Variables Associated With Feeding Tube Placement in Head and Neck Cancer

Objective  To identify clinical factors associated with enteral feeding tube placement in a head and neck cancer population.

Design  A self-administered survey was given to patients being treated for head and neck cancer while they were waiting to be seen in 1 of 4 otolaryngology clinics. The post hoc analysis presented here combines survey and chart review data to determine clinical and demographic variables associated with feeding tube placement.

Setting  Four otolaryngology clinics.

Patients  Otolaryngology clinic patients being treated for head and neck cancer.

Main Outcome Measure  Enteral feeding tube placement.

Results  Of the 724 patients eligible for this study, 14% (n = 98) required enteral feeding tube placement. Multivariate analysis found the following variables to be independently associated with feeding tube placement: oropharynx/hypopharynx tumor site (odds ratio [OR], 2.4; P = .01), tumor stage III/IV (OR, 2.1; P = .03), flap reconstruction (OR, 2.2; P = .004), current tracheotomy (OR, 8.0; P<.001), chemotherapy (OR, 2.6; P<.001), and increased age (OR, 1.3; P = .02). In addition, there was a curvilinear relationship between time since treatment and feeding tube placement, with about 30% having a feeding tube at 1 month posttreatment, tapering down during the first 3 years to about 8% and leveling off thereafter.

Conclusions  Identification of factors associated with an increased risk of feeding tube placement may allow physicians to better counsel patients regarding the possibility of feeding tube placement during treatment. Since feeding tube placement has been linked to decreased quality of life in head and neck cancer, such counseling is an integral part of the clinical management of these patients.

Each year, 500 000 new cases of head and neck cancer are diagnosed worldwide, with over 40 000 new cases in the United States alone. In the United States, 12 000 deaths per year are caused by head and neck cancer.¹ In addition to mortality, head and neck cancer causes significant morbidity related to problems with poor nutrition, decreased communication, pain, and decreased general functional capability.^2,3

Patients with advanced head and neck cancer are often nutritionally depleted at initial presentation. An increasing number of patients with head and neck cancer require enteral feeding supplementation during the course of their treatment because of increasingly aggressive multimodal therapy. Common indications for feeding tube placement in these patients include tumor-related factors (malnutrition or dysphagia) as well as treatment-related factors (dysphagia, odynophagia from surgery or irradiation, repeated aspiration, and pharyngocutaneous fistula).^4,5

Complications arising from gastrostomy tube placement contribute to morbidity, including infection, blockage of the tube, and aspiration pneumonitis.^6,7 More rare but serious complications include gastrointestinal bleeding, peritonitis, and metastasis of the primary tumor to the gastrostomy site.^6,8 Initiation of enteral tube feeding is also a significant cause of head and neck cancer–associated decreases in quality of life,^9,10 as documented by a previous study by our group.¹¹ To aid in prospective identification of patients at risk for prolonged enteral feeding, this study was designed to determine which clinical and demographic variables are associated with enteral feeding tube placement.

This multicenter cross-sectional study was part of a larger study investigating quality of life in patients with head and neck cancer. A research assistant distributed a self-administered questionnaire on quality of life to a convenience sample of patients with head and neck cancer while they were waiting to be seen in the otolaryngology clinic. Research assistants helped patients complete the questionnaire as needed. Clinical data were abstracted from medical records. Institutional review board approval was obtained at each site prior to the study.

Subjects were recruited from 4 medical centers including 3 Veterans Affairs (VA) hospitals, in Ann Arbor, Mich, Gainesville, Fla, and Dallas, Tex, and the University of Michigan Hospital from 2000 to 2002. Of the 1561 patients initially approached, 62% (n = 973) agreed to participate and met all eligibility requirements. The inclusion criterion was presence of cancer of the head and neck. Excluded were patients who, from the time of diagnosis and any time thereafter, were (1) pregnant; (2) younger than 18 years; (3) non-English speaking; or (4) found to have a tumor not arising from the upper aerodigestive tract (such as thyroid, parotid, or skin cancer). An additional 249 subjects had completed the survey less than 1 month after most recent therapy and were excluded from these analyses, leaving a sample size of 724 subjects. A 1-month threshold was chosen because it was thought that within 1 month after treatment, both patients and clinicians would regard enteral tube feeding as a short-term and nonextreme intervention, whereas the physical and psychosocial impact might be perceived as much larger when enteral feeding tube placement is required over an extended period after treatment.

