The burdens of cancer therapy

Clinical and economic outcomes of chemotherapy-induced mucositis


  • Linda S. Elting Dr.P.H.,

    Corresponding author
    1. Section of Health Services Research, Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston Texas
    • Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Boulevard, Unit 196, Houston, TX 77030-4009
    Search for more papers by this author
    • Fax: (713) 792-7990

    • Dr. Elting has received honoraria from Endo Pharmaceuticals.

  • Catherine Cooksley Dr.P.H.,

    1. Section of Health Services Research, Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston Texas
    Search for more papers by this author
  • Mark Chambers M.D.,

    1. Department of Head and Neck Oncology, The University of Texas M. D. Anderson Cancer Center, Houston Texas
    Search for more papers by this author
  • Scott B. Cantor Ph.D.,

    1. Section of Health Services Research, Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston Texas
    Search for more papers by this author
  • Ellen Manzullo M.D.,

    1. Department of General Internal Medicine, Ambulatory Treatment, and Emergency Care, The University of Texas M. D. Anderson Cancer Center, Houston Texas
    Search for more papers by this author
  • Edward B. Rubenstein M.D.

    1. Section of Medical Supportive Care, Department of Palliative Care and Rehabilitation Medicine, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author



Mucositis is a common but poorly studied problem among patients with solid tumors. The authors examined the clinical and economic outcomes of oral and gastrointestinal (GI) mucositis among patients receiving myelosuppressive chemotherapy.


A retrospective, random sample of 599 patients who developed chemotherapy-induced myelosuppression was followed for development of oral or GI mucositis and for development of subsequent episodes of bleeding or infection. Multilevel regression models of the risk of bleeding and infection were fit with chemotherapy cycles nested within patients.


Mucositis developed during 37% of 1236 cycles of chemotherapy. Episodes of bleeding were significantly more common during cycles with GI mucositis than during cycles without GI mucositis (13% vs. 8%; P = 0.04). Episodes of infection were significantly more common during cycles with mucositis (especially GI mucositis) than during cycles without mucositis (73% vs. 36%; P < 0.0001). The mean durations of hospitalization were 4 days, 6 days, and 12 days during cycles with no mucositis, oral mucositis, and GI mucositis, respectively. After accounting for the depth and duration of myelosuppression and for other predictive factors, GI mucositis was associated with both bleeding (odds ratio [OR], 2.0; P = 0.01) and infection (OR, 2.24; P < 0.0001), whereas oral mucositis was associated with infection only (OR, 2.4; P < 0.0001).


Mucositis was clinically and economically significant among patients with solid tumors who were receiving myelosuppressive chemotherapy. New preventive and therapeutic agents are needed. Cancer 2003;98:1531–9. © 2003 American Cancer Society.

DOI 10.1002/cncr.11671

Mucositis is a common cause of morbidity during chemotherapy. The incidence of National Cancer Institute (NCI) Grade 3–4 oral and gastrointestinal (GI) mucositis derived from clinical trials of standard-dose chemotherapy is estimated between 5% and 15%. Chemotherapy with 5-fluorouracil (5-FU) or irinotecan is associated with rates of oral or GI mucositis exceeding 15%.1 Among patients receiving high-dose regimens or concomitant radiotherapy to the head, neck, thorax, chest wall, abdomen, or pelvis, rates may exceed 40% and significantly affect quality of life.1–4 However, rates of mucositis after most chemotherapy regimens administered to patients with solid tumors probably are far lower. Studies of incidence across regimens are lacking.

Although the incidence of mucositis among patients with solid tumors may be modest, its outcomes are probably clinically significant. It has been hypothesized that mucositis leads to infection and bleeding through interruption of the mucosa. An association between mucositis and infection has been demonstrated for some populations (primarily bone marrow transplantation recipients)5–7 and for some organisms (particularly Streptococci).8, 9 The relation between mucositis and bleeding has not been studied as much. Neither infection nor bleeding has been studied in a large sample of patients with solid tumors who were undergoing chemotherapy.

