Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
IMPACT OF NUTRITION SUPPORT ON TREATMENT OUTCOME IN PATIENTS WITH LOCALLY ADVANCED HEAD AND NECK SQUAMOUS CELL CANCER TREATED WITH DEFINITIVE RADIOTHERAPY: A SECONDARY ANALYSIS OF RTOG TRIAL 90-03 Rachel Rabinovitch, MD,1 Barbara Grant, MS, RD,2 Brian A. Berkey, PhD,3 David Raben, MD,2 Kie Kian Ang, MD,4 Karen K. Fu, MD,5 Jay S. Cooper, MD6 for the Radiation Therapy Oncology Group 1 Department of Radiation Oncology, University of Colorado Health Sciences Center, Anschutz Cancer Pavilion, 1665 N. Ursula Street, Suite 1032, Box F706, Aurora, CO 80045. E-mail: [email protected] 2 Cancer Care Center, Saint Alphonsus Regional Medical Center, Boise, Idaho 3 RTOG Headquarters, Philadelphia, Pennsylvania 4 Department of Radiation Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas 5 Department of Radiation Oncology, University of California San Francisco, San Francisco, California 6 Department of Radiation Oncology, Maimonides Medical Center, Brooklyn, New York Accepted 1 August 2005 Published online 14 November 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hed.20335 Abstract: Background. The aim was to evaluate the relationship between nutrition support (NS) on host toxicity and cancer outcome in patients with locally advanced head and neck squamous cell carcinoma (HNSCC) undergoing definitive radiotherapy (XRT). Methods. We performed a secondary analysis of Radiation Therapy Oncology Group (RTOG) 90-03, a prospective randomized trial evaluating four definitive XRT fractionation schedules in patients with locally advanced HNSCC, which prospectively collected data on NS delivered before treatment (BNS), during treatment (TNS), and after definitive XRT. NS data and pretreat- Correspondence to: R. Rabinovitch Presented at the annual meeting of the American Society for Therapeutic Radiology and Oncology, 2002. C V 2005 Wiley Periodicals, Inc. Impact of Nutrition Support ment characteristics of the 1073 evaluable patients were analyzed against therapy toxicity and outcome. Results. Patients receiving BNS experienced significantly less weight loss by the end of treatment and less grade 3 to 4 mucositis than patients not receiving BNS. However, patients receiving BNS had a poorer 5-year actuarial locoregional control rate than patients receiving TNS or no NS (29%, 55%, and 57%, respectively, p < .0001) and a poorer 5-year overall survival rate (16%, 36%, and 49%, respectively, p < .0001). Patients receiving BNS were significantly more likely to have a higher T classification, N status, and overall American Joint Committee on Cancer (AJCC) stage and initial presentation with greater pretreatment weight loss, and a poorer Karnofsky Performance Status (KPS) than patients not receiving BNS. After adjusting for the impact of these prognostic factors through a recursive partition analysis, a multivariate analysis with a stratified Cox model found that BNS was still a highly significant independent prognostic factor for increased locoregional failure (hazards ratio [HR], HEAD & NECK—DOI 10.1002/hed April 2006 287 1.47; 95% confidence interval [CI], 1.21–1.79; p < .0001) and death (HR, 1.41; 95% CI, 1.19–1.67; p < .0001). Conclusion. In this study, the largest prospective evaluation of nutrition data in treated patients with cancer, BNS was associated with inferior treatment outcome in the patients with HNSCC undergoing XRT. These results should be considered hypothesis generating and encourage prospective clinical research and C 2005 identification of the mechanisms underlying this finding. V Wiley Periodicals, Inc. Head Neck 28: 287–296, 2006 Keywords: nutrition; radiotherapy; outcome; cancer; head and neck Malignancies of the head and neck frequently result in diminished nutritional intake because of local symptoms (painful and impaired chewing and swallowing) caused by physical invasion of regional structures. Head and neck cancers, similar to many other malignancies, may also induce constitutional symptoms (decreased appetite and cachexia), further negatively impacting on nutritional intake and maintenance of body mass. Specific to this population, patients with head and neck cancer commonly present with preexisting nutritional deficiencies because of poor lifetime personal habits, such as excessive alcohol consumption, tobacco use, and longstanding inadequate nutritional intake and a greater incidence of anemia.1,2 It is, therefore, a significant challenge to maintain the nutritional status of the patient with head and neck cancer throughout a definitive course of head and neck radiotherapy, given the profound acute impact of radiation on oropharyngeal mucosa. For all of these reasons, patients undergoing definitive head and neck radiotherapy frequently require alternate methods of nutrition support (NS), administered through various combinations of oral, enteral (via tube), and parenteral routes. Depending on the severity of pretreatment swallowing function, pain, and treatment-induced mucositis, NS may partially or completely replace normal oral nutritional intake. There is ample information that NS enhances