Table of Contents

Introduction
Evaluation of the Risk for Developing Aspiration Pneumonia:
    History and Physical Exam
    Modified Barium Swallow
Use of Feeding Tubes
Tube Feeding in Dementia
The Problem with Randomization
Final Conclusion

---

Tube Feeding, Aspiration Pneumonia, and Dementia

 

“When practice precedes evidence”

 

 

 

Lee Colglazier, M.D.

Wake Forest University, Baptist Medical Center

December 10, 2002

 

 

Introduction

 

The issue of tube feeding is a common dilemma faced by most general internists when taking care of patients on medical wards and in nursing homes.  Controversy surrounding the issue, along with contradictory conclusions drawn by the literature, makes diagnostic and therapeutic decision making difficult.  Although there are many reasons to provide feeding tubes, prevention of aspiration pneumonia is one of the most common in hospitalized and nursing home patients.  Caregiver and family concern for inadequate nutritional intake in patients with dementia is a common indication as well.   For the indication of preventing aspiration pneumonia, a common algorithm in many hospitals is to begin by identifying individuals considered at risk for aspiration pneumonia by history and physical.  Swallowing evaluations, both bedside and modified barium video fluoroscopy (MBS), are then routinely preformed in those considered at risk.  Finally, in those who fail modified barium video fluoroscopy, it is common practice to withhold oral intake, with the consideration of enteral nutrition via nasoenteric, percutaneous gastrostomy (PEG), or jejunostomy (J) feeding tubes.  It is also common practice to offer families of demented patients who have stopped eating the option of enteral nutrition through a feeding tube.  The purpose of this paper is to explore the evidence behind these practices.

Two cases will be used to illustrate the major issues that will be addressed in this paper.  Concern for aspiration pneumonia is frequent on inpatient wards; therefore the first case does not represent a specific patient, but is rather a representation of a common situation.  The second case addresses a more specific patient population.

 

Case # 1 

 

Mr. X is an elderly man admitted to the hospital with an exacerbation of one of his chronic medical problems.  On rounds, his nurse informs the team he has been observed several times to choke and cough while eating.  Speech and swallowing therapy is consulted, and after performing a bedside swallowing evaluation, recommend a modified barium swallow.  Later that day, the speech therapist informs the team the patient has failed all parts of the modified barium swallow, and recommends immediate discontinuation of oral intake because of concern for aspiration pneumonia.

 

Questions

1.        What elements of the history and physical are useful in predicting those at risk of aspiration pneumonia?

2.        What are the performance characteristics of the modified barium swallow in identifying those at risk for aspiration pneumonia?

3.        How effective are feeding tubes at preventing aspiration pneumonia in those at increased risk?

 

Case # 2

 

Mrs. Y is a 83 year old female with a history of progressive dementia of vascular and Alzheimer’s type over the past 10 years.  Four years ago she was placed in a nursing home when her family could no longer care for her.  Over the past 6 months, she had a progressive decline in functional status and cognitive ability, and became dependent on others for all of her activities of daily living.  Recently, her nutritional status declined to the point that she stopped eating, and lost 20 pounds in the last 3 months.  While her family recognizes her poor prognosis and has declared her DNR, they are afraid she will starve to death and request enteral nutrition.

 

Questions

1.        Does tube feeding reduce mortality in patients with advanced dementia?

2.        Does tube feeding reduce mortality in a general nursing home population?

3.        Which patient populations are likely to benefit from tube feeding?

 

Evaluation of the Risk for Developing Aspiration Pneumonia

 

History and Physical Exam

 

Autopsy studies have identified aspiration pneumonia as the primary cause of death in 65% of deaths from terminal illness.¹    The term aspiration pneumonia refers to the inspiratory sucking into airways of fluid or foreign bodies leading to pneumonia.  There are several potential sources for aspiration; including food, oral secretions, and refluxed gastric contents.  Most literature focuses on the contribution of swallowing dysfunction while eating to aspiration pneumonia, and thus being an indication for feeding tubes.  The first focus of this paper will be on the usefulness of history at identifying those at risk for aspiration of food, and the development of aspiration pneumonia.

 

Dysphagia

 

The process of chewing and swallowing requires at least 6 cranial nerves, the first 3 cervical segments, and the 26 muscles of the mouth, pharynx, and esophagus.  There are numerous mechanisms, which can interfere with the process, causing subjective difficulty in swallowing known as dysphagia.  Dysphagia is a complaint that often prompts the initial work up for swallowing evaluation.  The incidence of dysphagia in acute hospitalized patients has been found to be 10 – 15 %, in people with neurological disorders including stroke and head injuries 25 – 50%, and in nursing home residents 30 – 40%.¹  Perhaps because of the high frequency of dysphagia, it has not been found to be associated with an increased risk of aspiration pneumonia in hospital and nursing home settings.  In one prospective study of 69 nursing home residents using aspiration pneumonia as the primary outcome, there was no statistically significant difference in rates of dysphagia between those who did and did not develop aspiration pneumonia.²   There were, however, several medications found to be significantly associated with aspiration. (Table 1).