We hypothesized that if oral and GI mucositis increase the risk of infection and bleeding, then this effect would be most pronounced among patients with solid tumors who have developed thrombocytopenia and neutropenia. We further hypothesized that increases in these serious clinical events also would increase resource utilization. Therefore, we conducted a study of the outcomes and resource utilization associated with oral and GI mucositis among myelosuppressed patients with solid tumors or lymphoma.


A retrospective cohort consisting of a random sample of 635 adults with solid tumors or lymphomas was constructed from an existing database of patients with thrombocytopenia and neutropenia that had been used in the past for studies of bleeding.10, 11 Because our objective was to study bleeding and infection outcomes, this data set of patients with platelet counts that had fallen below 50,000 per μL and with absolute neutrophil counts that had fallen below 1000 per μL provided an ideal picture of outcomes in a high-risk population of patients with solid tumors. It permitted comparisons of the risk of infection and bleeding among patients with mucositis and similarly myelosuppressed patients without mucositis. However, it is important to note that the results of this study apply only to patients with solid tumors who have chemotherapy-induced myelosuppression.

Patients with leukemia, bone marrow transplantation or stem cell recipients, and patients younger than age 16 years were ineligible for the study. To avoid bias in ascertainment of outcomes and resource utilization, the sample was limited to patients who received all of their care at our center.

From the original 635-patient cohort, the medical records of 599 patients were available for review. Among these patients, chemotherapy cycles that were initiated between January 1, 1994 and December 31, 1995 during which myelosuppression occurred were included in the study. This resulted in a total of 1236 chemotherapy cycles. Patients were followed through December 31, 1996 for development of bleeding and infection and through December 31, 1999 for survival to correspond with previous data available in the database. Multiple cycles from a single patient were included to estimate the magnitude of the mucositis burden and mucositis-related infection and bleeding among patients with malignancies who were at risk of developing events and utilizing resources during every cycle of chemotherapy. The regression models were adjusted (see Statistical Considerations, below) for possible bias introduced by the correlation among multiple cycles within one patient.

One-half of all patients (48%) were males, and 19% were older than age 65 years. Patients with lymphoma, breast carcinoma, and sarcoma predominated, as would be expected from a population of solid tumor patients with chemotherapy-induced myelosuppression (Fig. 1). Twenty-five percent of patients had local disease, and 4% were receiving adjuvant chemotherapy in the absence of evidence of disease. Although the majority of patients (71%) had metastatic disease, 63% had a good performance status (Zubrod 0 or 1). Whereas 317 patients (53%) contributed only a single cycle of chemotherapy to the study, 282 patients contributed 2–16 cycles each. Only 10% of patients were chemotherapy naïve; 35% of patients had experienced oral mucositis, and 6% of patients had experienced GI mucositis during a previous cycle.

Figure 1.

Distribution of cancer diagnoses among the 599 patients. GI: gastrointestinal; GU: genitourinary; GYN: gynecologic.

Data Sources and Collection

All medical records of eligible patients were reviewed, including the paper medical record, electronic databases containing all hospital and clinic visits, tumor registry and survival data, diagnostic test results, blood products transfused, and pharmaceutical agents prescribed, as well as electronic records of patients enrolled on clinical research protocols. A 10-item set of variables (unknown to the abstractors) was collected both manually and electronically for estimation of the frequency of errors with manual transcription (< 4%).

Variables and Definitions

The primary grouping variable was the presence of oral or GI mucositis, as defined and characterized by the NCI Common Toxicity Criteria for mucositis not due to radiation.12 NCI criteria for esophagitis, gastritis, colitis, and typhlitis were aggregated and labeled GI mucositis. NCI criteria for stomatitis/pharyngitis were labeled oral mucositis. The duration of mucositis was measured using the last-value-carried-forward method.