host outcomes (ie, improves the cancer patient’s subjective quality of life) and improves or stabilizes patient body mass.3–6 Although many patients and clinicians express the notion that nutritional support may be beneficial in ‘‘fighting the cancer,’’ the limited data available from small groups of treated animals and human patients actually suggest that NS has a negative association with cancer outcomes (ie, increased tumor growth and decreased effectiveness of cancer therapy).7 There are little clinical data in humans evaluating the association of NS or other specific nutritional pa- 288 Impact of Nutrition Support rameters with host outcomes (acute and late toxicity) in a uniformly treated population of patients with cancer; there are also little data on the potential relationship of NS on cancer therapy outcome in such patients. This study, an unplanned re-analysis of the Radiation Therapy Oncology Group (RTOG) protocol 90-03 was performed to evaluate for an association of NS with host toxicity and cancer outcomes in patients treated with definitive radiotherapy for locally advanced head and neck squamous cell carcinoma (HNSCC). RTOG 90-03, a definitive radiotherapy treatment trial for patients with HNSCC, prospectively collected data on NS for patients at baseline (BNS), during treatment (TNS), and throughout follow-up. Issues related to this nutrition data have not been previously analyzed. This analysis constitutes the largest evaluation of prospectively collected NS data and its effect on treatment outcome in patients with cancer participating in a clinical cooperative group trial. MATERIALS AND METHODS RTOG 90-03 was a phase III prospective randomized trial designed to compare four definitive radiotherapy fractionation schedules for patients with locally advanced HNSCC: standard fractionation, twice daily hyperfractionation (HFX), accel erated fractionation with a split (AFX-S), and accelerated fractionation with a concomitant boost (AFX-C). The primary objectives of this trial were to (1) test the efficacy of hyperfractionation and two types of accelerated fractionation against standard fractionation with regard to locoregional control, and (2) compare acute and late toxicities between the different fractionation regimens. Between 1991 and 1997, 1113 patients were enrolled, and 1073 of these patients were analyzed for outcome. Neck dissection was allowed only for neck nodes >3 cm before radiation therapy, and patients did not receive chemotherapy during initial treatment. The overall details and results of this trial were published in 2000.8 This analysis examined pretreatment patient and tumor characteristics; type of NS used before, during, and after radiation therapy; incidence and severity of acute and late toxicities; elapsed days of treatment; and ultimate treatment outcome (Table 1). NS information was documented in the protocol’s data capture forms as follows: (1) oral liquid nutritional supplements, (2) enteral nutrition by means of a feeding tube, HEAD & NECK—DOI 10.1002/hed April 2006 Table 1. Variables analyzed. o o o o o o o o o Primary tumor site T classification, N classification, and AJCC stage Age Sex KPS Assigned protocol treatment Dysphagia at baseline (pretreatment) Mucositis at treatment completion Specific nutrition support (oral liquid nutrition supplements, enteral nutrition via feeding tube, or parenteral nutrition) n Baseline (before treatment initiation) n Completion of treatment n 3 mo after treatment n 6 mo after treatment o Pretreatment Hgb o Weight loss n 6 mo before treatment n At completion of treatment o Treatment outcomes n Incidence of treatment breaks n Elapsed days of treatment n Local regional control n Overall survival Abbreviations: AJCC, American Joint Committee for Cancer; KPS, Karnofsky Performance Status; Hgb, hemoglobin. and/or (3) parenteral nutrition. Of note, RTOG 90-03 did not mandate any guidelines for NS initiation or content. The distribution of patient and treatment characteristics for various patient subgroups was compared by the Pearson chi-square test for discrete data and the Wilcoxon test for continuous data. Locoregional control was estimated by the method of cumulative incidence,9 and statistical testing was performed with Gray’s test.10 Overall survival (OS) was estimated by the Kaplan-Meier method,11 and statistical testing was performed with the log-rank test.12 A recursive partitioning analysis (RPA) was performed to identify subgroups of patients that had distinct results with respect to survival and locoregional control using pretreatment prognostic factors for patients with HNSCC (age, sex, weight loss before start of treatment, hemoglobin, T classification, N status, overall American Joint Committee on Cancer [AJCC] stage, Karnofsky Performance Status [KPS], primary site, and baseline dysphagia). The RTOG previously used RPA methodology to classify patients with head and neck cancer.13 The analysis involves dichotomizing the patients using all logical splits of variables and finding the most significant difference between the two groups (adjusting the level of significance for