 

 

* Italics indicate factors not statistically significant.

 

 

 

A similar study by Langmore et al.³ conducted among nursing home, inpatient, and outpatient settings drew the same conclusion.  Among the most important historical findings associated with development of aspiration pneumonia was dependence on others for feeding.  Interestingly, level of alertness was not associated with increased risk of aspiration pneumonia.  In Cogen et al. ²⁶, a previous history of aspiration pneumonia was the strongest predictor of future aspiration.

 

Bedside Swallowing Evaluation

 

Along with dysphagia, observation of choking and coughing with eating has also been considered a warning sign for aspiration.  This has prompted the development of the bedside swallowing evaluation.  One type of evaluation, called the 3-oz water test, was studied by DePippo et al.⁴  In the trial, 44 patients at a stroke rehabilitation unit were identified who had one of the following features: 1) bilateral hemispheric stroke, 2) brain-stem stroke, 3) history of pneumonia during acute stroke phase, 4) coughing associated with feeding, 5) failure to consume half of meals, 6) prolonged time required for feeding, and 7) non-oral feeding program in progress.  Patients were given a 3-oz. cup of water and asked to drink without interruption.  Coughing while drinking or for one minute after drinking, or the presence of a post-swallow wet-hoarse voice quality was scored as abnormal.  Patients then underwent a modified barium swallow evaluation.  The outcome studied was performance on the MBS.

As shown in Table 2, the 3-oz. water swallow test identified 80% (16/20) of patients who aspirated during subsequent video fluoroscopic modified barium swallow examination. (Sens = 80%,  Spec = 54%,  (+) LR = 1.74,  (-) LR = 0.37).  The sensitivity was higher for identifying those who aspirated large amounts defined as >10% of bolus (Sens = 88%,  Spec = 44%, (+) LR = 1.57, (-) LR = 0.27) or more solid consistencies (Sens = 89%, Spec = 46%, (+) LR =1.65, (-) LR = 0.24).  The authors concluded that the 3-oz. water swallow test is a sensitive screening tool for identifying patients at risk for clinically significant aspiration who need referral of more definitive modified barium swallow evaluation.

Although the test was sensitive, the false positive rate was high (46%, 56%, and 54% for all aspiration, those aspirating large amounts, and those aspirating solid consistencies respectively).  The main limitation of the study was that aspiration pneumonia was not an endpoint, but rather performance on MBS.  The utility of the swallowing evaluation is therefore dependent on the utility of the MBS, which will be discussed later in the paper.  With regards to the MBS, the marginally elevated likelihood ratio for a positive test result does not make it useful to predict what the results of the MBS will be.  For example, a pretest probability of 50% in someone who fails the 3-oz. swallow will only give a posttest probability of 64%.  A negative test result (passing the swallowing study) similarly is not useful at identifying those that may pass the MBS.  A likelihood ratio of 0.37 for a pretest probability of 50% will only reduce the probability someone will passes the MBS to 27%.

 

 

 

Gag Reflex

 

                Another physical exam finding commonly used in clinical practice is the assessment of the gag reflex.  The assumption is that the gag reflex is a defense mechanism against aspiration, and the lack of one could put someone who eats at increased risk.  This has also been evaluated by Leder et al.⁵ at Yale University.  Over a 7-week period, 100 patients considered at risk for aspiration underwent video fluoroscopic evaluation of aspiration, with concomitant clinical evaluation of gag reflex.  The gag reflex was evaluated by touching the posterior tongue or pharyngeal wall with a tongue blade, and observing for elevation of larynx and pharynx with contraction of the soft palate.  Outcome measures were aspiration, as defined by penetration of material below the level of the true vocal cords on MBS.  Upper GI series and small bowel series were also used  to evaluate for aspiration.

In the results, 15 patients exhibited objective documentation of aspiration with video fluoroscopy, 14 of the 15 (93%) having a normal gag reflex.  As shown in Table 3, 20 patients were considered to have an abnormal gag reflex, 19 of the 20 (95%) without a gag reflex had no evidence of aspiration of fluoroscopy  (Sens. = 7%, Spec. = 78% (+) LR = 0.32, (-) LR = 1.19).  As shown in Table 4, using only the MBS, outcomes were similar (Sens. = 8%, Spec. = 63% (+) LR = 0.22, (-) LR = 1.46).  The authors concluded the presence of a gag reflex does not protect against aspiration, and the absence of a gag reflex does not predict aspiration.  The main limitation of the study is that aspiration pneumonia was not studied as an outcome measure, rather results of video fluoroscopic examination were.  Although gag reflex is not predictive of performance on MBS, its relation to aspiration pneumonia was not determined.  However, it is reasonable to conclude the gag reflex should not be used in the decision of whether or not to obtain a MBS.

 

 

 

 

 

Modified Barium Swallow

 

The modified barium swallow has become the standard for assessing swallowing difficulties in many hospitals.  Several studies attest to the clinical utility of the MBS, however there are an equal number of articles discounting the utility of the MBS (Table 5).  Several of the articles listed in Table 5 are retrospective, others do not use aspiration pneumonia as an outcome measure, and still others are confounded by use of feeding tubes in the study population.  The three articles, which were prospective and used aspiration pneumonia as an outcome measure, will be discussed below.

