点击显示 收起
1 From the Departments of Agricultural, Food & Nutritional Science (JLH, LM, CJF, and WVW) and of Oncology (SMW and VEB), University of Alberta, Edmonton, Canada, and the MD Anderson Cancer Center, Houston, TX (EDB)
2 Presented in part at the Cachexia in Aging and Cancer Conference, Chicago, 4-5 December 2000. 3 Supported by Alberta Cancer Board Palliative Care Research Initiative, the Natural Sciences and Engineering Research Council of Canada, and the Canadian Institute of Health Research. 4 Reprints not available. Address correspondence to VE Baracos, Department of Oncology, University of Alberta, WW Cross Cancer Institute, 11560 University Avenue, Edmonton, AB, Canada T6G 1Z2. E-mail: vickieb{at}cancerboard.ab.ca.
ABSTRACT
Background: Severe malnutrition and wasting are considered hallmarks of advanced malignant disease, and clinical research into anorexia-cachexia therapy and nutritional support for cancer patients is ongoing. However, information on typical dietary intakes and food choices for this population is notably lacking; proposed therapies for anorexia and wasting are not framed within the context of current intake.
Objective: The objective of the study was to characterize the food intake patterns of patients with advanced cancer.
Design: Patients with advanced cancer (n = 151) recruited from a regional cancer center and palliative-care program completed a 3-d dietary record a mean (±SD) 8 ± 7 mo before death. Food items were categorized according to macronutrient content and dietary use and subsequently entered into cluster analysis.
Results: Wide variations in intakes of energy (range: 4–53 kcal · kg body wt–1 · d–1;
Key Words: Malnutrition • advanced cancer • dietary assessment • dietary patterns • cachexia • cluster analysis
INTRODUCTION
Approximately 700 000 North Americans die of cancer each year, and most of these persons experience malnutrition and wasting, which evolve continuously during the course of their disease. This cancer-associated syndrome of anorexia-cachexia (1, 2) is associated with poor prognosis, loss of functional status, and poor quality of life. One major category of anorexia-cachexia therapy is based on the concept that cancer cachexia is, at least in part, a form of malnutrition. Current intervention strategies include the provision of orexigenic agents (3-7) to promote voluntary food intake and the supplementation of specific nutrients that may be deficient, such as amino acids and n–3 polyunsaturated fatty acids (7-11).
An understanding of the food preferences and dietary habits of a population is essential to the development of effective recommendations to maintain or improve health or quality of life (12-14). Thus, it seems surprising how little is known about these features in patients in the advanced stages of cancer. Clinical studies of pharmacologic intervention and nutrient supplementation targeting anorexia and wasting in patients with advanced cancer are not framed within the context of current intake or food preferences; in fact, the typical food choices of persons with advanced cancer are not well documented. A review of the recent literature found >50 randomized clinical trials of anorexia-cachexia therapy in cancer patients (15), but our literature review, conducted for this study, found no citation regarding dietary patterns in patient populations with advanced cancer and only a few studies in which such patients completed a dietary record for the purposes of computing total caloric intake (16-20). Much information confirms the presence of weight loss and the incidence of several nutrition-related symptoms (ie, nausea, vomiting, chemosensory dysfunction, and early satiety) in patients with advanced cancer, but we suspect that the context of terminal illness is simply one in which exist many (real and perceived) barriers to in-depth investigations of nutritional status. Patients often are frail and may decline suddenly or rapidly, and much of their available energy is devoted to their immediate needs. This is unfortunate, because patients at the end of life are a unique population, and research in subjects in other stages of disease is unlikely to be generalizable to this population.
Dietary pattern analysis can provide a broad picture of food and nutrient intakes to characterize the typical eating habits of a group of persons (21, 22). Numerous studies have shown cluster analysis to be a useful method of identifying different patterns of food consumption within a population (21-29). In the current study, we used records of dietary intakes to describe food and nutrient intakes and to identify dietary patterns in a population of patients with advanced cancer. Dietary patterns were related to nutrient intakes, weight loss, and patient survival.
SUBJECTS AND METHODS
Subjects
Subjects with advanced cancer (defined as locally recurrent or metastatic; n = 151) were recruited either from the Cross Cancer Institute, a cancer treatment center serving Edmonton and northern Alberta or from the Palliative Home Care program serving Edmonton. None of the patients was currently undergoing radiation or chemotherapy. All subjects were resident in their homes and were assumed to make food choices by personal preference. Consecutive patients were screened for eligibility and, as is typical for this setting, 75% of patients were ineligible to be approached because of cognitive impairment, severe illness, or imminent death. Of those meeting criteria for inclusion, 85% agreed to participate, and all of this group completed the study evaluations.