the multiple comparisons). The Impact of Nutrition Support process is repeated within each subgroup until all significant splits have been found and a set of prognostic classes has been determined. Within each final prognostic class, a univariate comparison was performed to assess for the impact of 6 BNS on locoregional failure and death from any cause. The Cox proportional hazards model14 was used for testing of 6 BNS on both locoregional failure and death in the entire set of patients stratified by the RPA prognostic classes. RESULTS Pretreatment characteristics of the 1073 evaluable patients are shown in Table 2. Patients were generally men, middle-aged, and had stage III or IV HNSCC; the oropharynx was the most common primary site. Before the initiation of radiation therapy, 27% of all patients (n ¼ 293) were receiving BNS, evenly distributed among protocol treatment arms. Of the patients receiving BNS, 50% used only oral liquid NS, 27% used only enteral nutrition by a feeding tube, 16% used a combination of oral and enteral supplementation, and 6% used parenteral support. Most patients (86%) required some form of TNS, with an even distribution among the treatment arms. Patients receiving BNS were significantly more likely to have a poorer KPS, higher primary tumor (T) classification, more extensive regional lymph node involvement (N status), greater overall AJCC stage, and a greater incidence of anemia Table 2. Pretreatment characteristics. Characteristic Sex Male Female Age, y Median Range AJCC stage II III IV Dysphagia None Mild Moderate Severe Unknown Result 79% 21% 61 30–90 4% (base of tongue and hypopharynx) 28% 68% 39% 23% 30% 8% <1% Abbreviation: AJCC, American Joint Committee on Cancer. HEAD & NECK—DOI 10.1002/hed April 2006 289 Table 3. Association between BNS and pretreatment characteristics. Pretreatment characteristic KPS 90–100 60–80 T classification T1–2 T3–4 N classification NO Nþ AJCC stage II–III IV Dysphagia at baseline None/mild Moderate/severe Hgb Normal Anemia* Assigned treatment Standard HFX AFX-S AFX-C % of patients by BNS status BNS þBNS p value 71 29 35 65 <.001 40 60 14 86 <.001 24 76 17 83 <.017 37 63 17 83 <.001 72 28 34 66 <.001 47 53 75 25 <.0001 25 24 25 26 24 25 28 23 .72 Abbreviations: BNS, baseline nutrition support; KPS, Karnofsky Performance Status; AJCC, American Joint Committee on Cancer; Hgb, hemoglobin; HFX, hyperfractionation; AFX-S, accelerated fractionation with a split; AFX-C, accelerated fractionation with a concomitant boost. *Anemia defined as <14.5 g% in men, and <13.0 g% in women. compared with patients who did not (Table 3). Patients receiving BNS had significantly greater weight loss in the 6 months preceding treatment (means of 7.8 kg vs 3 kg) than patients receiving no BNS (p < .0001). Patients presenting with grade 3 or 4 dysphagia before treatment initiation were significantly more likely to receive BNS (p < .0001) and TNS (p < .015) than those with grade 1 or 2 dysphagia, an observation consistent between treatment arms. By the completion of treatment, patients who received BNS had significantly less weight loss (median of 5% vs 7%,p< .0001) than those who did not. There was a trend (p ¼ .057) toward a lower incidence of grade 3 or 4 mucositis after treatment in patients who received BNS (100 of 293; 34%) compared with those who did not (311 of 780; 40%). At the time of this analysis (median follow-up, 64.2 months), there had been 553 local or regional failures and 757 deaths reported among the evaluable patients. As measured by univariate analysis, patients receiving BNS, TNS, and no 290 Impact of Nutrition Support NS had 5-year actuarial locoregional control rates of 29%, 55%, and 57%, respectively (Figure 1). The locoregional control rate for patients who received BNS was significantly less than for patients who received TNS and those receiving no NS, a consistent finding regardless of whether data were analyzed for all patients in the study (Figure 1) or by individual treatment arm (data not shown). Similarly, there was no significant difference in locoregional control rates between patients receiving no NS and patients receiving TNS, either in the entire cohort of patients or within each treatment arm. Univariate analysis also demonstrated that patients receiving BNS had significantly poorer 5-year actuarial OS compared with patients receiving TNS or no NS (16%, 36%, and 49%, respectively; p < .0001) (Figure 2). This significantly poorer OS rate for patients who received BNS was observed within each treatment arm (data not shown). OS for patients receiving TNS was significantly inferior to patients receiving no NS in only one of the treatment arms: accelerated fractionation with split (p ¼ .046). Patients receiving BNS were no less likely to receive their total prescribed radiation dose than other patients, means of 71.5 Gy vs 72.6 Gy, respectively (p ¼ .14). Patients receiving BNS also completed their treatment within the same time frame as other patients: means of 46.9 days versus 46.7 days, respectively (p ¼ .69). The results of the RPA analyses for locoregional control and OS can be seen in