The independent variables of interest included clinical and demographic variables. The dependent variable was enteral feeding tube placement at the time of the survey. Both nasogastric tubes and gastrostomy tubes were coded as feeding tubes. Some of the feeding tubes were placed prior to initiation of therapy and were coded according to whether they were present or absent at the time of survey. Thus, feeding tubes placed prophylactically, for previous tumor-unrelated treatment, and during the course of head and neck cancer treatment were all coded simply as present at the time of survey.

Clinical variables consisted of time since treatment, tumor stage, tumor site, current tracheotomy (excluding laryngectomy stoma), surgery (total laryngectomy, primary site excluding laryngectomy, any flap reconstruction, and/or neck dissection), and adjuvant therapy (chemotherapy and/or radiation). Tumor sites were categorized into 3 groups: oropharynx/hypopharynx, oral cavity/other, and larynx. These categories were chosen because oropharynx and hypopharynx sites behaved similarly in quality of life data from our previous research, as did oral cavity and “other” sites.^2,3 The “other” sites included nasal cavity, nasopharynx, sinus tumors, and unknown primary sites. Tumor stage was classified as stage 0/I/II vs stage III/IV. Flap reconstructions included both free and regional flaps. Nonflap reconstructions included primary closures, skin grafts, local flaps, and healing by secondary intention. Since patients were surveyed at different points in treatment, time since most recent treatment of head and neck cancer was measured in months since treatment.

Demographic control variables consisted of age, sex, race, marital status, education, and hospital site. Since there were so few African Americans and other race respondents compared with whites, race was classified into white and nonwhite for analysis. Similarly, marital status was classified as married vs not married (separated, widowed, divorced, or never married). Education was classified as high school or less, some college, or college or more. Hospital site was classified as University of Michigan or 1 of the 3 VA hospitals (Ann Arbor, Dallas, or Gainesville).

Descriptive statistics (means and frequencies) were computed on all demographic, clinical, and treatment variables. Bivariate analysis was conducted between the independent variables collected and the dependent variable, feeding tube placement. χ² Tests were used for categorical variables, t tests for interval level variables. Logistic regression was used to conduct multivariate analyses to determine the predictors of feeding tube placement as well as to determine the probability of having a current feeding tube given the number of predictor variables present.

Owing to power limitations, not all of the variables that were statistically significant in the bivariate analysis could be included in the multivariate analysis. Therefore, variables that showed the strongest statistical associations with feeding tube placement in the bivariate analysis were chosen, including tumor site, tumor stage, flap reconstruction, current tracheotomy, irradiation, and chemotherapy. While age was not significant in the bivariate analysis, it is known to be associated with swallowing and was hence included in the analysis. Since a curvilinear relationship was observed between time since treatment and likelihood of feeding tube placement, a time-squared variable was included in the multivariate analysis. While sex and race have been significant predictors in other head and neck cancer studies, they were omitted from the multivariate analysis because they were highly correlated with hospital site (VA vs university hospital) and were not significant in the bivariate analysis.

The data were entered into a Microsoft Access database and analyzed using SAS software (SAS Institute Inc, Cary, NC). Since all the respondents did not answer all of the questions, sample size varied for different analyses. For all tests, a 2-tailed P value of less than .05 was considered statistically significant.

