The role of infection in the morbidity and mortality of patients with head and neck cancer undergoing multimodality therapy

The Role of Infection in the Morbidity and Mortality of Patients With Head and Neck Cancer Undergoing Multimodality Therapy Maha Hussain, MD,* Julie A. Kish, MD,t Lawrence Crane, MD,$ Ahmed Uwayda, MD,§ Glen Cummings, PhD,* John F. Ensley, MD,? Efstathios Tapazoglou, MD," and Muhyi Al-Sarraf, MD 1 1

Cancer of the head and neck is a common cancer worldwide. The majority of patients present with locally advanced disease. Recently a great deal of improvement has been made in multimodality therapy of this disease, warranting more careful consideration of factors affecting quality of life, disease course, and treatment. Infection is clearly a factor. Analysis of 662 hospital admissions of 169 head and neck cancer patients was performed. A definite infection was documented in 86 febrile episodes, pneumonia contributed to 40'10, bacteremia to 13%, skin and soft tissue infection to 12%, and tracheobronchitis to 10%. Among the evaluated risk factors, foreign bodies, specifically intravenous (IV) cannulae and gastrostomy tubes, race, performance status, alcohol intake, and nutritional status were statistically significant variables that predicted for or were associated with infection. Infection contributed to 44% of the deaths. Cancer 67:716-721,1991.

ARCINOMA OF THE head and neck accounts for C about 5% of all new cancers in the US.' Greater than 60% of these patients present as clinically advanced Stages 111 and IV.' In addition to their cancers, this group of patients demonstrates the detrimental effects of heavy tobacco and alcohol abuse and their attendant medical disorders. These factors may compound the significant physical impairment produced by the tumor and result in a higher susceptibility to infection and malnutrition.

Infection is cited as a major cause of morbidity and mortality in cancer patients ~ v e r a l l . ~ , ~ Head and neck cancer patients are at particularly higher risk because of the disruption of anatomic and physiologic barriers. De-

From the Divisions of Hematology and Oncology and Infectious Dis- eases, Wayne State University, Harper-Grace Hospitals, Detroit, Mich- igan.

*Assistant Professor of Medicine, Division of Hematology and On- cology.

TAssociate Professor of Medicine, Division of Hematology and On- cology.

$Associate Professor of Medicine, Division of Infectious Diseases. §Former Fellow, Division of Infectious Diseases. ((Professor of Medicine, Division of Hematology and Oncology. Address for reprints: Maha Hussain, MD, Wayne State University/

Harper-Grace Hospitals, School of Medicine, Division of Hematology and Oncology, P.O. Box 02188, Detroit, MI 48201.

Accepted for publication July 2, 1990.

spite this, surprisingly little has been written concerning infections in this group of patients over the last decade.

In an autopsy review at M. D. Anderson Hospital of 94 patients with head and neck cancer who were treated from 1968 through 1970,43 deaths (46%) were related to in fe~ t ion .~ Since that report, the introduction of active chemotherapeutic agents such as cisplatin in combination with other agents plus radiation therapy and surgery has dramatically affected the care of these patients.' Significant advances have also occurred in general supportive care, i.e., antimicrobials and nutrition. These advances have made the multimodality treatment of this disease a model for other solid tumors.

The Head and Neck Cancer Service at Wayne State University (Detroit, MI) treats more than 200 newly di- agnosed and recurrent head and neck cancer patients per year in a multidisciplinary fashion. This analysis details the type and frequency of infection and the associated morbidity and mortality in head and neck cancer patients, and reviews the role of potential predisposing factors.

Materials and Methods

All admissions to the Head and Neck Cancer Service at Harper Hospital/Wayne State University were analyzed for the period of January 1, 1985 through December 3 1,

716

No. 3 INFECTION IN HEAD AND NECK CANCER Hztssain el a/. 717

1985. Patients who were not hospitalized are not included in this group. Charts identified by the cancer registry were each reviewed by a member ofthe Medical Oncology and Infectious Disease Divisions.

Inclusion criteria consisted of histologically proven cancer of the head and neck, and adequate documentation of infection.