Figures 3 FIGURE 1. Actuarial locoregional control by level of nutrition support (all patients). Patients who received baseline nutrition support before treatment initiation (BNS) had significantly inferior locoregional control compared with patients receiving nutrition support during treatment only (TNS) or no nutrition support at baseline or during treatment (NS). HEAD & NECK—DOI 10.1002/hed April 2006 FIGURE 2. Actuarial overall survival (OS) by level of nutrition support (all patients). Patients who received baseline nutrition support before treatment initiation (BNS) had significantly inferior OS compared with patients receiving nutrition support during treatment only (TNS) or no nutrition support at baseline or during treatment (NS). and 4, respectively. The tree for locoregional control endpoint has nine prognostic classes with sample sizes ranging from 10 to 428; the tree for the OS endpoint yielded six prognostic classes with sample sizes ranging from 62 to 497. Tables 4 and 5 show hazard rates for locoregional failure and death, respectively, comparing the patients who did and did not receive BNS within each of the RPA-derived prognostic classes.With the exception of one class in which the ratio was equal to 1, all of the hazard ratios are greater than 1, indicating that in each class patients who received BNS had an increased risk of failure, al- though the effect did not achieve significance, likely because of small sample sizes. The Cox proportional hazards model was then used to evaluate the impact of BNS after stratifying for derived prognostic classes of each endpoint. BNS was found to be highly significant for an increased risk of locoregional failure (hazard ratio [HR], 1.47; 95% confidence interval [CI], 1.21–1.79; p < .0001) and death (HR, 1.41; 95% CI, 1.19–1.67; p < .0001). An analysis of treatment outcome taking into account each of the routes of BNS administration (ie, oral, enteral, and parenteral) was performed using a Cox model stratified by the derived prognostic classes in the RPA. These results (Table 6) continue to demonstrate significantly higher risks of locoregional failure and death for each of the routes of BNS (p ¼ .0033 and .0047, respectively). It should be noted that the effects of oral and enteral nutrition were similar to one another for each outcome, whereas the parenteral effect was larger. DISCUSSION The recognized goals of nutritional intervention in the patient with cancer are to reverse or reduce nutritional deficiencies,15 preserve lean body mass, aid in recovery and healing,16,17 and maximize quality of life (QOL).18 These ‘‘host’’ benefits of nutritional support for the patient with cancer (ie, benefits to the patient’s own body and QOL) FIGURE 3. Recursive partition analysis (RPA) derived tree of locoregional control. Impact of Nutrition Support HEAD & NECK—DOI 10.1002/hed April 2006 291 FIGURE 4. Recursive partition analysis (RPA) derived tree of overall survival. are established in the literature.6,19–21 Klein et al22 reviewed 28 prospective randomized trials of F TPN in patients with cancer and reported pooled reduction in postoperative complications (p ¼ .01) and postoperative mortality (p ¼ .02) for patients receiving TPN. Consistent with these Table 4. Effect of baseline nutrition support on locoregional failure within each RPA class and in all patients. Class 1 2 3 4 5 6 7 8 9 All BNS status No. of patients No. of events BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS 12 43 63 120 6 21 40 388 13 43 28 50 9 1 88 48 32 59 291 773 6 14 41 68 5 13 22 130 9 20 18 20 8 1 75 39 22 37 206 342 HR (95% Cl) 1.43 (0.55–3.72) Impact of Nutrition Support Table 5. Effect of baseline nutrition support on death within each RPA class and in all patients. 1.36 (0.92–2.02) 1.25 (0.44–3.55) 1.98 (1.26–3.12) Class 1 1.96 (0.87–4.38) 2 1.82 (0.96–3.44) 3 – 4 1.34 (0.90–1.99) 5 1.16 (0.68–1.98) 6 1.47 (1.21–1. 79)* p < .0001 Abbreviations: RPA, recursive partitioning analysis; BNS, baseline nutrition support; HR, hazard rate; Cl, confidence interval. The HR is > 1 in all classes analyzed, demonstrating an increased risk of local failure for patients using BNS. There is a significant increased HR for locoregional failure associated with BNS in all patients, stratified for known prognostic variables. *From a model stratified by the derived prognostic classes. 