The probability of having an enteral feeding tube if 1 or more of the significant clinical (but not demographic) risk factors were present was calculated using a logistic regression. The 5 clinical variables included in this analysis were oropharyngeal and/or hypopharyngeal site, stage III or IV cancer, any flap reconstruction, current tracheotomy, and chemotherapy. Count variables were created based on the sum of predictor variables present for each subject (0, 1, 2, 3, 4, or 5). The count variables were then used in a logistic regression to generate the probability of having a feeding tube given the number of predictor variables. It should be noted that only 1 subject had all 5 chosen clinical variables. This did not provide enough data to accurately assess the probability if all 5 variables were present; therefore, no estimate of probability is given for 5 variables. Age was not included in this model because it was considered a demographic rather than a clinical variable. In addition, an arbitrary cutoff between younger and older was difficult to justify.

Descriptive statistics are reported in Table 1 and Table 2. Fourteen percent (n = 98) of patients in the study had an enteral feeding tube in place at the time of the survey. Tumor sites were evenly distributed among hypopharynx/oropharynx, oral cavity/other, and larynx. Almost two thirds of patients (64%; n = 449) were classified as stage III/IV. Almost three quarters of the study population (71%; n = 516) underwent some type of head and neck surgery (excluding biopsies) as part of their overall treatment. About one fifth of all study patients (21%; n = 153) received some form of surgical flap reconstruction (either regional or free). A small percentage of patients had a tracheotomy at the time of survey (4%; n = 28). About three quarters of patients (78%; n = 567) underwent radiation therapy. Of the patients who received chemotherapy (25%; n = 178), most also received irradiation (n = 168) (Table 1).

The mean age of the patients surveyed was 61 years at the time of survey. The mean number of months since most recent treatment was 35 months. Most of the patients were male (84%; n = 611) and white (89%, n = 634). Slightly less than two thirds of study patients were married (61%; n = 440). Most of the study patients had only a high school degree or less education (52%; n = 374), with a smaller percentage attending some college (32%; n = 227) or obtaining a 4-year college degree or more (16%; n = 116). Just over half were treated at the University of Michigan (56%; n = 409), the rest at a VA hospital.

Clinical variables associated with statistically significant increases in rate of enteral feeding tube placement were tumor site of oropharynx/hypopharynx, tumor stage of III/IV, flap reconstruction, tracheotomy, irradiation, and chemotherapy (Table 3 and Table 4). Age did not show a significant association with feeding tube placement in the bivariate analysis. There was a curvilinear relationship between time since most recent treatment and feeding tube placement. Although 14% of the entire population required a feeding tube, the frequency was greatest 1 to 3 months after treatment and then declined dramatically during the first 2 years after treatment, leveling off at about 8% after that time (Figure). Sex, race, marital status, educational level, and hospital site did not have a statistically significant impact on feeding tube placement.

While the cancer sites of oropharynx/hypopharynx and oral cavity/other compared with larynx were both significant predictors of feeding tube placement in the bivariate analysis, only oropharynx/hypopharynx remained significant in the multivariate analysis (odds ratio [OR], 2.4; P = .01) (Table 5). Stage III/IV disease continued to show a significant association with feeding tube placement (OR, 2.1; P = .03). Both flap reconstruction (OR, 2.2; P = .004) and tracheotomy (OR, 8.0; P<.001) remained significant in the multivariate analysis, with current tracheotomy being the strongest predictor of feeding tube placement of all the independent variables. Irradiation was no longer significant in the multivariate analysis (OR, 1.9; P = .12); however, chemotherapy did remain significantly associated with feeding tube placement (OR, 2.6; P<.001). While increased age was not a significant predictor of feeding tube placement in the bivariate analysis, in the multivariate analysis each decade of age was associated with a 30% greater odds of enteral feeding tube dependence (OR, 1.3; P = .02). Time and time-squared since most recent treatment showed statistically significant associations with the likelihood of feeding tube placement (OR, 0.98; P<.001 and OR, 4.6; P<.001, respectively). While the odds ratio of 1.9 for irradiation was not significant, smaller odds ratios for age and time since treatment were significant because each of these was tested against its own standard error, which was influenced by the scaling of the variables and by correlations with other predictors.