The first study by Aviv et al.⁶ was conducted at the Columbia Presbyterian Medical Center.  The study group included 20 stroke patients admitted to a neurovascular unit.  All patients had a stroke within the past 4 weeks, and all had complaints of dysphagia accompanied by coughing or choking while swallowing.    Modified Barium Swallow (MBS) and laryngopharyngeal sensory discrimination testing (LPSDT) were preformed.  LPSDT involves delivering pulses of air to the mucosa innervated by the superior laryngeal nerve, a sensory nerve felt to be important for protective mechanisms.   Patients were prospectively followed for at least 2 years to identify the frequency of aspiration pneumonia.  Those who failed MBS were treated with a PEG tube, with the assumption that PEG tubes prevent aspiration pneumonia.  The authors’ purpose was only to assess the negative predictive value of the test.

As shown in Table 6, MBS identified 10 patients not at risk for aspiration pneumonia, with 4 out of 10 developing aspiration pneumonia in the follow up period.  In table 7, MBS + LPSDT identified 5 pts not at risk, none of which developed aspiration pneumonia in the follow up period.  The authors concluded the negative predictive value was significant, and that the addition of LPSDT provided a more sensitive test.

 

 

 

One major limitation of the study was the very small sample size.  Another limitation was the differential treatment of those who failed versus those who passed MBS.  Those who failed MBS received a feeding tube, which the authors assumed prevented aspiration pneumonia.  Since feeding tubes were assumed to prevent aspiration pneumonia, the authors concluded pneumonia rates in these groups did not accurately reflect what they would have been without intervention.  This logic prevents the calculation of accurate sensitivity, specificity, and likelihood ratios.   A false positive rate would also not be accurate, because its calculation requires the total incidence of pneumonia, which the authors conclude is falsely low.  However, since the incidence of pneumonia was 0 in those that passed MBS and LPSDT, the authors felt they could still state the false negative rate was 0%, and therefore conclude MBS and LPSDT is a useful test.  There are several problems with the conclusion.  First, it is based on the assumption that tube feeding prevents aspiration pneumonia.  Second, not knowing other performance characteristics of the test such as likelihood ratios does not provide clinically useful information, especially if the test results will be used to decide upon an invasive procedure (the potential for false positives is not known).  Also, because of the small sample size the calculated 95% confidence interval for the false negative rate is 0 - 84%, and therefore not useful.

 

 

 

 

 

 

 

 

The next study by Teasell et al.⁷ done in Canada in 1996 also concluded the clinical utility of the MBS.  The study group consisted of 441 consecutive stroke patients admitted to a stroke rehabilitation unit within 4 months of their stroke over an 8-year period.  Video fluoroscopic modified barium swallow was preformed on 106 patients who were considered at risk of aspirating.  Failing MBS was considered if any amount of barium passed the vocal cords.  The amount or consistency was not considered.  The outcome measured was development of aspiration pneumonia during the inpatient stay.

No mention was made of use of feeding tubes during the study.  10 out of the 84 patients who failed the MBS (11.9%) developed pneumonia while in the hospital.  2 out of the 357 patients within the study (0.6%) who were presumed non-aspirators developed pneumonia.  The authors concluded the MBS was a useful test at predicting aspiration pneumonia, quoting a 20-fold difference in aspiration rates between the two groups. 

However, one limitation of the study was that the authors used as their control group those not considered at risk for aspiration pneumonia.  Therefore, the control group consisted not only of the 22 patients who passed the MBS, but also 335 of the original group who did not take the MBS as they were considered at low risk for aspiration pneumonia.  As only 6% of the control group took the MBS, the potential exists for a low risk population to potentially dilute false negative test results.  To adequately assess a test’s characteristics, only outcomes in those who took the test should be evaluated.  Table 8 lists the results of aspiration after removing the people who did not undergo the MBS and recalculating test characteristics using only those who did.  (Sens. = 83%, Spec = 21%, (+) LR = 1.05, (-) LR = 0.81).  Calculation of 95% confidence reveals the likelihood ratios are not statistically significant.

 

 

The next prospective study by Schmidt et al.⁸ also concluded clinical utility of MBS.  The study group included patients on an inpatient stroke and rehabilitation unit, who had a diagnosis of stroke.  26 patients with video fluoroscopic evidence of aspiration were compared to 33 randomly selected, case-matched, dysphagic controls without video fluoroscopic evidence of aspiration.  Patients were followed for 18 months post-stroke.  The patients who failed MBS did receive dietary modification and taught feeding techniques, but were not given feeding tubes.

After 18 months, 5 of the 26 patients who failed MBS developed aspiration pneumonia, compared to 1 of the 33 who passed MBS.  The Odds ratio calculated for pneumonia was 7.6 for those aspirating any amount of barium.  For thickened liquids or more solid consistencies the OR was 5.6.  The Odds ratio for death was 9.2 for those aspirating thickened liquids or more solid consistencies compared with those who did not aspirate or who aspirated thin liquids only.  All were statistically significant (p < 0.05).  The authors concluded that MBS was a clinically useful study. 