292 benefits of NS, patients who received BNS in this RTOG study reaped host benefits as well: less weight loss by treatment completion and a lower incidence of grade 3 to 4 mucositis after the conclusion of therapy. One can further infer that the addition of BNS allowed these patients to complete therapy in an identical time frame as patients not receiving NS, despite their poorer KPS and greater tumor burden. A commonly held presumption exists that as a result of the documented host benefits derived by All BNS status No. of patients No. of events BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS BNSþ BNS 24 79 166 145 16 81 41 359 32 58 12 50 291 772 21 56 149 121 13 56 29 184 25 44 12 41 249 502 HR (95% Cl) 1.43 (0.87–2.37) 132 (1.03–1.68) 1.27 (0.69–2.32) 1.82 (1.23–2.70) 1.00 (0.61–1.63) 3.02 (1.56–5.85) 1.41 (1.19–1.67)* p < .0001 Abbreviations: RPA, recursive partitioning analysis; BNS, baseline nutrition support; HR, hazard rate; Cl, confidence interval. The HR is > 1 in all but class 5. There is a significant increased HR for death associated with BNS in all patients, stratified for known prognostic variables. *From a model stratified by the derived prognostic classes. HEAD & NECK—DOI 10.1002/hed April 2006 Table 6. Effect of route of BNS administration locoregional failure and death (each stratified by derived prognostic classes). Route of BNS Locoregional failure None Oral Enteral Parenteral Death None Oral Enteral Parenteral HR (95% Cl) p value — 1.21 (0.97–1.50) 1.30 (1.01–1.66) 2.40 (1.36–4.21) .0033 — 1.27 (1.05–1.52) 1.22 (0.98–1.53) 1.75 (1.07–2.87) .0047 Abbreviations: BNS, baseline nutrition support; HR, hazards rate; Cl, confidence interval. the patient with cancer from NS, a consequent negative impact on the malignancy should result. Patients commonly express this belief— that good nutrition support is important to ‘‘help my body fight the cancer.’’ Preclinical and clinical data, however, suggest just the opposite. That is, whereas NS and appropriate dietary intake benefit the host by preserving body mass, minimizing toxicity to therapy, and improving QOL, NS seems to similarly benefit the malignancy as well.7 Although the cancer clinician may be surprised by this concept, there is a growing body of literature, reported primarily in nutrition journals— and hence largely unseen by most clinical oncologists—describing the tumor-enhancing effects of NS in both animals and humans. Researchers experienced in feeding and ‘‘treating’’ tumor-bearing rodents are familiar with this relationship. As early as 1909, Moreschi23 documented that tumors transplanted into mice fed ‘‘ad libitum’’ (as much as desired) grew better than tumors transplanted into underfed mice. The literature contains several descriptions of diet restriction inhibiting tumor growth in tumor-bearing animals.24,25 Conversely, evidence exists that total parenteral nutrition (TPN) delivery to tumor-bearing animals increases tumor mass26–28 and stimulates a host of markers indicative of tumor cell growth.29The stimulatory effect of TPN on tumor growth parameters is recognized in rodent research, and recent work has focused on manipulating specific dietary components to counteract this phenomenon.30–32 Ye et al,30 Millis et al,31 and He et al32 have all independently compared various nutrition formulations and found that manipulation of specific amino acid concentrations—focusing on arginine and methionine balance—inhibited tumor growth Impact of Nutrition Support in tumor-bearing rodents. Interestingly, controversy exists in the literature as to whether TPN stimulates tumor growth in the animal model to a greater degree than oral dietary intake.28,33 In humans, epidemiologic data consistently demonstrate an association between obesity and an increased incidence of numerous cancers (breast, uterus, esophagus, liver, prostate, gallbladder, larynx, kidney, cervix, ovary, brain, colon, and lymphoma).34–36 This observation indirectly refutes the notion that well-fed individuals have better immune or other mechanisms to prevent cancer induction or cancer growth. Clinical data directly evaluating the impact of NS in cancer-bearing humans has involved small patient numbers and frequently non-uniformly staged or treated patients. Nevertheless, the available information also is consistent with the rodent model described previously. A retrospective analysis of the protein and energy intake of 37 patients with metastatic melanoma or renal cell cancer showed a 50% reduction in complete and partial responses to high-dose interleukin-2 therapy among patients receiving 14 days of concurrent parenteral nutrition compared with controls with oral intake only.37 Overall survival was shown to be unchanged between the two groups; however, time to progression of disease was decreased, and tumors progressed 17% faster in patients receiving parenteral nutrition. Baron et al38 evaluated tumor samples obtained from 14 untreated patients with HNSCC before and after TPN administration and compared them with normal tissue samples from the same patients. They identified a significant increase in the percentage of hyperdiploid cells in tumors after TPN delivery, but no such changes were observed in normal mucosa. Bozzetti et al39 evaluated the impact of preoperative TPN versus none in 19 malnourished patients with gastric cancer. Increased 3 H-thymidine labeling index, a marker for S-phase cell fraction or tumor cell proliferation, was identified in half of the endoscopically biopsied tumor specimens. One of the largest prospective trials published to date randomized 92 patients with operable gastrointestinal cancer (gastric, colon, and rectal tumors) and malnutrition to one of four interventions before definitive surgery: parenteral NS, parenteral NS and chemotherapy, chemotherapy, or nothing (control).40 Parenteral NS resulted in a significant stimulation of tumor