When the significant clinical (nondemographic) predictors were used in the logistic regression analysis to predict the probability of having an enteral feeding tube, the probability ranged from 0.6% if none of the predictors were present to 59% if 4 of the clinical predictors were present (Table 6). The 5 clinical predictors were oropharyngeal/hypopharyngeal site, stage III/IV cancer, any flap reconstruction, current tracheotomy, and chemotherapy.

In this head and neck cancer patient population, statistically higher odds of having enteral feeding tubes at least 1 month after completion of therapy were found in patients with 1 or more of the following variables: oropharyngeal/hypopharyngeal site, advanced stage, flap reconstruction, current tracheotomy, chemotherapy, increased age, and decreased time since treatment. While at least 2 small studies have attempted to identify clinical variables that predict feeding tube placement in patients with head and neck cancer,^12,13 to our knowledge this study is the first to examine data from a very large cross section of patients that allowed analysis of a greater number of independent variables.

Multivariate analysis revealed that patients with tumors of the oropharynx/hypopharynx had roughly twice the odds of requiring feeding tube placement when compared with patients with laryngeal tumors. These results are consistent with prior studies demonstrating that the site of the primary tumor has a predictable impact on the severity of dysphagia experienced by patients with head and neck cancer. Tumors in the posterior oral cavity, including the base of the tongue, soft palate, retromolar trigone, and tonsillar fossa usually cause more severe dysphagia on surgical excision owing to the critical role that the tongue base and posterior pharyngeal wall play in swallowing.^14,15 This results from the relative contributions of oral and pharyngeal structures to the complex act of swallowing, the extent of resection required at a given site, and the loss of sensation that accompanies the interruption of nerve function with surgery.

Advanced-stage cancers are more likely to distort and debilitate the mucosa, muscles, nerves, and other tissues that are critical to the complex physiology of swallowing and thereby be associated with greater risk of pretreatment or subsequent dysphagia. In addition, patients with advanced-stage cancers usually require more aggressive therapy, particularly larger doses of radiation to larger fields, larger resections of tissues for those treated with primary site surgery, more complex flap reconstructions, or more rounds of chemotherapy.

Flap reconstructions were found to have an independent association with feeding tubes after controlling for other clinical and demographic variables. Tissue flaps can directly contribute to dysphagia in a number of ways. They may obstruct bolus passage if they are bulky, decrease the amount of propulsive force available, or interfere with the normal sensation needed to guide the bolus down the oropharynx.¹⁶ In addition, the surgical defect created is typically greater in patients who require a flap for closure compared with those who have primary site surgical procedure without a flap. Unfortunately, there were not enough patients who received free flaps or regional flaps to include these groups separately in the multivariate analysis. Thus, we could not determine which kind of flap reconstruction, if any, might be associated with an increased incidence of feeding tube placement.

Several mechanisms may explain the strong association between tracheotomy and feeding tube placement. Increased aspiration risk is an indication for both tracheotomy and feeding tube placement, meaning that if a patient requires one procedure, he or she is very likely to require the other procedure as well. In addition, a large body of literature exists describing the negative effects of tracheotomy on swallowing function.^17,18 Since our survey data simply recorded the presence or absence of a tracheotomy at the time of survey, we are unable to conclude if both short-term and long-term tracheotomies are associated with feeding tube placement.

Patients who underwent irradiation appeared more likely to receive feeding tubes in the bivariate analysis, but this association did not remain significant in the multivariate analysis. This was an unexpected result because it is widely assumed that there is a strong association between radiation therapy and dysphagia in patients with head and neck cancer.¹⁶ This is possibly owing to the exclusion of patients who were less than 1 month out from their most recent treatment, who might have had temporary feeding tubes related to early irradiation adverse effects, whereas those further out from treatment might no longer have had feeding tubes at the time they were surveyed. It is possible that a larger sample size might have detected an association between irradiation and enteral feeding tube placement, but it probably would not be as strong a predictor as the other variables identified in this study.