The authors chose to calculate the Odds ratio because of the case matched nature of the study, even though the study was prospective.  The results do indicate there is an association between failing MBS and development of aspiration pneumonia, however it does not help one understand the predictive value of the test.  Since the study was prospective, other test performance characteristics can be calculated (Table 9).

 

 

Although one is more likely to develop aspiration pneumonia if one fails MBS, 21 out of 26  (80%) of patients who failed MBS never developed AP, giving a positive predictive value of only 20%.  The false positive rate of 40% is significantly high for a test, which may be used to justify invasive procedures, as well as have a negative impact on quality of life in by taking away the enjoyment of eating.  Again, limitations of the study were differences in treatment between study groups.  Those who failed MBS received dietary counseling, which may have lowered the incidence of pneumonia.  Also, there would be less chance of bias if an entire group who took the MBS were followed together, instead of selecting a case matched cohort who passed the MBS.

 

Use of Feeding Tubes

 

When first evaluating the value of a therapeutic intervention, it is important to address whether the intervention itself could cause harm.  One of the first trials to assess the risk of gastric tubes was not preformed on patients being tube fed, but rather in surgical patients in whom nasogastric (NG) tubes were placed routinely after surgery.  A common practice of the past and present, nasogastric tubes are often

inserted postoperatively to prevent abdominal distention.  W. J. Mayo, of the Mayo Clinic

once stated he “would rather have a resident with a nasogastric tube in his pocket than a stethoscope.”⁹ This practice was questioned by Argof et al.⁹ in the 1970s.  The study group consisted of 300 consecutive surgical patients who underwent abdominal operations between 1975 and 1976.  The first consecutive 150 patients were assigned to an NG tube, and the second consecutive 150 patients were not.  The outcome measured was pneumonia defined as: 1) temp of 38.5 to 39 on 1^st to 3^rd postoperative day, 2) infiltrate on CXR, 3) positive sputum cultures, and 4) deterioration of pulmonary function.  23 out of the 150 NG patients (15%) developed pneumonia, as opposed to 2 out of 150 NG free patients (1.5%).  The difference was statistically significant ( p < 0.05).  The author concluded the practice of routine NG tube insertion is not justified.

 

 

                The study design had some limitations.  First, it was not randomized.  Since patients were selected consecutively, the potential exists for differences between the two study groups.  Even though it was not randomized, since patients were consecutively assigned there is less chance for conscious selection bias of placing NG tubes in sicker patients.  In trying to apply the study to use of feeding tubes, this study is limited by the younger and generally healthy surgical patient population compared to the nursing home and medical ward patients commonly tube fed.  Also, patients in this study were not tube fed, but rather attached to wall suction, which could potentially allow for complications.  In addition, nasoenteric tube feeding is often done with smaller bore tubes, often times placed in the duodenum.  Currently, fine bore tubes (such as the Dobhoff tube) have not been studied head to head against NG tubes.  However, despite these limitations, the significantly high complication rate of aspiration pneumonia with a number needed to harm of 7 for NG tubes brings up obvious concerns about nasoenteric feeding for the preventing of aspiration.

Without addressing benefit, the above study points out potential harm that may be caused by nasoenteric tube feeding.  One theory regarding the possible contribution of nasoenteric tube fed patients to aspiration pneumonia is the reduction of the lower esophageal pressure by the tube.  In Winterbauer et al.¹⁰, aspiration was demonstrated on video fluoroscopy 38% of the time in NG fed patients.  However, PEG and J tubes do not pass through the lower esophageal sphincter, and are theorized to have less incidence of aspiration pneumonia.  Still, in anecdotal cases (Cole et al.¹¹) food has been video fluoroscopically observed to reflux and aspirate into the lungs.  Homer-Ward et al.¹² addressed NG versus PEG feeding in 1996.  In a randomized prospective trial consisting of 30 stroke patients, 16 were randomized to a PEG tube and 14 to NG for feeding.  The outcome studied was the 6-week mortality rate.  At the end of 6 weeks, 2 of the 16 patients (12.5%) in the PEG group died, compared with 8 of the 14 (57%) NG fed patients (RR = 4, NNH = 2.7, p < 0.05).  It is a reasonable conclusion from studies such as Argof et al.⁹ and Homer-Ward et al.¹² , that if tube feeding is going to prove to be beneficial, it is more likely be with a PEG or J tube than a nasogastric or dobhoff tube.

The initial studies regarding use of feeding tubes were all retrospective.  Use of feeding tubes especially with the invention of the MBS was theoretical and preceded any evidence either supporting or discounting it.  Its widespread use created difficulty in randomizing patients against oral feeding, and such a trial has not been done.  The next study by Croghan et al.¹³ involves residents of Johns Hopkins Geriatrics Center.  Records of 40 residents who underwent video fluoroscopic swallowing evaluation between 1987 – 1989 were retrospectively reviewed for age, sex, primary symptoms or clinical conditions prior to swallowing study, major diagnosis, concurrent illness, result of swallowing study, outcomes (feeding tube placement, re-hospitalization, prolonged nursing home stay, pneumonia, death, and pneumonia death.  The authors were unable to determine if pneumonia was caused by aspiration due to the retrospective nature.