The primary outcome measures, episodes of bleeding and episodes of infection, were defined as follows: Bleeding was characterized as either minor (World Health Organization [WHO] Grade 1 or 2, including petechiae, ecchymoses, superficial bleeding of gums, microscopic hematuria, blood-tinged sputum, mild epistaxis, and vaginal bleeding not requiring red cell transfusion) or major (WHO Grade 3 or 4, including epistaxis, vaginal bleeding or major organ hemorrhage requiring red cell transfusion, or fatal hemorrhage). Infection was defined according to published criteria for reporting clinical trials of antibiotic therapy13–17 and was categorized as either documented infection or fever of unknown origin. Documented infections were further categorized as either clinically or microbiologically documented; and microbiologically documented infections were categorized as gram-positive, gram-negative, fungal, viral, or polymicrobial. All bleeding and infection outcomes during chemotherapy cycles with mucositis occurred after the development of mucositis.

Secondary outcomes that were measured for descriptive purposes included infection-related and bleeding-related deaths, delays and dose reductions in the next planned cycle of chemotherapy, fatigue, oral or GI pain, and weight loss. Cause of death among cancer patients often is multifactorial and is difficult to attribute to one cause when the determination must be based on retrospective review of the medical record. Death during episodes of bleeding or infection was defined as infection-related and bleeding-related death. All other deaths were considered due to other causes. Chemotherapy delay was defined as a delay > 7 days in the next planned cycle of chemotherapy. Dose reduction was defined as a decrease ≥ 20% or discontinuation of the planned dose of any antineoplastic agent during the next cycle. These events were not mutually exclusive: Cycles with mucositis, bleeding, or infection often were accompanied by chemotherapy delay or dose reduction. Fatigue and pain were defined and categorized according to the NCI Common Toxicity Criteria. Weight loss was measured in kilograms from Day 1 of the chemotherapy cycle until the first of three possible events: 1) Day 1 of the next cycle, 2) recovery of the platelet count if no further therapy was received, or 3) death.

Potentially confounding variables were defined as follows: thrombocytopenia was defined as platelet counts < 50,000 per μL and was categorized by the nadir. Neutropenia was defined as an absolute neutrophil count < 1000 per μL and was categorized as < 100 per μL, 100–1000 per μL, and > 1000 per μL. Durations of thrombocytopenia and neutropenia were computed using the last-value-carried-forward method. The extent of disease dissemination at the initiation of each cycle was characterized as no evidence of disease, local disease, or disseminated disease. Performance status was measured on Day 1 of each cycle using the Zubrod score,18 with poor performance defined as a performance status < 2. Agents shown in our previous research to affect bleeding or platelet function also were measured, including tricyclic antidepressants and phenothiazines, heparin, penicillins or cephalosporins, and antihistamines.10, 11

Resource utilization was measured by collecting the number of emergency department visits and hospital days during the 1236 chemotherapy cycles. We also collected information about the administration of platelet transfusions, total parenteral nutrition, liquid diet, fluid replacement therapy, opioid analgesics, antibacterial antibiotics, antifungal agents, and antiviral agents.

Cost was not measured. However, we estimated cost based on the duration of hospitalization using a daily cost of $1000 per day. This figure approximates the U.S. mean Medicare payments, which range from $670 to $1080 per day, for common infection Diagnosis-Related Groups (79, 80, 89, 90, 416, 418, 419, 420, and 421) in 2002 U.S. dollars. We used an estimate at the upper limits of the range based on the assumption that infections in myelosuppressed patients with cancer would be more costly per day than infections in the general population. Furthermore, in the absence of data on actual cost, we used Medicare payments as a surrogate.

Statistical Considerations

The unit of analysis was 1 chemotherapy cycle, which was defined as the interval between the first day of chemotherapy and either 1) the start of the next cycle; 2) the day of platelet recovery (> 100,000 per μL), if no further therapy was received; or 3) 60 days or death, whichever came first, if no further therapy was received and if no platelet recovery occurred. We calculated the sample size based on the bleeding outcome, because we predicted that bleeding would be less common than episodes of infection. A sample of 635 patients was chosen originally based on predictions that each patient would contribute 2 eligible cycles and allowing for 0.5–1.0% failure to locate medical records. The target sample size of 1250 cycles was chosen to allow for 10 bleeding episodes per variable (on the basis of predictions that the bleeding rate would be 8% and the multiple-variable model would have ≤ 10 variables).19