Patients undergoing chemotherapy had markedly increased odds of receiving a feeding tube compared with patients who did not undergo chemotherapy. However, it should be noted that most chemotherapy protocols use chemotherapy in conjunction with irradiation. Thus, the number of patients in our study receiving chemotherapy alone was very small (<1%), and the 2.6 OR observed really represents the odds associated with combined chemoradiation treatment (after controlling for the other factors in the analysis, including irradiation).

The rate of feeding tube placement in patients undergoing chemoradiation therapy for head and neck cancer has not been carefully documented. In the present study, the observed rate of feeding tube placement in patients receiving chemoradiation was 28.0%. Published rates in patients with head and neck cancer undergoing radiotherapy or chemoradiotherapy range widely from 13% to 85%.^19-21 The incidence of severe (grade 3 or 4) mucositis is increased in patients receiving chemoradiation compared with those receiving radiotherapy alone in several randomized trials,²² making it probable that higher rates of mucositis and other soft tissue damage noted with chemoradiation therapy are causally related to the higher frequency of enteral feeding tube use in these patients. In fact, our clinical experience indicates that patients with laryngeal or hypopharyngeal tumors develop severe mucositis, stricture, and dysphagia after chemoradiation treatment, and therefore enteral feeding tubes are placed prophylactically in many patients undergoing chemoirradiation. Therefore, the association we have noted between chemoradiation therapy and enteral feeding tubes may be due in part to feeding tubes being placed “in anticipation” of treatment sequelae. Patients who undergo pretreatment, prophylactic enteral feeding tube placement often have the feeding tubes in place for more than 1 month after completion of therapy and so might have been included in this study. This could have skewed the odds calculation to some degree.

Each decade of age was associated with a 30% greater odds of feeding tube placement. This is likely related to an increasing number of medical comorbidities in each successive decade of life as well as the neural, muscular, and structural changes in the pharynx and larynx that occur with aging.

As expected, patients who more recently underwent treatment had a higher rate of feeding tube placement. The percentage of patients with feeding tubes decreased from 30% to about 8% in the first 3 years and leveled off thereafter (Figure). Much of this might be explained by the routine placement of feeding tubes as a part of organ preservation protocols, as noted. Survivorship may also play a role, whereby the patients living the longest had less severe disease, required less aggressive therapies, and had better outcomes from treatment, while patients who were in worse clinical condition may not have survived to be surveyed after completion of their treatment. Also, patients farther out from completion of cancer treatment may be better able to adapt to or minimize their dysphagia with swallowing therapy.

Based on the clinical risk factors associated with use of an enteral feeding tube in this study, a probability model for assessing the chance of having a feeding tube inserted was generated (Table 6). Each clinical risk factor was associated with an increased probability of enteral feeding tube dependence in this data set, such that patients with no clinical risk factors were unlikely to have an enteral feeding tube (probability, 0.6%), while those with 1, 2, 3, and 4 clinical risk factors had a much higher probability of requiring a feeding tube (probabilities, 10%, 18%, 26%, and 59%, respectively). This probability model does not take into account patient age or time since treatment because these variables were considered demographic rather than clinical variables and are continuous variables rather than dichotomous (absent/present). Therefore, the model in Table 6 may underestimate the risk of feeding tube placement in elderly patients and overestimate the risk in patients who are farther out from treatment. Nevertheless, we believe that this probability model is conceptually easier to grasp than the odds ratios generated from multivariate analyses. Patients and clinical caregivers can more easily generate a rough approximation of the probability of having an enteral tube placed at a time at least 1 month after therapy using this model. As such, we believe that this may be a valuable tool for clinical teaching, patient education, and outcomes research. While this clinical prediction tool may have utility in predicting feeding tube placements, ideally it should be validated in a prospective study of patients with head and neck cancer to confirm its clinical validity prior to using the model as a predictive tool for an individual patient.

This cross-sectional study uses data collected from patients at 1 point in time; however, this point in time varied from patient to patient with respect to initiation of treatment. While time was included in our analysis, no actual longitudinal follow-up of an a priori population of patients was conducted. Our results are affected by factors that influence whether a patient comes to the clinic, which include time since diagnosis and the severity of their ongoing clinical problems. While we think that there may be different variables associated with short-term and long-term enteral feeding tube placement, the sample size was not large enough to make this comparison.