The following definitions for various infections were used. ( 1) Pneumonia: radiographically documented pulmonary infiltrate persisting 48 hours or longer, associated with at least one compatible sign, i.e., fever (oral temper- ature 2 38"C), purulent sputum, high leukocyte count, or left shift, and one compatible symptom, i.e., chest pain, chills, and cough. Microbiologic etiology required either compatible gram stain and isolation of a pathogen from expectorated sputum and blood or pleural fluid, or compatible gram stain and recovery of the same pathogen from the sputum on two consecutive occasion^.^ All other pneumonias were categorized as unknown microbiology. (2) Skin and soft tissue infection: purulent discharge ap- pearing in an incision, with its microbiology based on the results of the gram stain or cultures. ( 3 ) Bacteremia: di- agnosed when at least one blood culture yielding at least one organism and the patient had consistent signs and symptoms of septicemia.

Infections were considered as contributing to death when blood cultures obtained during the last week of life in patients with clinical evidence of infection were posi-

To consider pneumonia as a cause of death, clinical and microbiologic evidence for extensive pulmonary infection were required.

The primary tumor sites were categorized as follows: oral cavity, larynx, oropharynx, paranasal sinuses, pharynx, salivary gland, and others to include unknown primary, and nasal melanoma. The TNM staging system was utilized with grouping into clinical stages to simplify the analytic process,' and Eastern Cooperative Oncology Group (ECOG) criteria for performance status were used.'

Predisposing Factors

Several variables were evaluated by chi-square tests or Fisher's exact test where appropriate to define their po-

TABLE l . Patient Characteristics

Characteristics No. of patients

Total 169 Sex

Men I08 Women 61

Black 80 White 89

Median age (yr) No. of admissions 662

Race

6 1 (range, 34-86)

TABLE 2. Admitting Diagnoses

Diagnosis No. of admissions (YO)

Chemotherapy 298 (45%) Surgery and/or endoscopy 168 (25%) Dehydration 30 (4.59'0) Malnutrition 23 (3.4%) Infection 18 (3%) Miscellaneous 125 (19%) Total 662

tential roles in leading to infection.' In addition, significant factors were evaluated by stepwise logistic regression analysis." Variables evaluated were as follows: age, sex, race, stage, tumor site, performance status, history of alcohol intake, concurrent cardiac disease, pulmonary disease, diabetes mellitus, modality of therapy (surgery, chemotherapy, radiation therapy), leukocyte count, presence of foreign bodies in the form of gastrointestinal (GI) access tubes (feeding tube, gastrostomy, or nasogastric tube), intravenous (IV) cannulae (Hickman catheter, subclavian, or peripheral IV cannulae), and tracheostomy.

We also evaluated nutritional status by utilizing the following assessment criteria which are based on weight change and serum albumin." Weight change reflects the adequacy of caloric intake. Two assessments were used to determine weight when applicable: (1) Weight as a per- centage of ideal body weight (ideal body weight was de- rived from established tables) = (current weight/ideal body weight) X 100. A weight that is 80% to 90% of ideal body weight is suggestive of mild calorie malnutrition, 70% to 80% is suggestive of moderate malnutrition, and less than 70% is severe malnutrition; (2) percent loss of body weight = (usual weight - current weight/usual weight) X 100, whenever premorbid weight was available. A 10% loss of body weight or more was a significant weight change. Serum albumin concentration of 2.8 g% to 3.5 g% is rep- resentative of mild visceral protein depletion, 2.1 g% to 2.7 g% of moderate depletion, and less than 2.1 g% is indicative of severe depletion."

For statistical analysis of risk factors, the most serious or life-threatening infectious episode was used since several patients had more than one infection.

Results

Tables 1 and 2 illustrate patient characteristics and admitting diagnoses, respectively. Table 3 lists the distribution of patients by site, stage of tumor, and performance status. One hundred three of 169 (60%) patients presented with advanced stage (111 & IV).