We conducted descriptive analyses of patients with and without mucositis and of the associations between mucositis and 1) the outcomes and 2) the resources described above (Tables 1–3. We then developed separate, multiple-variable models for the infection and bleeding outcomes. We included factors associated with mucositis that were identified in the descriptive analyses as well as confounding factors that were identified from the literature. A cycle-level analysis with multiple cycles per patient would fail to account for within-patient correlations for potentially important factors. Therefore, multilevel regression models were fit with cycles nested within patients. This strategy permitted analysis at the cycle level while adjusting for within-patient correlations. Variables were added to the model (in ascending order by P value, with the most significant first) during sequential regression analyses until comparison of log-likelihood ratios failed to show significant improvement in the model (Table 4).

Table 1. Characteristics of Patients with and without Oral and Gastrointestinal Mucositis
CharacteristicNo mucositis (n = 778) (%)Mucositis
Oral only (n = 274) (%)GI only (n = 91) (%)Oral and GI (n = 93) (%)
  1. GI: gastrointestinal; ANC: absolute neutrophil count; 5-FU: 5-fluororacil; XRT: radiotherapy.

ANC < 100/μL (mean days)2211
Platelets < 50,000/μL (mean days)991010
Prior chemotherapy696 (89)246 (90)77 (85)81 (87)
Prior oral mucositis250 (32)144 (53)26 (29)48 (52)
Prior gastritis 29 (4) 15 (5) 4 (4) 8 (9)
Prior esophagitis 47 (6) 13 (5)10 (11) 7 (8)
Prior colitis 51 (7) 15 (5)19 (21)11 (12)
Prior peptic ulcer 11 (1)  4 (1) 1 (1) 2 (2)
Good performance status (0–1)493 (64)173 (63)51 (56)52 (56)
5-FU with or without leucovorin 63 (8) 59 (22) 7 (8)15 (16)
Prior XRT    
 None604 (78)218 (80)69 (76)66 (71)
 Any174 (22) 56 (20)22 (24)27 (29)
 Head and neck872399
 Thoracic/chest wall5225410
Current XRT    
 None763 (98)266 (97)84 (92)92 (99)
 Any 15 (2)  8 (3) 7 (8) 1 (1)
 Head and neck2201
 Thoracic/chest wall5450
Table 2. Outcomes of Patients with Oral and Gastrointestinal Mucositis
OutcomeNo mucositis (n = 778) (%)Oral mucositis onlyGI mucositis onlyBoth oral and GI mucositis
Any grade (n = 274) (%)Grade 3–4 (n = 70) (%)Any grade (n = 91) (%)Grade 3–4 (n = 21) (%)Any grade (n = 93) (%)Grade 3–4 (n = 40) (%)
  1. GI: gastrointestinal.