                The group demographics were as follows: 20 female, 20 male, and mean age 68.6.  Primary symptoms were suspected aspiration (68%), dysphagia (50%), aphasia (30%), and dysarthria (23%).  Of the 22 patients with aspiration on MBS, 15 had feeding tubes placed and 7 did not.  The group with feeding tubes placed had a higher incidence of pneumonia and death (Figure 1) than groups that did not (P < 0.05).  Also, there was a statistically significant higher death rate in those that were fed via NG versus PEG tubes.

The major limitation to the study was the lack of randomization.  A conclusion of causality cannot be drawn given the possible the groups who had feeding tubes placed represented a population at higher risk for aspiration.  However, the study does show those with feeding tubes still have a significantly high mortality rate from aspiration pneumonia.

The next study by Pick et al.² was a prospective observational study of those in a nursing home to assess risk factors for development of aspiration pneumonia.  The study group included 257 patients at the Canandaigua Veterans Affairs Medical Center in New York.  All patients present at the VA nursing facility (257 patients) on a given date were followed prospectively for 8 months.  All had major degenerative neurological diagnosis, and were unable to perform ADLs requiring complex nursing treatments.  24% had feeding tubes.  Pneumonia was defined as either a) new infiltrate on CXR, b) new tachypnea with a respiratory rate > 18, c)new fever > 100 degrees, or d) change in mental status not defined by another cause.  Aspiration was defined as witnessed (choking after emesis or choking while eating) or unwitnessed (coughing while eating, displacement of an enteral feeding tube, or presence of vomitus or tube feeding on a pillow within 24 hours of diagnosis of pneumonia).  Patients were then followed for 3 more years with death as primary endpoint.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

At the end of the study there were 98 aspiration events attributed to 69 patients, 16 of which died.  39 out of 69 (57%) patients in the aspiration group were tube fed, as opposed to 27 out of 192 (14%) in the non-aspiration cohort.  The odds ratio for tube feeding was 7.9 (95% CI, 4.0 – 15.5).  There was no difference in incidence of aspiration between nasogastric and PEG tube cohorts (Figure 2).  At the 3-year follow up, 40% of the non-aspiration cohort had died, as compared with 83% of the aspiration cohort. (P = 0.001)

Again, because of the retrospective nature of the study, causality cannot be established, as it cannot be known what the mortality rate would have been if the group had not been tube fed.  This study further supports, however, that despite feeding tubes, there is still a significant mortality rate in those who are at risk of aspirating.

Another prospective observational study by Langmore et al.³ at a VA medical center in Ann Arbor, Michigan had similar results.  The study group included 189 elderly subjects in outpatient clinics, inpatient acute care wards, and nursing home at a VA center in Ann Arbor, Michigan.  Multiple risk factors were assessed, and they were followed for 4 years.  At the end of the study 41 of the 189 (21.7%) patients developed aspiration pneumonia.  27% of aspiration group were tube fed, as compared to 9% of non-aspiration group (OR = 3.031 p = 0.05).  The largest risk factor for aspiration was dependency on others for feeding (41% of subjects who developed aspiration pneumonia compared to 6% of patients without aspiration pneumonia.), with an odds ratio of 19.98 (p = 0.0001).

Looking at a population of healthier patients, Feinberg et al. ¹⁴ studied residents of the Philadelphia Geriatric Center skilled nursing facility.  No patients were considered critically ill, suffering from a rapidly terminal illness, or had significant congestive heart failure.  All had clear CXRs prior to study.  No patients had a history of artificial feeding.  The group consisted of 96 females and 56 males, with an average age of 86.  100 were referred because of dysphagia identified by their primary care physician or speech pathologist.  All underwent video swallowing examinations (MBS.)  Aspiration was defined as barium entering the laryngeal ventricle and beyond into the airway.  Patients who had abnormal MBS were referred to speech pathologist.  Management strategies included diet alterations, behavioral and environmental manipulation, improved positioning during mealtime, feeding and oral rehabilitation, and recommendations for artificial feeding.  NG tubes were used for less than 4 weeks, and PEG tubes were used for greater than 4 weeks.  Patients were followed for 36 months, having periodic examinations by a speech pathologist.  Any change of signs or symptoms warranted a repeat MBS.  Pneumonia was defined as fever, cough, tachypnea, altered mental status, rales/rhonchi, leukocytosis, and chest x-ray revealing infiltrate.

At the beginning of the study, there were 50 non-aspirators, 51 minor aspirators, and 51 major aspirators/oral feeders.  At the conclusion of study, there were  37 non-aspirators, 38 minor aspirators, 47 major aspirators/oral feeders, 30 major aspirators/artificial feeders.  Diagnosis of patients were: dementia in 31%, prior stroke in 21%, disuse deconditioning in 14%, acute stroke in 13%, Parkinson’s in 9%, and unknown in 13%.  Data was measured in terms of patient months within groups.  Frequency of pneumonia in groups were as follows: Non-aspiration = 0.6%; Minor aspiration = 0.9%; Major aspiration/oral feeding = 1.3%; Major aspiration/artificial feeding = 4.4% (p<0.001).  The difference among the first 3 groups were not significant, the only significant difference was within the tube fed group. 