proliferation as measured by an increase of the percent of tumor cells in S phase, DNA content, and DNA index. HEAD & NECK—DOI 10.1002/hed April 2006 293 Overall, clinical outcomes with regard to tumor control and survival were not evaluated. Although there is a clinical trial in small cell lung cancer that did not identify a negative effect of intravenous hyperalimentation on cancer therapy outcome,41 it is important to note that there is no published documentation supporting the idea that delivery of unmanipulated diets/NS to humans or animals enhances tumor control or improves cancer therapy outcome. Our analysis of the 1073 patients treated and evaluable on RTOG 90-03 comprises the largest evaluation of NS on cancer treatment outcome in a uniformly staged and treated group of patients with cancer, albeit in an originally unplanned analysis. The significant and large negative association of BNS on both locoregional control and OS described here is striking and lends considerable weight to the repeated observation that NS is associated not only with improved patient outcomes but inferior cancer outcomes as well. The magnitude of the negative effect of BNS on treatment outcome in the initial univariate analysis of this report is impressive: patients who received BNS had a 28% absolute reduction in locoregional control compared with patients not using any NS (29% vs 57% at 5 years), a relative decrease of 49%. Similarly, the absolute difference in OS rate for patients receiving BNS compared with those never using NS was 33% (16% vs 49% at 5 years), a relative decrease of 67%. Because these observations may have important clinical implications, we made a concerted effort to determine that the negative association observed was indeed related to the BNS and not the severe imbalance of the poorer prognostic features observed in those patients who received BNS (advanced T classification, N status, and AJCC stage, poor KPS, undesirable primary tumor site, and greater pretreatment weight loss). To account for them, an RPA was performed identifying discrete classes of patients, which took into account each of the poor prognostic factors on locoregional control and OS. (RPA is a statistical tool used to determine the association of prognostic variables with a specific outcome). Only then was BNS evaluated within each individual prognostic class for its impact on locoregional failure and death, allowing us to isolate any effect of BNS from the other prognostic factors already taken into consideration. Hazard rates greater than 1 (indicating a higher risk of failures) were seen in all but one class for both locoregional failure and death (Tables 4 and 5). Evaluation for an 294 Impact of Nutrition Support association of BNS on the entire group of evaluable patients, stratified for the various RPA prognostic classes, was highly significant (p < .0001), demonstrating a 47% increased risk of locoregional failure and a 41% increased risk of death in patients receiving BNS. This method of analysis, in our opinion, is sufficiently rigorous to account for the differences in the prognostic profiles of the patients who did and did not receive BNS. We, therefore, believe that these data confirm an association between use of BNS and greater locoregional failure and death in the patients evaluated on this study. This negative association is consistent with the direct and indirect evidence summarized earlier, supporting the theory that NS has a negative impact on cancer therapy outcome. The findings of this report raise many questions. What are the physiologic, cellular, and molecular mechanisms for decreasing the effectiveness of radiotherapy when patients receive BNS? How is the patient who receives BNS nutritionally different than someone who does not require supplementation? Will our observation be validated in the setting of other malignancies and with other anticancer modalities? How can the negative association of BNS on treatment outcome be mitigated by either altering the composition of the BNS or by introducing other interventions? We acknowledge that an analysis of NS and cancer outcome was not designated as an up-front endpoint in this study, and the collected information regarding BNS is quite limited (ie, no data captured on duration of support, nutritional content of support). However, the sheer magnitude and statistical significance of the negative impact of BNS identified on multivariate analysis (47% increased risk of locoregional failure and 41% increased risk of death, both p < .0001) after accounting for other associated negative prognostic factors through creation of RPA trees, lends weight to the credence of these data. The value of retrospective analyses is precisely that which was achieved here, to mark effects and interactions not initially considered in the study design. At a minimum, our data are hypothesis generating and justify a prospective analysis. Evaluating BNS prospectively, the logical next step, will, however, present substantial challenges. BNS is often independently initiated by the patient with cancer before evaluation by an oncologist, preventing