Dose reduction next cycle 86 (11) 68 (25)15 (21)18 (20) 5 (24)18 (19) 8 (20)
Delay next cycle 72 (9) 30 (11) 7 (10) 6 (7) 2 (10) 5 (5) 1 (3)
 Mean days1619111319810
Death unrelated to bleeding or infection 21 (3)  5 (2) 3 (3) 1 (1)
 Any 63 (8) 22 (8) 4 (6)11 (12) 1 (5)13 (14) 7 (18)
 Oral bleeding  6 (0.8)0002
 GI bleeding  5 (0.6)  2 (0.7) 5 (5.5) 2 (2.2)
 Bleeding-related death  9 (1)  2 (1) 2 (2) 3 (3)
 Mean days2232712
 Any283 (36)187 (68)61 (87)66 (73)15 (71)81 (87)37 (93)
 Fever of unknown origin142 (18)110 (40)36 (51)30 (33) 5 (24)35 (38)14 (35)
 Documented infection141 (18) 77 (28)25 (36)36 (40)10 (48)46 (49)23 (58)
  Unknown organism10661171962514
 Mean days of fever3.
 Infection-related death 13 (2)  7 (3) 3 (3) 8 (9)
 Any 38 (5) 33 (12)12 (17) 3 (3) 0 (0) 5 (5) 0 (0)
 Grade 3–491040030
Oral or GI pain       
 Any 10 (1)110 (40)44 (63)22 (24) 9 (43)39 (42)25 (63)
 Grade 3–424140882523
Weight loss       
 Any421 (54)168 (61)37 (53)54 (60)12 (57)60 (65)27 (68)
 Mean, kg2.
Table 3. Resource Utilization Associated with Oral and Gastrointestinal Mucositis
ResourceNo mucositis (n = 778) (%)Oral mucositis onlyGI mucositis onlyBoth oral and GI mucositis
Any grade (n = 274) (%)Grade 3–4 (n = 70) (%)Any grade (n = 91) (%)Grade 3–4 (n = 21) (%)Any grade (n = 93) (%)Grade 3–4 (n = 40) (%)
  1. GI: gastrointestinal; TPN: total parenteral nutrition.

ER visits (mean visits)
Hospital days (mean days)
Therapeutic platelet transfusions (mean no.)
TPN0  8 (3) 7 (10) 3 (3) 1 (5)10 (11) 7 (18)
Liquid diet  2 (< 1) 12 (4) 8 (11) 0 (0) 0 (0) 2 (2) 1 (3)
Fluids 19 (2) 28 (10)16 (23)11 (12) 7 (33)11 (12)11 (28)
Therapeutic opioids  6 (< 1) 19 (7)12 (17) 6 (7) 4 (19)13 (14)12 (30)
Antibacterial antibiotics283 (36)187 (68)61 (87)66 (73)15 (71)81 (87)37 (93)
Antifungal, antiviral       
 Any 90 (12)146 (53)53 (76)26 (29) 7 (33)73 (79)37 (93)
 Therapeutic only28102411025533
 Prophylactic only53283155101
 Any 50 (6) 93 (34)39 (56)10 (11) 2 (10)42 (45)25 (63)
 Therapeutic only216129202821
 Prophylactic only28321082144
 Any 81 (10)124 (45)47 (67)24 (26) 7 (33)65 (70)32 (80)
 Therapeutic only2110144925431
 Prophylactic only5622314581
Table 4. Relationship between Mucositis and Bleeding or Infectiona
FactorOR95% CIP value
  • OR: odds ratio; 95% CI: 95% confidence interval; GI: gastrointestinal; ANC: absolute neutrophil count.

  • a

    Multiple variable model parameters were adjusted for multiple cycles per patient.

Factors associated with bleeding   
 Oral mucositis present1.110.65–1.920.70
  Duration > 7 days0.690.32–1.470.33
 GI mucositis present2.001.18–3.370.01
  Duration > 7 days0.800.27–2.360.70
 Prior bleeding episode6.893.59–13.25< 0.0001
 Drug affecting platelet function5.803.40–9.90< 0.0001
 Platelet nadir   
  < 10,000.μL3.892.25–6.72< 0.0001
 Duration of thrombocytopenia > 7 days1.881.17–3.010.009
Factors associated with infection   
 Oral mucositis present2.401.76–3.26< 0.0001
  Duration > 7 days1.190.77–1.850.43
 GI mucositis present2.241.53–3.28< 0.0001
  Duration > 7 days1.010.47–2.160.98
 ANC nadir   
  < 100/μL1.851.31–2.630.0006
  100–1000/μL2.221.66–2.98< 0.0001
 Duration of neutropenia > 7 days1.080.76–1.540.65

Descriptive analyses were computed using BMDP Dynamic software (Version 7; BMDP Statistical Software, Inc., Los Angeles, CA) and SPSS software (Version 11.0; SPSS Inc., Chicago, IL). All regression analyses were computed with SAS software (Version 6.12; SAS Institute, Cary, NC) using the GENMOD (General Estimating Equations) procedure with an unstructured covariance matrix.