All participants spoke English and provided written informed consent. The study was reviewed and approved by the Research Ethics Board of the Alberta Cancer Board.
Data collection
Dietary records (detailing intake for 3 consecutive days, including 1 weekend day) were used to assess subjects’ nutrient intakes and meal patterns; this method has been shown to adequately reflect current dietary intake (30, 31) and provide mean estimates of group dietary intake (32) in patients with advanced cancer. The record consisted of 6 fields to be completed each day, corresponding to 3 main meals (breakfast, lunch, and dinner) and 3 between-meal snacks (morning, afternoon, and evening). A research assistant instructed participants on completion of the food record and later reviewed the records with each participant for accuracy and completeness.
Each subject’s height and weight were measured; if a participant was unable to stand without support, the most recently recorded values were taken from the medical chart. Body mass index was calculated in kg/m2. Information regarding history of weight loss, defined as a weight loss in the 6 mo immediately preceding the study, was self-reported. Date of death was confirmed from institutional records.
Data analysis
Nutrient intakes were estimated by using the Canadian Nutrient File Database of the FOOD PROCESSOR II nutrient analysis software (version 9.0; Esha Research, Salem, OR). Analysis focused on energy and protein intakes and the percentage of total energy contributions (% of kcal) from fat, carbohydrate, and protein. Mean energy and protein intakes were expressed per person per day and per kg body wt (BW)/d.
For the dietary pattern analysis, food items were classified into 1 of 20 food categories on the basis of similarities or differences in macronutrient composition and culinary role (23-25, 33; Table 1). The food selection data were standardized by the percentage of energy contribution to total energy intake (28, 29). Because cluster analysis is sensitive to outliers, average energy contribution values for any food category that were 5 SD from the mean were assigned the next lowest value of energy contribution for that food category (24). This procedure was carried out in <0.5% of all data points.
View this table:
TABLE 1. Definition of food categories used in cluster analysis of 3-d dietary record food items1
The chosen cluster solution was validated in the following manner to ensure that the resulting dietary patterns were representative of the sample: the cluster analysis was run on the entire data set with a range of previously defined cluster numbers, randomly selected proportions of the data set, and subsets of the data set classified by sex (24, 33). The food groups that consistently defined and separated the clusters were identified. Once the 3-cluster solution was validated, one-factor analysis of variance (with Tukey-Kramer test for post hoc pairwise t tests) was used to compare the mean energy contribution from each food category, energy and nutrient intakes for selected macronutrients, and continuous clinical variables across the 3 clusters. Differences in nutrient intakes and weight loss across clusters were tested after control for total energy intake. Chi-square analysis was used to test proportional differences among clusters (sex and prevalence of weight loss). All statistical analyses were performed by using SAS for WINDOWS software (version 8.2; SAS Institute Inc, Cary NC).
RESULTS
Characteristics of the study population are shown in Table 2. Because preliminary analyses showed similar diet patterns and nutrient intakes in men and women, further analyses were not classified by sex. Dietary information was recorded, on average, at 7.8 mo before death (range: 0.5 –24 mo).
View this table:
TABLE 2. Characteristics of study population
Energy and nutrient intakes showed a striking degree of variability (Table 3). Energy intake was determined to be normally distributed; the mean energy intake was 1610 ± 686 kcal/d (range: 290–4065 kcal/d) or 25 ± 10 kcal · kg BW–1 · d–1 (range: 4–53 kcal · kg BW–1 · d–1) (Table 3). Eighty-one percent (n = 122) of the participants had an energy intake <34 kcal · kg BW–1 · d–1. Mean protein intake was 64 ± 29 g/d (range: 9–164 g/d) or 1.0 ± 0.4 g · kg BW–1 · d–1 (range: 0.2–2.7 g · kg BW–1 · d–1).
View this table:
TABLE 3. Energy and nutrient intakes in the total study population and the 3 dietary intake patterns1
The average energy contributions from the 20 food categories are shown in Table 4. In the study population as a whole, meats provided the highest proportion of energy (13%), with dessert (10%), fruit (9%), milk (9%), and white bread (8%) following in descending order. The supplement food category, consisting mostly of liquid meal replacement or enteral formulae, provided a mean of 5 ± 10% of the energy intake for the entire population of the study. In those subjects taking any supplements (n = 48), this food category provided 17 ± 12% of total energy.
View this table:
TABLE 4. Percentage energy contributions (% kcal/d) from food categories in the total study population and the 3 dietary intake patterns1
Dietary patterns identified by cluster analysis
The validation testing of the 3-cluster solution consistently identified milk, meat, and fruit as the food categories providing the greatest division among clusters (Table 4). The milk and soup pattern contributed 2–3 times the amount of energy from milk than did the other clusters (P < 0.0001). We noted frequent use of fat-reduced milk products by subjects in this group, and only 25% of those in this group chose whole milk or cream. An important feature of this pattern is that a high proportion of the total caloric intake was contributed by the smallest variety of foods: that is, >50% of calories were provided by milk, cereal, soup, and supplement.