Limitations of the study included lack of randomization, and lack of comparison in co-morbidity between the tube fed and non tube-fed groups.  Although it cannot be concluded that tube feeding causes aspiration pneumonia, the study does raise questions as to the efficacy of feeding tubes in preventing aspiration pneumonia.

Other studies support the conclusion that a significant percentage of tube fed patients continue to aspirate, especially those with a history of AP.  In Cogen et al.¹⁵ charts of 44 consecutive patients with a history of aspiration pneumonia who were fed via a jejunostomy tube were retrospectively reviewed.   They found 31.6 % continued to develop aspiration pneumonia.  A larger study also by Cogen et al.¹⁶ drew similar conclusions, where a review of 109 consecutive nursing home residents fed via G tube revealed 40.7% with a history of aspiration pneumonia continued to develop pneumonia.

 

Tube Feeding in Dementia

 

Volicer et al.¹⁷ specifically examined issues of aspiration in demented patients.  73 patients with dementia of the Alzheimer’s type (DAT) hospitalized for long term care on dementia unit were studied.  Functional status was determined by the mini-mental status exam, the MACC Behavioral Adjustment Scale, the Boston Diagnostic Aphasia Examination, and the Bedford Alzheimer Nursing Scale.  Eating status, diet, and caloric requirements were recorded.  From this 3 categories were derived: group 1 (17 patients) who could feed themselves independently, group 2 (13 patients) who had to be fed by nursing staff, but otherwise had no difficulties, group 3 (18 patients) who refused food although they were able to swallow it, and group 4 (18 patients) who had definite eating difficulties choking on liquid and/or solid food, despite maintaining oral intake.  Mortality rates were followed for 2 years.

At the end of 2 years, 6 patients died (35.3%) in group 1, 5 patients died (38.5%) in group 2, 10 patients died (55.6%) in group 3, and 10 patients died (43.5%) in group 4.   Only 1 of the 73 patients was tube fed.  There was not a statistically significant difference in mortality rates between the groups (p=0.41).  Mortality rates for subgroups based on weight were also not statistically significant.  The authors concluded aspiration risk in dementia patients was not a significant predictor of death.  Tube feeding impact was not evaluated in this study.  Since mortality rates were equally high in groups with and without aspiration risk, and those who were and were not eating, the authors concluded that intervention in different subgroups is not necessary. 

Sanders et al.¹⁸ also studied patients with dementia.  The study group was a cohort of 361 consecutive patients requiring PEG feeding between August 1992 and July 1997 in two British hospitals.  Indications for PEG included oropharyngeal malignancy, stroke, dementia, and neurological disorders such as Parkinsonism, multiple sclerosis, and cerebral palsy.  In the dementia group, metabolic and nutritional causes of dementia were excluded.  Patients were followed for 5 years with mortality as the primary endpoint. 

At the end of the study, the overall mortality rate was 28% at 1 month, 44% at 3 months, 52% at 6 months, and 63% at 1 year.  In the dementia group, mortality was 54% at one month, 78% at 3 months, 81% at 6 months, and 90% at 1 year.  (Figure 3.)  The mortality difference between dementia group and stroke or oropharyngeal malignancy groups was statistically significant (p<0.0001).

 

Figure 3: Survival graph of patients undergoing gastrostomy insertion divided by diagnosis¹⁸

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Limitations of this study include lack of comparison to orally fed groups, and lack of data of cause of mortality.  However, the dramatic 1-year mortality rate of 90% still questions the role of tube feeding in demented patients.

   

Meier et al.¹⁹ studied demented patients at Mt. Sinai hospital in New York City.  99 hospitalized patients with advanced dementia and a surrogate decision maker were followed during hospitalization, and subsequently 3 years after discharge.  The average age was 85 years old, 29% lived at home and 70% from a nursing home.  44% of patients were admitted with a diagnosis of pneumonia.  51 out of 99 (50%) of patients received a feeding tube, 17 out of 99 (17%) already had a feeding tube, and 31 out of 99 (31%) left the hospital without a feeding tube.  Patients were followed for 3 years with mortality as the primary endpoint.

The median survival for patients receiving or not receiving feeding tubes was not statistically significant (195 days and 189 days respectively, p = 0.9).  The 6-month mortality was 50% in both tube fed and non tube fed groups.  Limitations of the study include lack of cause of mortality, as well as lack of randomization.  Still, as in Meier et al.¹⁹, the mortality in advanced dementia with tube feeding is very high, and would seem to question its benefit.