Mucositis was common among these patients with solid tumors who were receiving myelosuppressive chemotherapy. Overall, 303 of the 599 patients experienced oral and/or GI mucositis during 37% of 1236 cycles of chemotherapy. Oral mucositis developed during 22% of 1236 cycles, GI mucositis developed during 7% of cycles, and both oral and GI mucositis developed during 8% of cycles. Severe mucositis (NCI Grade 3 or 4) occurred less frequently, during 11% of cycles.

The depth and duration of neutropenia and thrombocytopenia were distributed similarly among cycles with and without mucositis (Table 1). Two factors were associated with oral mucositis: prior oral mucositis (41% vs. 31%; P < 0.0001) and prior gastritis (41% vs. 29%; P = 0.08). A history of gastritis also was associated with NCI Grade 3 or 4 oral mucositis (23% vs. 9%; P = 0.0003). GI mucositis was more common among patients who previously experienced esophagitis (22% vs. 14%; P = 0.10) or colitis (31% vs. 13%; P < 0.0001). Concurrent radiotherapy was associated with mucositis at the corresponding site and overall (51% vs. 37%), although the difference did not reach statistical significance due to the small number of cycles (31 cycles; P = 0.13). Oral mucositis occurred more frequently during cycles of chemotherapy with 5-FU (51%) than during cycles without 5-FU (27%; P < 0.001).

Clinical Outcomes of Mucositis

A reduction in the dose of the next cycle of chemotherapy was twice as common after cycles with mucositis as it was after cycles without mucositis (23% vs. 11%; P < 0.0001) (Table 2). Dose reductions occurred as frequently after Grade 1 or 2 mucositis (23%) as they did after Grade 3 or 4 mucositis (21%; P = 0.76); therefore, the decision to reduce the dose may have resulted from any of several highly correlated factors, such as mucositis-related episodes of infection or bleeding, neutropenia, or thrombocytopenia. We developed a multiple-variable model of dose reduction, with adjustment for within-patient correlation. In this analysis, dose reduction indeed was related to neutropenia (odds ratio [OR], 1.56; 95% confidence interval [95% CI], 1.07–2.28; P = 0.02). However, after accounting for the influence of neutropenia, Grade 1 or 2 oral mucositis increased the risk of dose reduction > 2-fold (OR, 2.31; 95% CI, 1.61–3.31; P < 0.0001), and Grade 3 or 4 oral mucositis nearly doubled the risk (OR, 1.95; 95% CI, 1.13–3.35; P = 0.02). Surprisingly, episodes of GI mucositis (P = 0.34), infection (P = 0.64), and bleeding (P = 0.38) were not associated with an increase in the risk of dose reduction. No difference was observed for delays in the next cycle, reflecting the similar distributions of neutropenia and thrombocytopenia between cycles with and without mucositis.

Although the depth and duration of thrombocytopenia were similar among patients with and without mucositis, episodes of bleeding were more common during chemotherapy cycles that were complicated by GI mucositis (13%) compared with cycles that were complicated by oral mucositis (8%; P = 0.08) or cycles without mucositis (8%; P = 0.04) (Table 2). Episodes of major bleeding complicated 6% of cycles with GI mucositis, compared with 1.8% of cycles with oral mucositis (P = 0.03) and 3.2% of cycles without mucositis (P = 0.08). Oral bleeding was uncommon and was not associated with oral mucositis. However, GI bleeding was significantly more common among patients with GI mucositis (3.8%) than among patients without GI mucositis (0.7%; P = 0.001). Half of the episodes of GI bleeding occurred during cycles with GI mucositis.

The risk of infection was significantly higher during cycles with mucositis (73%) than during cycles without mucositis, despite their similarity in depth and duration of neutropenia (36%; P < 0.0001) (Table 2). The risk of infection increased with increasing severity of mucositis, reaching 93% during cycles with Grade 3–4 oral and GI mucositis. More importantly, these findings were not restricted to fevers of unknown origin. The risk of documented infections increased from 18% during cycles without mucositis to 58% during cycles with Grade 3–4 oral and GI mucositis (P < 0.0001). In particular, gram-positive infections occurred during 3% of cycles without mucositis and during 11% of cycles with GI mucositis (P < 0.0001).