The fruit and white bread pattern showed the highest mean energy contribution from fruit (largely from fruit juice) and white bread. The meat and potato pattern had a significantly higher intake of these food categories and of butter, margarine and added fats and other food categories. Whereas the meat category provided a large fraction of total energy (16 ± 8%), that dietary pattern had the widest variety of food energy sources overall.
Nutrient intakes by cluster
Significant differences in mean energy and nutrient intakes were evident across dietary intake patterns (Table 3). The milk and soup pattern had lower energy and protein intakes than did the meat and potato pattern (Table 3). Because the percentage of total energy derived from protein did not differ significantly between clusters, the observed dissimilarities in protein intake are associated with caloric intake, which itself is a function of dietary pattern.
Clinical variables by cluster
The history of weight loss was greatest in the dietary patterns with the lowest average nutrient intakes (Table 5). Body mass index and time to death did not differ significantly across clusters (P = 0.4963 and 0.2122, respectively). The wide range of energy intakes appeared to be unrelated to time to death (Figure 1). When the patient population was stratified by time to death, no significant relation was found between nutrient intakes or percentage of energy derived from different foods and the time to death (data not shown).
View this table:
TABLE 5. Clinical variables by dietary intake pattern1
FIGURE 1.. Energy intake in relation to time to death in 150 patients with advanced cancer.
Meal pattern analysis
Subjects were categorized according the frequency of eating episodes reported for the 3-d recording period; a significant (P < 0.0001) relation between frequency of eating and total caloric intake was observed (Figure 2). The incidence of missed meals was quite low, such that 81% of all subjects reported consumption of breakfast, lunch, and supper on all 3 study days. The variations in total energy intake and of eating frequency were largely accounted for by variations in consumption of between-meal snacks (Figure 2). Clinical variables by the number of eating episodes are shown in Table 6.
FIGURE 2.. Calorie intake by meal and number of eating episodes over 3 d in 151 patients with advanced cancer as recorded on 3-d dietary records.
View this table:
TABLE 6. Caloric intake and clinical variables by frequency of eating episodes over 3 d1
DISCUSSION
Methodologic considerations
Our data provide a unique glimpse into dietary habits near the end of life, focusing on the typical nutrient intakes and dietary patterns described by persons in the advanced stages of cancer. Although our sample size is smaller than dietary pattern analyses reported in healthy populations, we argue that this is a natural limitation in studies investigating nutritional intake near the end of life. As such, the data reported here represent a privileged access to patients’ dietary habits and nutrient intakes in the advanced stages of neoplastic disease.
The accuracy of self-reported food records has been drawn into question because obese subjects tend to overestimate or underestimate the intake of certain foods (34-36); the accuracy of the dietary records of terminally ill patients affected by wasting syndromes remains to be determined. Bruera et al (30) found good correspondence between 24-h food records and actual energy and protein intakes in patients with advanced cancer. The collection of records for an extended time may provide a better estimate of usual intake; however, given the relative frailty and vulnerability of the patients, 3 d of data collection (16-20) reflects a viable compromise.
Dietary pattern analysis has been shown to be a valid approach to dietary assessment of a population (21-29, 33); however, several limitations exist. Cluster analysis is an empirical statistical method and as such has a subjective component. The validity of the 3-cluster solution was tested to ensure that the identified dietary patterns are representative of the study sample. Furthermore, there appears to be reasonable consistency among studies describing dietary patterns (12, 23-26, 29, 33).
Our population is likely to be representative of the cancer patients included in several recent, large, clinical trials of cancer cachexia intervention that included subjects who were cognitively intact and had recent weight loss, significant self-reported anorexia although capable of oral food intake, and a life expectancy of 6–7 mo (2, 6-10, 37). By using inclusion criteria similar to those of such trials, we attempted to capture that segment of the population for whom dietary and anticachexia therapies would be initiated. Our exclusion criteria were intended to separate advanced cancer patients whose nutritional issues would be dissimilar. For example, cognitive impairment is prevalent (38), and, in the current study, proxies (family or caregiver) made the food selections. The place of food in the context of imminent death is largely governed by family and the health care team, and it was initially evaluated by using qualitative methods (39, 40)
Low energy and protein intakes with a wide degree of variation
We observed striking heterogeneity in nutrient intakes and food choices. Energy intake showed wide variation (4–53 kcal · kg BW–1 · d–1). Higher than average intakes were associated with a meat and potatoes pattern diet and an ability to eat up to 6 times/d, whereas low intakes associated with low-energy-density foods and low frequency of eating. Whereas objectively determined nutrient requirements are lacking for patients with advanced cancer, several benchmarks exist for comparison. The average resting energy expenditure of advanced cancer patients comparable to those studied here is 22.0–23.6 kcal · kg BW–1 · d–1 (16, 41, 42); by this benchmark, 40% of our patients likely had energy intakes insufficient to support basal metabolism. Even the patient subsets showing the highest intakes (ie, 27.1 kcal · kg–1 · d–1 in meat and potato pattern eaters and 30.5 kcal · kg–1 · d–1 in the most frequent eaters) showed significant weight loss. Lundholm et al (20) recently reported that weight maintenance can be achieved in a similar population consuming 34 kcal · kg–1 · d–1, which would place the minimum energy intake for weight maintenance at 1.5x resting energy expenditure, in the upper range of current recommendations (43, 44).