participation in a clinical trial evaluating this parameter de novo. The ethics of randomizing a nutritionally compromised patient with HEAD & NECK—DOI 10.1002/hed April 2006 cancer to 6 BNS will need thoughtful discussion. (The RTOG is currently comparing a standard NS preparation to Juven— a NS that is rich in arginine, glutamine, and b-hydroxy b-methyl butyrate, in cancer patients with cachexia. The intended endpoints of this study, however, are changes in weight, lean body mass, fatigue, and QOL. This trial will enroll patients with advanced stages of different malignancies and is not designed to evaluate survival or any specific cancer therapy).42 Other approaches include attempts at reproducing our results through similar retrospective analyses of clinical trials if BNS information is available and determining whether the effect demonstrated here is replicated in similarly treated patients and applicable to other disease sites and treatment modalities. Integrating detailed objective NS parameters into future prospective clinical trials and designing such trials with the intent of evaluating the impact of BNS on outcome would bring us closer to validating (or refuting) the implications of our findings. This would require documentation of routes and duration of supplementation delivery, as well as the nutritional content of both supplements and normal dietary intake for all patients. Using simple subjective nutritional assessment scales such as the Subjective Global Assessment tool43 would assist the clinician in establishing pretreatment nutrition status and determining whether this well-established tool, as an example, correlates with treatment outcome. Given the enormous range of possibilities for the physiologic mechanisms underlying our observations, which our results cannot even begin to hint at, the next generation of trials might include evaluation of an array of relevant plasma cytokines and tumor biomarkers drawn before, during, and after therapy in an attempt to identify discrete molecular associations with outcome or differences between patients receiving and not receiving BNS. Cytokines that are currently thought to play a role in cachexia and other metabolic alterations in the patient with cancer include prostaglandin E2 (PGE2), tumor necrosis factor-a (TNFa), interleukin (IL)-1, IL-6, interferon-g (IFNg), and proteoglycan 24K.44 Integrated collaboration between nutrition and oncology researchers is sorely needed. In conclusion, the available published literature demonstrates that NS is associated with improved host outcomes, defined as improved QOL and tolerability of treatment. However, here we identify a Impact of Nutrition Support clinically meaningful and statistically significant negative association of BNS on cancer treatment outcome in a group of patients with HNSCC treated with definitive radiotherapy, consistent with other available clinical and preclinical data. If the bound-to-be-controversial results of this analysis are confirmed in the future, the benefits of BNS on the one hand will need to be weighed against the counter impact of such support on tumor control. Designing interventions will then urgently be needed for the nutritionally compromised patient with cancer, currently situated between this proverbial rock and a hard place. REFERENCES 1. Cancer facts & figures 2000. Atlanta, GA: American Cancer Society; 2000. p 1–40. 2. Feldman JG, Hazan M. A case controlled investigation of alcohol, tobacco and diet in head and neck cancer. Prev Med 1975;4:4444–4463. 3. Rivadeneira DE, Evoy D, Fahey TJ, et al. Nutritional support of the cancer patient. CA Cancer J Clin 1988; 48:69–80. 4. Chlebowski RT, Palomares MR, Lillington L, et al. Recent implications of weight loss in lung cancer management. Nutrition 1996;12:42–47. 5. Nixon DW. Cancer, cachexia and diet: lessons from clinical research. Nutrition 1996;12:52–56. 6. Van Bokhorst-de Van der Schuer MA, Langendoen SI, Vondeing H, et al. Perioperative enteral nutrition and quality of life of severely malnourished head and neck cancer patients: a randomized clinical trial. Clin Nutr 2000;19:437–444. 7. Canada T. Clinical dilemma in cancer: is tumor growth during nutrition support significant? Nutr Clin Practice 2002;17:246–248. 8. Fu KK, Pajak TF, Trotti A, et al. A radiation therapy oncology group (RTOG) phase III randomized study to compare hyperfractionation and two variants of accelerated fractionation to standard fractionation radiotherapy for head and neck squamous cell carcinomas: first report of RTOG 9003. Int J Radiat Oncol Biol Phys 2000;48: 7–16. 9. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: John Wiley and Sons; 1980. 10. Gray RJ. Class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16:1141–1154. 11. Kaplan EL, Meier P. Nonparametric estimation of incomplete observations. J Am Stat Assoc 1958;53:457– 481. 12. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chem Rep 1966;5:163–170. 13. Cooper JS, Farnan NC, Asbell SO, et al. Recursive partitioning analysis of 2,105 patients treated in RTOG studies of head and neck cancer. Cancer 1996;77:1905–1911. 