Fatigue was more common during cycles with mucositis (9%) than during cycles without mucositis (5%; P = 0.007), which also was true for weight loss (54% vs. 62%; P = 0.01) (Table 2). It is not surprising that oral or GI pain occurred frequently (37%) during cycles with mucositis. However, it is noteworthy that oral or GI pain also occurred during 1% of cycles without mucositis.

Seventy-seven patients died (13%), corresponding to 6% of the 1236 cycles (Table 2). Thirty-one deaths (40%) were infection-related, 16 deaths (21%) were bleeding-related, and the remaining deaths (30 deaths; 39%) were unrelated to those outcomes. Deaths due to causes other than infection or bleeding were distributed similarly among patients with and without mucositis. However, bleeding-related deaths occurred during 3% of cycles with GI mucositis compared with only 1% of cycles without GI mucositis (P = 0.06). Infection-related deaths were significantly more common during cycles with both GI and oral mucositis than among any other group (9% vs. 2%; P = 0.0009).

Resource Utilization during Mucositis

During chemotherapy cycles with mucositis, the average duration of hospitalization was significantly longer (Table 3), and liquid diets, total parenteral nutrition, fluid replacement, and antifungal and antiviral therapy and prophylaxis were more common. Cost was estimated rather than measured in the current study. However, assuming a conservative estimate of the cost of a hospital day ($1000, as described above; see Materials and Methods), the excess hospital days would have a major economic impact. The average cost of hospitalization would be $3893 per cycle without mucositis, $6277 per cycle with oral mucositis, $9132 per cycle with GI mucositis, and $9161 per cycle with both oral and GI mucositis. The incremental costs would be $2725 and $5565 per cycle with NCI Grade 1–2 mucositis and Grade 3–4 mucositis, respectively.

Multiple-Variable Model of the Relationship between Mucositis and Bleeding or Infection

Although mucositis was related to episodes of bleeding or infection in univariate analysis, multiple-variable models of bleeding and infection were developed to examine the numerous possible factors that could be correlated with both mucositis and outcomes of bleeding or infection. Hierarchic models were developed to adjust for within-patient correlation.

After accounting for the depth and duration of thrombocytopenia, prior episodes of bleeding, and administration of drugs known to affect platelet function, the risk of bleeding was doubled among patients with GI mucositis (P = 0.01) (Table 4). Neither oral mucositis nor the presence of infection was associated with an increased risk of bleeding. After accounting for the depth and duration of neutropenia, the risk of infection increased 2-fold among patients with oral mucositis or with GI mucositis (P < 0.0001). Episodes of bleeding and the duration of mucositis were not related to the risk of infection.


We observed mucositis during 37% of chemotherapy cycles, but this observation should be interpreted with caution. The incidence of mucositis varies with the chemotherapy regimen, and regimens vary over time and among centers. Furthermore, our study was restricted to cycles during which myelosuppression developed. The incidence of mucositis may be different among patients who receive similar regimens but do not develop myelosuppression.

Significant associations between mucositis and the development of serious clinical outcomes (specifically, bleeding and infection) and high resource utilization also were observed. The strength of the current study was its large sample of similarly myelosuppressive chemotherapy cycles in a largely unstudied population: patients with solid tumors. This design permitted examination of the impact of mucositis on the development of bleeding or infection among cycles with similar primary risk factors for these events (depth and duration of myelosuppression). However, the design was observational, and the data were collected retrospectively. Consequently, patients with and without mucositis may differ in important factors, other than the degree of myelosuppression, which may affect the development of bleeding or infection. Although a number of these were included in the multiple-variable models, it is possible that other, unmeasured factors were represented unequally among the groups. If these factors were associated with mucositis, then the resulting bias would produce spuriously large differences between cycles with and without mucositis.