Infections One hundred two febrile episodes were documented in

67 patients (Table 4). Sixteen episodes were excluded from

718 CANCER February 1 199 I Vol. 61

TABLE 3. Distribution of Patients by Site, Stage, and Performance Status

No. of patients

Site Larynx Oral cavity Oropharynx Paranasal sinus Pharynx Salivary Other

Stage 1

I1 I11 IV Recurrent Unknown

0 1 2 3 4

Performance status (ECOG score)

41 34 44 13 25 4 8

1 1 16 34 69 23 16

1 62 56 34 16

ECOG: Eastern Cooperative Oncology Group.

the statistical analysis as follows: seven attributed to causes other than infections (atelectasis, serous otitis, superficial thrombophlebitis); three febrile neutropenic episodes that were treated with empiric antibiotics without an identi- fiable source, and six episodes unrelated to the tumor or its treatment, i.e., symptomatic and asymptomatic urinary tract infections and wound infection secondary to hernia repair.'*

A definite infection was documented in 86 febrile episodes; these are detailed in Table 4.

Table 5 lists the etiologic microorganisms for the pneumonias and for the bacteremias, respectively. There were 32 deaths, 14 (44%) of which were infection related. Pneumonia was the contributor to 43% of the infection- related deaths (six of 14) or 18% (six of 3 2 ) of the total deaths. Other infection-related deaths were as follows:

TABLE 4. Type, Distribution, and Incidence of Infection

No. of episodes Percent

Pneumonia Bacteremia Cellulitis and wound infection Tracheobronchitis Tumor infection Orocutaneous fistula Hickman exit site infection Osteomyelitis Clinical sepsis Peritonitis, neck abcess, enterocolitis, septic

Total phlebitis, and Candidu esophagitis

34 I I 1 1 9 4 4 3 3 2

5 86

40 13 13 10 4.6 4.6 3.5 3.5 2

5.8

TABLE 5. Microbiology of Pneumonias and Bacteremias

Microorganism No. of patients

Pneumonia Aeromonus hydrophila Anaerobes Hernophilus infuenzae Klehsiella oxytoca Klebsiellu pneumoniae Pseudomonas

aeruginosu Poly microbial Staphylococcus aureus Streptococcus pneumoniae Unknown

Staphylococcus uureus Streptococcus pyogenes Enterococcus Pseudomanus aeruginosa Pol ymicrobial

Bacteremia

1 I 2 1 4

2

1 2 1

13

5 2 1 1 2

three secondary to bacteremia, two secondary to urosepsis, two secondary to clinical sepsis, and one death secondary to Candidu peritonitis. Other causes of death were as follows: 14 deaths were related to terminal cancer, two deaths were of unclear etiology, and single deaths were attributed to massive pulmonary embolus and cerebrovascular ac- cident. Sixty-nine percent (22/32) of the deaths were in Stage 111 and IV patients; 45% (10/22) of the deaths in Stage I11 and IV patients were secondary to infection. Of the infection related deaths, 71% (10/14) were in Stage 111 and IV patients; in contrast, no infection-related deaths occurred in Stage I or I1 patients. One Stage I1 patient died of a massive pulmonary embolus.

Of the 18 patients who were admitted with a diagnosis of infection, 12 (67%) were Stage 111 and IV, three had GI tubes, three had IV cannula, and one patient had both an IV cannula and a GI tube on admission. Only six of 18 had a performance status ECOG score 111 or IV.

Risk Factors

Of the risk factors evaluated in Table 6, statistical significance was attributed to performance status, IV cannulae, and malnutrition ( P < 0.001): gastrostomy tubes (P < 0.01): alcohol consumption and race ( P < .05). No statistical significance was associated with tumor stage, site, or patient's sex, nor was there any relationship between infection and the specific treatment modality (surgery, radiation, or chemotherapy) or any concomitant medical condition or tracheostomy. There was no difference in the leukocyte count between the patients with infection and those without infection. Maximum, mini- mum, and median leukocyte count for the whole group was 17.8 X 103/pl, 0.8 X 103/pl, and 7.5 X 103/fil, respectively. Among those with bacteremia, only one pa-

No. 3 INFECTION IN HEAD AND NECK CANCER Hussain et d. 719

TABLE 6. Results of Risk Factors Analysis

Factor 1 Factor 2 Chi-square df P value

Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection Infection

Stage Site ECOG score Sex Race Chemotherapy Radiotherapy Surgery GI tubes Gastrostomy tube Nasogastric tube Feeding tube IV canula Tracheostomy Alcoholism Malnutrition Heart disease Diabetes Lung disease

