. 33%). The group that received the tool received more dysphagia diets (89% vs. 76%). However, tube feeding was rare in both groups even after 9 months (1 person vs. 3 people), so whether there is a difference in use could not be determined (PMID 22091750). The conclusion of a literature review examining 4 types of decision aids is the same. Decision conflict and knowledge consistently improved, but the preference for tube feeding did not change (PMID 36150408).

This final piece of data best illustrates the nature of this decision. Information reduces decision conflict, but it does not make the choice for you. Noting that simply handing over information does not change the preference for or use of tube feeding, the authors of the literature review wrote that a more interactive type of intervention is needed (PMID 36150408). Whether that is a limitation of the tools or a property of this decision has not yet been determined.

Common Misconceptions

"Putting in a tube prevents aspiration pneumonia." 44% of acute stroke patients fed via nasogastric tube developed pneumonia, and most of them did so on days 2 to 3 (PMID 15145999). Pneumonia in 58% and aspiration pneumonia in 13.6% have been reported in stroke patients who received a PEG (PMID 39190711). A Cochrane review found no difference in aspiration pneumonia between PEG and nasogastric tubes (PMID 25997528).

"It has been confirmed that tube feeding makes outcomes worse." The strong association between tube feeding at discharge and 90-day disability and death in stroke is from observational data, and the authors called this a marker of poor prognosis (PMID 31866129). This association alone cannot separate whether the tube is the cause. The evidence neither confirms nor rules out the harm of tubes. In progressive dementia, nasogastric tubes were associated with an increased risk of pneumonia even after adjusting for other factors (aHR 1.41, AHR 1.60) (PMID 35489380, PMID 40796169).

"PEG is better than a nasogastric tube" or vice versa. Neither holds true without accounting for timing. FOOD Trial 3 reported that PEG in the early acute phase had 7.8% more deaths or poor outcomes (PMID 15733717), and data from different phases and designs point in the opposite direction (PMID 34514181, PMID 41496648). A Cochrane review sees no difference in mortality (PMID 25997528).

"Once someone becomes unable to swallow, they usually stay that way." In a cohort of 427 acute stroke survivors, the vast majority of those who experienced Dysphagia were eating by mouth by 6 months, and 5% needed a tube at 6 months (PMID 34280690). Even 24.2% of those unable to eat by mouth on day 10 were eating by mouth at 3 months (PMID 22947895). However, in a hospital data review of 48 people whose severity persisted for up to 3 weeks, half (50%) died within 3 months (PMID 41232430).

"Putting in a tube in progressive dementia makes you live longer." There are no randomized trials (PMID 19370678). In two cohorts from Hong Kong, 1-year survival showed no difference, and pneumonia was higher in the nasogastric tube group (PMID 35489380, PMID 40796169). However, in the first cohort, that pneumonia association was significant only in the group with both Dysphagia and behavioral feeding problems, and was not significant in the group with Dysphagia alone (P=.30) (PMID 35489380). There are also data pointing in the opposite direction showing increased survival, but that study compared tubes to cases with no artificial nutrition rather than hand feeding (PMID 29178502).

Honest Unresolved Questions

• There has been no large-scale trial randomizing this question again since FOOD. The 5.8% reduction in mortality with early tube feeding has remained unchanged for over 20 years with a confidence interval crossing zero (PMID 15733717).
• There are no randomized trials at all for tube feeding in advanced dementia (PMID 19370678). What exists are only observational data, and observational data cannot eliminate confounding by indication in principle.
• There are not yet any recovery prediction tools confirmed to hold true when applied elsewhere. The PRESS seen in this article fit well where it was developed, but its goodness of fit broke down in validation outside the United States (PMID 30742198, PMID 41687691).
• Quality-of-life outcomes have rarely been measured. Among the studies reviewed by the Cochrane dementia review, not a single one looked at quality of life (PMID 19370678).
• Decision support tools change knowledge and conflict, but they have not changed preferences for tube feeding (PMID 22091750, PMID 36150408). Whether that is a limitation of the tools or the nature of the decision has not yet been determined.
• The gap between guidelines and practice is wide. Despite recommendations to delay early PEG, 53% of PEGs in US data were performed within 7 days of admission (PMID 27965430).