Retrospective collection of data also may result in under-reporting of mucositis, bleeding, or infection. Under-reporting of infection is least likely. Because all patients received all their care at our institution and infection is a potentially life-threatening event among myelosuppressed patients, these events should be reported in medical records. Under-reporting of mild mucositis or mild bleeding is far more likely. When mucositis was underreported, mild cases would be included in the no mucositis group, biasing our estimates of association with bleeding and infection toward no difference. Under-reporting of mild bleeding events would produce erroneously low estimates of association. Thus, in both instances, associations between mucositis and bleeding would be understated.

The Clinical Burden of Mucositis

Our data suggest that the development of mucositis is an important chemotherapy dose-limiting factor among patients with solid tumors. The risk of dose reduction doubled when mucositis was present. Although the value of maintaining very high doses of chemotherapy is controversial, an association between maintaining standard doses and improved survival has been demonstrated for some cancers.20

Infection episodes and, to a lesser extent, bleeding episodes are the most common, potentially fatal complications of myelosuppressive chemotherapy. Our findings provide systematic evidence supporting the postulated relation between mucositis and these serious clinical outcomes; both were significantly more common during cycles with mucositis. Two findings involving GI mucositis are of particular clinical importance: the high rate of GI bleeding (4%) associated with GI mucositis compared with cycles without GI mucositis (< 1%) and the high rate of documented infections among patients with GI mucositis. We observed a threefold increase in documented infections during cycles with GI mucositis. In fact, the frequency of documented infections among patients with solid tumors who had GI mucositis approached that observed among bone marrow transplantation recipients with mucositis, a far more immunosuppressed population.7

We also observed that infection-related and bleeding-related death rates were significantly higher among patients with GI mucositis. These findings are of particular concern, because high rates of GI toxicity have been reported with new chemotherapy agents (namely, irinotecan) that were not represented in this analysis. If our observations generalize to such agents, then an increase in these serious clinical outcomes may be forthcoming.

Pain is a clinically significant component of mucositis and an important factor in decreased quality of life among cancer patients. We observed NCI Grade 3–4 pain during 37% of cycles with oral or GI mucositis, but opioid analgesia was prescribed for only 8% of patients. Undertreatment of pain among patients with cancer has been reported previously in many settings. It appears to be particularly problematic in patients who have the short-lived although severe pain that characterizes mucositis.21–24 This observation underscores the clinical burden of mucositis and points to goals for improvement in the clinical management of this problem.

The Economic Burden of Mucositis

Considering the serious outcomes associated with mucositis, it is not surprising that resource utilization was increased significantly among these patients. Because mucositis disturbs oral intake as well as absorption of nutrients, an association with nutritional adjuncts would be expected. In this respect, our results did not disappoint; fluid replacement therapy, liquid diets, and total parenteral nutrition were significantly more common during cycles with mucositis. Because of the association with infection, antibiotic therapy and prophylaxis also were significantly more common among patients with mucositis.

It is of even greater economic import that the serious outcomes of mucositis led to significantly more days of hospitalization per cycle: economically significant events. Cycles with GI mucositis had the longest durations of hospitalization, exceeding 10 days per cycle during cycles with NCI Grade 3 or 4 GI mucositis. Our conservative estimates suggest that the incremental cost of hospitalization may exceed $3500 per cycle with mucositis. Sonis and colleagues reported hospital charges exceeding $40,000 per patient with oral ulcerations during the 100-day posttransplantation engraftment period.7 Although mucositis is more frequent and may be more expensive per episode among bone marrow transplantation recipients, the annual number of cycles of myelosuppressive chemotherapy administered to patients with solid tumors far exceeds the number of bone marrow transplantations. From a societal perspective, the overall cost of mucositis among patients with solid tumors would far exceed the overall cost among transplantation recipients.

Our results support the hypothesis that mucositis is associated with increased incidence of serious clinical outcomes and with excess resource utilization among patients with solid tumors who receive myelosuppressive chemotherapy. Therapies are needed to reduce the incidence and mitigate the severity of this common problem.