6.88 5 2.81 6

42.00 4 1 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0.230 0.832 o.ooo* 0.120 0.036T 0.292 0.536 0.51 I 0.0074 0.0054 0.524 0.214 o.ooo* 0.150 0.041t o.ooo* 0.110 0.340 0.072

df: degrees of freedom; ECOG: Eastern Cooperative Oncology Group;

* P < 0.001. t P < 0.05. *P<O.OI. Fisher’s exact test was used for all 2 X 2 tables (one degree of freedom);

GI: gastrointestinal; IV: intravenous.

chi-square was used for all other tables.

tient was leukopenic with a total leukocyte count of 0.8

Foreign bodies have classically been associated with infection, therefore, the presence of foreign bodies was evaluated in detail. As anticipated, IV cannulae (central or Hickman) have a significant relation to infection ( P < 0.00 1). When analyzed by type, however, there was no difference in the statistical significance between the Hick- man catheter and the central venous catheter. Of the 11 patients with bacteremia, eight had IV cannula. Gastroin- testinal access tubes as a group also carried a significant relation to infection ( P < 0.0 1). However, when they were analyzed by type, gastrostomy versus nasogastric tubes versus feeding tubes, gastrostomy tubes emerged as the only significant variable ( P < 0.0 1).

The presence of malnutrition is also empirically related to infection. In this analysis, we objectively defined a patient’s nutritional status. The nutritional status was evaluated by utilizing two parameters: weight change and albumin level as defined earlier. For patients to be classified as malnourished, they had to fulfill both criteria. However, there was no difference between mean serum albumin of those who were infected when compared with the non- infected group: mean albumin was 3.080 g% k 1.167 for the former group and 3.012 g% -+ 1.627 for the latter group. Therefore, weight change was the only criteria from which nutritional assessment could be made retrospec- tively. Patients demonstrating malnutrition as defined by a weight change had a statistically significant higher risk

x 1 0 ~ / ~ 1 .

of infection which was 3.5 times greater than control. In this population malnutrition often coexists with ethanol (ETOH) abuse. A history of alcohol consumption did correlate significantly with infection ( P < 0.05). Nutri- tional status and alcohol consumption may manifest their influence through poor performance status, ( ie . , higher ECOG score) which correlated quite significantly with infection. Being black also correlated with a higher chance for infection.

Risk factors were evaluated by stepwise logistic regression (BMDPLR). Because of the colinearity between gastrostomy tubes and IV lines, only one of the two could be used in further analysis. With stepwise logistic regression analysis of all risk factors, only IV cannulae and performance status were predictive for infection (Table 7).

Since pneumonia was a major contributor to infections and infection-related mortality, it was analyzed separately against the same risk factors as was infection in general. Nevertheless, the same risk factors held the same statistical significance, whether pneumonia and the rest of the infections were analyzed as separate groups or were analyzed as one group.

Discussion

In a study from M. D. Anderson4 in 1970,46% of head and neck cancer deaths were related to infection. In a recent retrospective review by Crane13 evaluating 7.5 years of admissions of head and neck cancer patients at Wayne State University, 65 infections were found in 55 patients. Pneumonia contributed to almost 70% of the infections and was the major cause of infection-related mortality. However, case finding was accomplished by coded discharge diagnosis of infection. In our current review, 86 definite infections were documented in 67 patients over a 12-month period. This reflects our more rigorous case finding approach of reviewing every admission in a defined time period, and may reflect the improvement in the supportive care and multimodality cancer-specific therapy resulting potentially in prolonged survival in some subsets of this group of patients.14

As in Crane’s review and M. D. Anderson’s review?’3 pneumonia constitutes the majority of infections. In our review, it constituted 40% of total infections and resulted

TABLE 7. Results of Stepwise Logistic Regression

Factor I Factor 2 F df P value

Infection IV 12.4 1/136 o.ooo* Infection ECOG 4.7 1/136 0.0347

IV: intravenous; ECOG: Eastern Cooperative Oncology Group; F: ;

* P < 0.001. t P < 0.05.

df: degrees of freedom; F: F ratio.