To Go Deeper

• Divide the design first. Depending on whether it is randomized or observational, the conclusion of tube feeding literature changes. Benefits seen in observational data sometimes become nonsignificant when only randomized trials remain or when adjustments are made (sensitivity analysis in PMID 41496648, adjustment in PMID 29367443).
• Divide the outcome indicators. We look at whether they are nutritional indicators (albumin and weight), tube maintenance (intervention failure), clinical events (pneumonia and death), or functional outcomes (mRS and FOIS). In the FOOD trial 3, deaths were the same and functional outcomes differed (PMID 15733717).
• Check the timing. We look at whether it is the first 2 to 3 weeks of the acute phase, the recovery phase, or long-term care. If you mix data from these three phases, contradictory conclusions hold true at the same time.
• Check the population. We look at whether it is stroke, progressive dementia, or cancer. If the subjects are different, the question of the decision itself is different.
• If you leave just one thing, it is this: The bad outcomes seen in observational data of tube feeding are mixed together with the reason for receiving the tube, and before peeling that away, it cannot be read as the effect of the tube.

Source Papers

Design and cohorts are listed together. Figures already in the body are not repeated. ⚠️ indicates data showing indication bias in the body.

Where This Decision Stands — PMID 26863627 (2016) 570 consecutive admissions for acute ischemic stroke · PMID 31866129 (2020) 201 prospective cases in a stroke unit, FOIS scale and "tube feeding at discharge as a prognostic marker" · PMID 17498602 (2007) Commentary on institutional factors driving up tube feeding use in the United States.

The FOOD Trials — PMID 16409880 (2006) 5,033 patients across 131 hospitals in 18 countries, overall report of the three-trial bundle · PMID 15733717 (2005) Main report of Trials 2 and 3 (early vs. avoided, PEG vs. nasogastric tube) · PMID 12750536 (2003) Nutritional status-outcome observational data from the same cohort, showing the gap with Trial 1.

PEG or Nasogastric Tube? — PMID 25997528 (2015) Cochrane review of 11 papers with 735 patients · PMID 41496648 (2026) Meta-analysis of 10 papers with 1,416 patients on post-stroke dysphagia, ⚠️mortality benefit disappears when leaving only randomized trials · PMID 34514181 (2021) Propensity score adjustment in a convalescent rehabilitation ward.

What Tubes Cannot Prevent — PMID 15145999 (2004) 100 prospective acute stroke cases with nasogastric tube feeding · PMID 39190711 (2025) Meta-analysis of 22 papers with 996,567 stroke patients, ⚠️control group definition absent in abstract · PMID 39985956 (2025) 30 papers on safety incidents by feeding method · PMID 18651972 (2008) Endoscopic confirmation of tube position and worsened swallowing.

Why Association Is Not Causation — PMID 29367443 (2018) Propensity score matching of 1,448 Ontario registry subjects, ⚠️benefit vanishes with a single adjustment · PMID 39976967 (2025) 143,331 elderly dementia patients · PMID 34280690 (2021) 427 prospective acute stroke cases, 5% needing a tube at 6 months.

The Axis of Time — PMID 22947895 (2012) 3-month outcomes for 723 patients unable to take oral nutrition at day 10 · PMID 41232430 (2025) 48 patients with severe conditions persisting up to 3 weeks · PMID 16136260 (2005) Mean 2-year follow-up after gastrostomy · PMID 30742198 (2019) Development and validation of the PRESS prediction score · PMID 41687691 (2026) External validation in 149 patients across 2 US institutions, failure of fit · PMID 27965430 (2017) Timing of PEG placement in 34,623 US inpatient sample cases · PMID 38112750 (2024) Single-institution data showing how rehabilitation transfer pathways shape outcome numbers.

Progressive Dementia — PMID 19370678 (2009) Cochrane review, 0 randomized trials · PMID 25039796 (2014) American Geriatrics Society position statement on careful hand-feeding · PMID 35489380 (2022) 764 patients in Hong Kong, ⚠️pneumonia association significant only in a subgroup · PMID 40796169 (2025) Prospective propensity score matching across 2 geriatric hospitals · PMID 32736992 (2021) Meta-analysis of 12 papers · PMID 19426943 (2009) Tube insertion rates and median survival in US nursing homes · PMID 29178502 (2017) 185 patients in Japan, ⚠️data pointing in the opposite direction — comparator was no artificial nutrition rather than hand-feeding.

The Structure of Decision — PMID 28793390 (2018) Review of evidence for tube maintenance methods · PMID 27348795 (2016) Withholding and withdrawing, time-limited trials · PMID 22091750 (2011) Randomized trial of a decision aid for 256 nursing home pairs · PMID 36150408 (2022) Literature review of 4 decision aids.