 

The Problem of Randomization

 

A common limitation to all of the studies above is the lack of randomization.  One could make the case that groups who were tube fed in these trials may have been different than the groups who were not tube fed.  As it makes sense that sicker patients could be more likely to be tube fed, they may also be more likely to have a higher mortality rate.  To date there are no studies randomizing patients to tube feeding versus oral feeding.  Even though it lacks evidence, tube feeding has practically become the standard of care in many settings.  There are also many ethical considerations to designing such a study where tube feeding would be withheld, or those considered at aspiration risk would be allowed to eat.  Also, decisions around tube feeding are emotional for family members and recruitment of high numbers could be difficult.  For these reasons, there may never be a randomized trial with tube feeding to prove its benefit or harm.  Yet without optimal randomized trials, physicians still need to make evidenced based decisions regarding treatment in patients who are at risk of aspirating, or who have dementia and are not eating.  Two studies have attempted to account for lack of randomization by designing prospective trials that follow tube fed patients, as well as case matched cohorts who are not being tube fed.  Although it is not randomization, it is an improvement on prospective observational comparisons that are full of confounding variables.

Mitchell et al.²⁰ made an attempt to find a solution to lack of randomization.  The design was a prospective study of newly initiated tube fed nursing home patients who had recent progression to severe cognitive impairment compared to a matched, non tube fed cohort with similar co-morbid conditions, using survival as the primary outcome measure.   Subjects were all residents greater than 65 who lived in a nursing home in Washington State.  Subjects were derived from the Minimum Data Set, which is a federally mandated functional assessment designed to identify needs of residents in long-term care facilities.  Trained nurses administer the assessment and submit it to central government offices who assess for completeness.  The original pool consisted of 272 nursing homes, with 36,498 residents in 1991.  Residents were identified with severe cognitive impairment using the Cognitive Performance Scale (CPS), which scores residents from 0-6, with 6 being residents who are totally dependent on others for feeding, including those in a coma.  A score of 5 means severe cognitive impairment, but not totally dependent on others for feeding.  From the data set, 1386 residents were identified with a CPS score of 5 or less who progressed within 24 months to a score of 6, but did not have a diagnosis of coma.  Also, subjects were not included if they had a feeding tube at baseline.  The median age was 87 (65 – 107), with 341 (24.6%) being men.  From the data set, independent variables felt to be associated with feeding tube placement were analyzed.  Variables included age, gender, health status, weight loss, body mass index, presence of pressure ulcers, chewing or swallowing problems, recurrent lung aspirations, unstable medical condition including terminal diagnosis, baseline cognitive status, baseline functional status, and resuscitation (DNR) status.  Patients were followed for 639 days for placement of a feeding tube, mortality, as well as the other variables.  Survival time was calculated from the time a patient became a level 6.

 

Table 11. Variables associated with increased risk of tube feeding20
Variables		Odds Ratio (95% Confidence Interval)
	Age < 87	1.85 (1.25 – 2.78)
	Recurrent Lung Aspiration	5.46 (2.66 – 11.20)
	Chewing and Swallowing Problems	3.00 (1.81 – 4.97)
	Pressure Ulcer	1.64 (1.23 – 2.95)
	Stroke	2.12 (1.17 – 2.62)
	Less Baseline Functional Impairment	2.07 (1.27 – 3.36)
	No DNR order	3.03 (1.92 – 4.85)
	No dementia	2.17 (1.43 – 3.22)

 

               

 

At the end of the study, less than 4% of patients were lost to follow up.  Of the 1386 study patients, 135 (9.7%) underwent placement of a feeding tube.  Table 11 lists the variables found to be statistically significantly associated with tube feeding placement.  Patients were followed for 639 days.  Cox proportional hazards linear regression was used to examine the association between feeding tube status and mortality.  In the unadjusted model, feeding tube was not associated with survival (RR = 1.06, 95% CI 0.81 – 1.39).  Adjusting for the variables in Table 11, feeding tubes were again not associated with survival (RR = 0.90, 95% CI 0.67 – 1.21) (Figure 4).  Looking at individual subgroups of the independent risk factors, there again was no difference found.

                Although the study tried to control for confounding variables, one major limitation for the study is that the possibility still exists that patients who were tube fed were still sicker and would have died sooner without being tube fed.

 

Figure 4: A, adjusted 24 month survival comparison of residents with severe cognitive impairment with (dotted line) and without (dashed line) feeding tubes.²⁰

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The focus of the previous study was in severely cognitively impaired patients.  A follow up study again by Mitchell et al.²¹ looked specifically at a younger, less cognitively disabled cohort who were being tube fed for chewing and swallowing difficulties.  The same population as the last study was used, 36,498 residents of the 272 nursing homes in Washington in 1991 analyzed based on the MDS (Minimum Data Set).  5,266 residents were identified to have chewing or swallowing problems, with a median age of 87 years and 72.4% female.  The cohort was then categorized into groups that did and did not have feeding tubes at baseline, and those who obtained feeding tubes during the study were excluded.  Variable were again chosen that might be associated with weather or not a nursing home resident would undergo feeding tube placement, as well as clinical characteristics that might influence survival.  Variables included age, gender, weight loss, body mass index, pressure ulcers, recurrent lung aspiration, presence of an unstable medical condition, and DNR status.  Cognitive function was again assessed using the Cognitive Performance Scale (CPS).  Baseline functional status was assessed using the Activities of Daily Living scores.  Survival time was calculated from time of baseline assessment.  Residents were followed for one year.