720 CANCER February 1 199 1 Vol. 67

in 18% of deaths. Since aspiration-related atelectasis is common in this population, pneumonia poses a diagnostic dilemma. This dilemma of diagnosis is illustrated by studies of Andrews and colleague^,'^ where a correlation of the histopathologic and clinical diagnosis was made in patients dying of adult respiratory distress syndrome. Twenty-four patients were evaluated; pneumonia was present histologically in 58% of the study patients, 36% of whom had been thought to have only lung injury. Con- versely, among patients who had only diffuse lung injury histologically, 20% were thought to have pneumonia clinically. Therefore, 29% of the cases were misdiagnosed.

Head and neck cancer patients are predisposed to aspiration and thus pneumonia by many factors. There is anatomic barrier disruption by the tumor, surgical pro- cedures performed and presence of foreign bodies in the GI tract. Associated illnesses such as chronic lung disease resulting in poor mucocilliary clearance. l 3 Alcoholism, which in our study emerged as a significant factor associated with infection, leads to depressed cough reflex, altered leukocyte migration, and T-cell depre~sion.’~ The increased incidence of gram-negative colonization in al- coholics suggests a loss of normal adherence barriers. 16,17

Mucosal adherence of pathogenic bacteria is a prerequisite for events leading to pneumonia.

Radiation therapy of the salivary glands results in de- creased salivary flow and secretory IgA levels.18,” These several salivary components are thought to play a role in the defense against pneumonia. Chemotherapy can also result in mucositis and myelosuppression.

In our analysis, the actual therapeutic modality did not seem to alter the risk for pneumonia; however, this needs to be addressed in a prospective fashion.

We analyzed several specific risk factors in relation to infection. Certain foreign bodies, performance status, nutritional status, alcohol intake, and race emerged as significant; the latter raised the question whether other variables determined by one’s race are operational or is the difference purely related to socioeconomic status. This is not clear at this time. Performance status reflects the effects of several factors including the effects of the cancer, concurrent medical illness, and the general nutritional status. As expected, there is a strong correlation between performance status and infection.

The statistical significance of foreign bodies, particularly intravenous cannulae and gastrostomy feeding tubes, may reflect the potential for bacteremia with IV lines and pneumonia with gastrostomy tubes. It is interesting to note from our statistical analysis that the risk factors’ significance for pneumonia when analyzed as a separate group and for the whole group of patients with infection (with pneumonia included) did not change.

Whether the presence of feeding gastrostomy tubes and IV catheters selected poorer risk patients (inadequate nu-

trition, more advanced disease, elc.) is not clear. Certainly malnutrition is a factor in this group of patients resulting from local tumor effect, anorexia, chronic alcohol abuse, advanced age, and associated systemic disease. In our study, about one third of all the patients were assessed to be nutritionally deficient; malnutrition was a significant factor relating to higher infection risk. Marks2’ estimated that more than a third of all cancer victims die as a direct result of nutritional factors. In a prospective study by Brookes2’ of the nutritional status of 114 patients with untreated primary squamous cell cancer of the head and neck, 37.7% ofthe patients had nutritional deficits. Those who were adequately nourished had 57.5% survival at 2 years in comparison with 7.5% in the undernourished group, indicating the importance of nutritional status as an adverse prognostic variable. In addition to a good clinical and dietary history and physical examination, nutritional assessment should involve evaluation of the body’s fat stores and somatic proteins by using anthropometric and biochemical measurements. Minimally, it is suggested that this assessment should include current weight, recent weight change, determination of serum albumin, trans- fenin, total lymphocyte count, testing for delayed hyper- sensitivity reaction to common skin antigen, and muscle function evaluation. ’ However, in light of the retrospective nature of our analysis, such a detailed nutritional assessment was not performed. It is of serious concern that despite the improvement in our diagnostic and treatment abilities, almost 50% of these patients with Stage 111 and IV head and neck cancer died secondary to infection. This demands more prospective evaluation of the problem of infection in this group of patients.

Conclusion

In our study, 44% of the head and neck cancer patients’ deaths were infection related, with pneumonia as the major contributor. The actual incidence of infection in general and pneumonia in particular may be underestimated by virtue of the retrospective nature of this study. This analysis clearly emphasizes the need for more careful at- tention to diagnosis and documentation of infection to ensure more effective management. In addition, a better understanding of the effects of chemotherapy and radiation therapy on the local defense mechanisms in head and neck cancer patients is needed, i.e., effect on bacterial flora, mechanisms of adherence and colonization, fibro- nectin, salivary lysozymes, and proteases.