Dietary patterns and eating habits
Patients in our study population consumed daily main meals and, with varied frequency, between-meal snacks. The frequency of eating emerged as an important variable in total energy intake; greater total caloric intake was largely derived from the consumption of food outside of the 3 main meals of the day, which suggests the need for a supportive environment for snacking behavior and implies the utility of snacks with high nutritional value.
The sample clustered into 3 distinct dietary patterns that were determinants of energy and protein intake. A prevalent dietary pattern that emphasized meat and potatoes had a higher fat content than did other dietary patterns and a relatively even distribution of energy across food categories. Similar diet patterns have consistently been identified in healthy populations with comparable age and sex distributions (24-26, 29, 45), which may reflect a capacity to follow typical meal patterns despite illness or perhaps a determination to maintain a sense of normalcy in the domain of usual eating habits. We were able to identify groups of persons at higher risk of malnutrition according to specific nutrient intakes and prevalence of weight loss—ie, those persons following the milk and soup and the fruit and white bread dietary patterns. Food categories labeled "milk" and "fruit" have repeatedly defined diet patterns, although the energy contribution from these food categories was notably higher in our patients than has been reported in comparable healthy populations (23-25). Both of these dietary patterns featured a generally low energy density.
In healthy populations, dietary patterns have been related to factors such as ethnicity, age, and sex (46). In our sample, no differences were observed between the sexes, and because the sample was relatively homogeneous in terms of age and ethnicity (white), these factors were not assessed. Origins of the observed eating behaviors are unknown. Food choice may be motivated by a desire to control the cancer or prevent recurrence (47-49) or by the presence of symptoms such as pain, chronic nausea, chemosensory abnormalities, constipation, early satiety, fatigue, anxiety, and depression (13, 14, 50, 51). Further investigation is needed to explore these relations.
Implications for anorexia-cachexia therapy
Our subjects corresponded to study populations in several trials of cancer cachexia therapy (2, 6-10, 37). Therapeutic approaches generally involve a nutrition component (7, 9, 18-20, 36); however, attempts to increase or supplement dietary intake may not generate lean-tissue weight gain (6, 7, 52). Even weight maintenance may be hard to achieve, given that intakes of 34 kcal · kg–1 · d–1 would be necessary (20). Orexigenic therapies or supplements could not be expected to induce weight gain unless such agents were capable of raising voluntary intakes above 1.5x resting energy expenditure, which in many cases would represent a doubling or even tripling of daily intakes. These conclusions may be important to help inform expectations of nutritional supports and to limit the development of unrealistic goals.
Our results suggest that the foods consumed by patients with advanced cancer correspond largely to typical foods eaten by healthy persons. It is interesting that 70% of our subjects did not select available commercial supplements for consumption. Whereas supplements have been used in clinical trials (7, 17-19), compliance appears to be an issue, which may be related to the use of a food product not otherwise selected by the majority of these patients, to the fact that it may displace a large fraction of the intake of normal foods, or to specific features of a product (13, 14, 53). It may be helpful to acknowledge the emphasis on normal foods and to take into account the apparently limited preference for complete nutritional supplements. If a note of concern may be raised about the use of foods with particularly low energy density, it may make sense to create a platform for nutritional support based on frequently selected or preferred foods, such as those shown here.
ACKNOWLEDGMENTS
JLH and VEB were responsible for the conception and design of the study; EDB, SMW, and VEB were responsible for provision of study participants; JLH, EDB, and SMW were responsible for the collection and assembly of data; JLH, LM, CJF, WVW, and VEB were responsible for the analysis and interpretation of data; JLH, CJF, WVW, and VEB were responsible for the draft of the manuscript; and all authors gave final approval of the manuscript. None of the authors had a financial or personal conflict of interest.
REFERENCES