                Of the 5,266 residents, 551 (10.5%) had feeding tubes.  Variables that were associated (p<.10) with both survival and feeding tube status in bivariate analysis were considered as potential confounders.  These variables included baseline CPS score, weight loss, female sex, unstable condition, recurrent lung aspiration, a diagnosis of dementia, recent weight loss, low body mass index, atherosclerotic heart disease, and presence of pressure ulcers.  The potential confounding variables were entered into a Cox proportional hazards regression model.  After controlling for confounding variables, feeding tube status still was independently associated with poorer survival (RR = 1.44, 95% CI 1.17 – 1.76) (Figure 5).  An analysis of individual subgroups (age, baseline cognitive status, functional status, and aspiration) still revealed poorer survival.

                Among the 551 residents with a feeding tube, 121 (22%) died.  Of the 403 survivors, 108 (25.1%) eventually became and remained free of the feeding tube during the 12 months.  The only significant factor associated with tube removal was age < 87 years (OR = 1.66, 1.03 – 2.69).  The residents were younger and tended to have less dementia, diabetes, and aspiration history.

 

 

Figure 5: One Year Survival Comparison of residents with chewing and swallowing problems with (dotted line) and without (solid line) feeding tubes²¹

 

               

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Final Conclusion:

 

                This paper examined the evidence behind the current practice of identifying those at risk of aspiration pneumonia both by history, physical, and MBS, and the role of tube feeding for the purpose of preventing aspiration pneumonia.  It also explored evidence behind the use of feeding tubes in hospitalized and nursing home populations, including patients with dementia.

            Currently there is no accurate way of identifying those who will aspirate.  Perhaps the most useful element of the history and physical is a history of aspiration pneumonia, and dependence on others for feeding.  The use of the MBS is not clearly defined, mainly because there has not been a large trial in which patients who haven taken the MBS are simply followed without intervention.  It is clear from the data in the literature that those who fail a MBS are at increased risk of developing aspiration pneumonia.  What is not clear is how much more likely.  Most studies are limited by lack of numbers, differential treatment of the groups who pass and fail the test, and retrospective designs.  In the prospective studies available, the false positive rates in the studies appear to be high, around 40%.  This is an unacceptably high rate especially if the test is used to justify the invasive procedure of placing a feeding tube, as well as the dramatic reduction of life quality by taking away a patient’s enjoyment of eating. 

                The use of tube feeding has become ubiquitous and in some cases standard of care prior to any clear data supporting its benefit.  Because of that, and because there are many complicated ethical and emotional issues associated with families of patients, there may never be a randomized trial of tube feeding to oral feeding.  In a population of hospitalized and nursing home patients, there is much retrospective data associating tube feeding to worse outcomes.  Some have argued the reason for this is the bias of tube fed patients having poorer health at baseline.  To the contrary, Mitchell et al.²⁰ found that tube fed patients were more likely to be younger with a better cognitive and functional status.   Mitchell et al.²⁰ is perhaps the best attempt of a prospective study of tube feeding to account for confounding variables and match a prospective tube fed cohort with a cohort of similar co-morbidity.  Even after adjustment for confounding variables, there was still no evidence of lowered mortality.  There are many reasons to explain why mortality could be the same or even worse in hospital and nursing home tube fed populations.  Feeding tubes have been found to cause agitated behavior in cognitively impaired patients, which could lead to the use of restraints, psychotropic medication (which was shown by Pick et al.² to be associated with aspiration), immobility, fecal incontinence, and functional decline.  These conditions could lead to pressure ulcers and other complications, as demonstrated by Peck et al.²³ who showed tube fed patients were found to have restraints and pressure ulcers more often than matched controls without feeding tubes. There are also mechanical and surgical complications related to feeding tubes, which can cause high morbidity and mortality.  Also, feeding tubes do not prevent aspiration of oral secretions or refluxed gastric contents, which may have a significant role in the etiology of aspiration pneumonia.

                The reason the literature on tube feeding is not clear is that the practice preceded any evidence supporting it.  Given the potential for harm, there should be clear evidence to support the practice if it is to continue.  Although there is a lack of randomized trials, the best available evidence does not support the use of feeding tubes to lower mortality or morbidity in most hospitalized and nursing home patients.  Considering the potential harm they can cause, along with the high false positive rate with the current method of identifying those at risk for aspiration, use of feeding tubes should be pursued with caution, if at all.  It is possible that tube feeding is beneficial in population subgroups, as Mitchell et al.²¹ revealed patients with the lowest mortality are ones whose feeding tubes are placed only temporarily, for reversible conditions such as stroke with chance of neurological recovery.  Decisions about tube feeding should be individualized for these patients.  Research also needs to be directed at further identifying populations who will benefit.  For patients without chance of improvement of their underlying conditions, wording to patients and families should be chosen carefully.  Studies have found one of the biggest reasons families²⁴ and nursing home resident’s themselves²⁵ choose feeding tube placement is the belief it will prolong life.  It should be made clear to patients and families that this is likely not the case.

               

 

References