Several factors in this study correlated with the devel- opment of infections. These factors include the presence of IV cannulae, gastrostomy tube, race, performance status, alcohol intake, and nutritional status. Most predictive were IV cannulae and performance status. The design of future trials involving head and neck cancer patients

No. 3 INFECTION IN HEAD AND NECK CANCER * Hussain et d. 721

should control the impact of infection by addressing these factors.

REFERENCES

1. Al-Sarraf M. Head and neck cancer: Chemotherapy concepts. Semin

2. Ketchel SJ, Rodriguez V. Acute infections in cancer patients. Semin

3. Bodey GP. Infections in cancer patients. Cancer Treai Rev 1975;

4. Inagaki J, Rodriguez V, Bodey GP. Causes of death in cancer pa-

5. Tilltson JR, Lerner AM. Pneumonias caused by gram negative

6. Chang HY, Rodriguez V, Narbon G et al. Causes ofdeath in adults

7. American Joint Committee on Cancer. Manual for Staging of Can-

8. DeVita V Jr, Hellman S, Rosenberg S, eds. Cancer: Principles and

9. Dixon WS, Massey FJ. Introduction to Statistical Analysis. New

10. Dixon WJ, Brown MB. BMDP Biomedical Computer Programs.

1 1. Grant JP, Custer PB, Thurlow J. Current techniques of nutritional

Oncol 1988; 15(1):70-85.

Oncol 1978; 5:167-179.

2:89-128.

tients. Cancer 1974; 33:568-573.

bacilli. Medicine (Baltimore) 1966; 45:65-76.

with acute leukemia. Medicine (Baltimore) 1976; 55:259-268.

cer, ed. 2. Philadelphia: JB Lippincott, 1983; 25-54.

Practice of Oncology, vol. I , ed. 3. Philadelphia: JB Lippincott, 1989.

York: McGraw Hill, 1969; 243.

Berkeley: University of California Press, 1979; 517.1-517.13.

assessment. Surg Clin North Am 1981; 61(3):437-463.

12. Louria DB. Controversies in the management of infectious complications of neoplastic disease: Introduction and epidemiology. Am J

13. Crane L. Infections in patients with head and neck cancer. In: Jacobs JR, Cnssman JD, Valeriote FA, Al-Sarraf M, eds. Head and Neck Cancer: Scientific Perspectives in Management and Strategies for Cure. New York Elsevier, 1987; 145-158.

14. Rooney M, Kish J, Jacobs J et a/. Improved complete response rate and survival in advanced head and neck cancer after three course induction therapy with 120 hours 5-FU infusion and cis-platinum. Cancer

15. Andrews CP, Coalson JJ, Smith JD et al. Diagnosis of nosocomial bacterial pneumonia in acute, diffuse lung injury. Chest 1981; 80:254- 258.

16. Mackowiak PA, Martin RM, Jones RS. Pharyngeal colonization by gram-negative bacilli in aspiration-prone persons. drch Intern Med

17. Smith FE, Palmer DL. Alcoholism, infection and altered host defenses: A review of clinical and experimental observations. Journal of Chronic Disease 1976; 29~35-49.

18. Eneroth CM, Henrikson CO, Jakobsson PA. Effects of fractionated radiotherapy on salivary gland functions. Cancer 1972; 30:1147-1153.

19. Sonis ST. Oral complications of cancer therapy. In: DeVita VT, Hellman S, Rosenberg SA, eds. Cancer: Principles and Practice of On- cology, ed. 2. Philadelphia: JB Lippincott, 1985; 2014-2121.

20. Marks PN. Nutrition and cancer. Arch Intern Med 1977; 17:417. 21. Brooks GB. Nutritional status: A prognostic indicator in head

Med 1984; 76~414-420.

1985; 55:1123-1128.

1978; 138: 1224- 1227.

and neck cancer. Head and Neck Surgery 1985; 93:69-74.

Documents

The role of infection in the morbidity and mortality of patients with head and neck cancer undergoing multimodality therapy