Because our survey data did not distinguish between prophylactically placed feeding tubes and those placed subsequent to initiation of antitumor therapy, the statistical associations found in this study may reflect at least in part which variables influence clinicians in their assessment of which patients will need a feeding tube. Thus, the clinical variables we have found to be associated with feeding tube placement are not necessarily those that lead to clinically significant dysphagia, but these variables are associated with feeding tube placement in standard clinical management of these patients. Nonetheless, clinicians will continue to place enteral feeding tubes based on their best clinical judgment, and these results are a useful description of current clinical practice.

Despite our efforts to include minority patients by collecting data from 3 VA hospitals, there were still small numbers of minorities (11%) studied. Since a little less than half of the patients were from VA hospitals, whose patient base is largely composed of men, there were fewer women in the study (16%) than are in the general population of patients with head and neck cancer (24%). While we were able to control for time since most recent treatment, all of the patients were not surveyed at the same point in their treatment. Also, the results of this study are only generalizable to patients seen in otolaryngology clinics and not those seen in other settings and practices. Since feeding tube placement is now becoming a standard part of chemoradiation protocols, assessing patients at known times after the completion of their treatment may have given a more accurate picture of the rate of feeding tube dependence after the short-term treatment–associated toxic effects have resolved.

The number of variables included in our analysis was statistically limited by the sample size, even with the considerable number of patients recruited. Therefore, a number of important clinical distinctions, such as the type of radiotherapy and chemoradiation regimens, were not included in our initial questionnaire. It is certainly likely that increasing irradiation dose would have a positive correlation with the need for feeding tube placement. In addition, there are differences between the amount of mucositis caused by sequential and concomitant chemoradiation treatments. The relationship of these different modalities to the incidence of feeding tube placement would be of interest for further study.

In conclusion, in patients with head and neck cancer, statistically higher odds of having enteral feeding tubes in place more than 1 month after completion of therapy was found when 1 or more of 5 clinical factors (oropharyngeal/hypopharyngeal site, advanced stage, flap reconstruction, tracheotomy, and/or chemotherapy) and 2 demographic variables (age and time since most recent treatment) were present. Patients with greater numbers of clinical risk factors had an incrementally greater probability of needing enteral feeding tubes. Since previous work has demonstrated that enteral feeding tube dependence is a strong predictor of large decrements in health-related quality of life, these current findings may provide clinicians and patients with some insight into which patients may be at risk for such feeding tube–associated decrements in quality of life.

Correspondence: Sonia A. Duffy, PhD, RN, Veterans Affairs Ann Arbor Healthcare System, Health Services Research and Development (11H), PO Box 130170, Ann Arbor, MI 48113-0170 (sonia.duffy@med.va.gov).

Submitted for Publication: March 24, 2005; final revision received November 20, 2005; accepted January 27, 2006.

Financial Disclosure: None.

Funding/Support: This research was supported by grant IIR-98500 from the Department of Veterans Affairs, Washington, DC; by GlaxoSmithKline, Research Triangle Park, NC, through the Managed Care Forum; and by the National Institutes of Health, Bethesda, Md, through the University of Michigan's Head and Neck SPORE (Specialized Program of Research Excellence) grant 1 P50 CA 97248.

Previous Presentation: This research was presented at the Sixth International Conference on Head and Neck Cancer; August 8, 2004; Washington, DC.

Acknowledgment: We thank the otolaryngology clinic personnel at the VA hospitals in Ann Arbor, Gainesville, and Dallas and at the University of Michigan Medical Center for their cooperation in the study. We also thank Cynthia Wampler, BS, Lynn Gregory, MSN, Sharon Stephens, RN, and Carol Bishop, MSN, for their participation in data collection and management. Finally, we thank the patients who participated in this study for their time and cooperation.