14. Cox DR. Regression models and life tables. J R Stat SocSeries B 1972;34:187–229. 15. De Cicco M, Panarello G, Fantin D, et al. Parenteral nutrition in cancer patients receiving chemotherapy: effects on toxicity and nutrition status. J Parenteral Enteral Nutr 1993;17:513–518. 16. Gianotti L, Braga M, Nespoli L, et al. A randomized controlled trial of preoperative oral supplementation with a HEAD & NECK—DOI 10.1002/hed April 2006 295 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 296 specialized diet in patients with gastrointestinal cancer. Gastroenterology 2002;122:1763–1770. Bertrand PC, Piquet MA, Bordier I, et al. Preoperative nutritional support at home in head and neck cancer patients: from nutritional benefits to the prevention of the alcohol withdrawal syndrome. Curr Opin Clin Nutr Metab Care 2002;5:435–440. Eldridge B, Rock CL, McCallum PD. Nutrition and the patient with cancer. In: Coulston NM, Rock CL, Monsen ER, editors. Nutrition in the prevention and treatment of disease. San Diego, CA: Academic Press; 2001. p 397– 412. Shike M, Russel DM, Detsky A, et al. Changes in body composition in patients with small-cell lung cancer: the effect of total parenteral nutrition as an adjuvant to chemotherapy. Ann Intern Med 1984;101:303–309. Isenring E, Bauer J, Capra S. The scored patient-generated subjective global assessment (PG-SGA) and its association with quality of life in ambulatory patients receiving radiotherapy. Eur J Clin Nutr 2003;57:305–309. Ravasco P, Monteiro-Grillo I, Camilo ME. Does nutrition influence quality of life in cancer patients undergoing radiotherapy? Radiother Oncol 2003;67:213–220. Klein S, Simes J, Blackburn GL. Total parenteral nutrition and cancer clinical trials. Cancer 1986;58:1378– 1386. Moreschi C. BezIehung zwishcen ernahrung und tumorwachstum. Z Imunitatsforsch 1909;2:651–675. Grossie VB Jr, Ota DM, Ajani JA, et al. Influence of total parenteral nutrition on tumor growth and polyamine biosynthesis of fibrosarcoma-bearing rats after induced cachexia. J Parenteral Enteral Nutr 1998;12: 441–444. Kritchevsky D. Caloric restriction and cancer. J Nutr Sci Vitamino 2001;47:13–10. Terosian MH, Daly JM. Nutritional support in the cancer-bearing host: effects on host and tumor. Cancer 1986;58:1915–1929. Popp MB, Wagner SC, Brito OJ. Host and tumor responses to increasing levels of intravenous nutritional support. Surgery 1983;94:300–308. Popp MB, Kirkemo AK, Morrison SD, et al. Tumor and host carcass changes during total parenteral nutrition in an anorectic rat-tumor system. Ann Surg 1984;199:205– 210. Chance WT, Cao LQ, Fischer JE. Response of tumor and host to hyperalimentation and antiglutamine treatments. J Parenteral Nutr 1990;14:122–128. Impact of Nutrition Support 30. Ye SL, Nawfal WI, Driscoll DF, et al. Tumor and host response to arginine and branched chain amino acid-enriched total parenteral nutrition. Cancer 1992;69:261–270. 31. Millis RM, Diya CA, Reynolds ME, et al. Growth inhibition of subcutaneously transplanted hepatomas without cachexia by alteration of the dietary arginine-methionine balance. Nutr Cancer 1998;31:49–55. 32. He YC, Wang YH, Cao J, et al. Effect of complex aminoacid imbalance on growth of tumor in tumor-bearing rats. World J Gastrenterol 2003;9:2772–2775. 33. Stallion A, Foley-Nelson T, Chance WT, et al. Parenteral vs enteral nutrition in tumor bearing rats. J Parenteral Enteral Nutr 1994;18:148–153. 34. Moller H, Mellemgaard A, Lindvig K, et al. Obesity and cancer risk: a Danish record-linkage study. Eur J Cancer 1994;30A:344–350. 35. Wolk A, Gridley G, Svensson M, et al. A prospective study of obesity and cancer risk (Sweden). Cancer Causes Control 2001;12:13–21. 36. Brown P, Allen AR. Obesity linked to some forms of cancer. W V Med J 2002;98:271–272. 37. Samlowski WE, Wiebke G, McMurray M, et al. Effects of TPN during high-dose interleukin-2 treatment for meta-static cancer. J Immunol 1998;21:65–74. 38. Baron PL, Lawrence W Jr, Chan WM, et al. Effects of parenteral nutrition on cell cycle kinetics of head and neck cancer. Arch Surg 1986;121:1282–1286. 39. Bozzetti F, Gavazzi C, Cozzaglio L, et al. Total parenteral nutrition and tumor growth in malnourished patients with gastric cancer. Tumori 1999;85:163–166. 40. Jin D, Phillips M, Byles JE. Effects of parenteral nutrition support and chemotherapy in the phasic composition of tumor cells in gastrointestinal cancer. J Parenter Enteral Nutr 1999;23:237–241. 41. Clamon GH, Feld R, Evans WK, et al. Effect of adjuvant central IV hyperalimentation on the survival and response to treatment of patients with small cell lung cancer: a randomized trial. Cancer Treat Rep 1985;69:167–177. 42. RTOG 0122. A double-blind study of nutritional intervention for the treatment of cachexia using JuvenR nutritional supplement. http://rtog.org/members/protocols/ 0122. 43. Detsky AS, McLaughlin JR, Baker JP, et al. What is subjective global assessment of nutritional status? J Parenteral Enteral Nutr 1987;11:8–13. 44. Albrecht JT, Canada TW. Cachexia and anorexia in malignancy. Hematol/Oncol Clin North Am 1996;10:791–800. HEAD & NECK—DOI 10.1002/hed April 2006