|
||||||
| |||||||
|
|
IMPORTANCE OF WEIGHT LOSS DEFINITION IN THE PROGNOSTIC EVALUATION OF NON-SMALL CELL LUNG CANCER From the Cuneo Lung Cancer Study Group ( www.culcasg.org), at the Pulmonary Unit of "S. Croce e Carle" Hospital, Cuneo I-12100, ItalyAuthors:
ABSTRACT In both clinical practice and scientific reporting, various definitions of weight loss (WL) are currently in use. A comparison of their efficiency would be appropriate. In this report, we describe the first clinical evaluation of different WL definitions.388 new consecutive non-small cell lung cancer lung cancer patients were prospectively studied at the Pulmonary Unit of "S.Croce and Carle" Hospitals from 1995 to 1999. Multiple anthropometric, clinical and pathological data, along with 8 WL-related variables, were analysed. Patients’ length of survival was estimated by the Kaplan-Meier function and the Cox’s multivariate regression. In univariate analysis, all WL variables were prognostically significant. Among them, total WL (i.e., the percent difference between the weight at diagnosis and the last weight recorded while in good health, dichotomised by the threshold level of 11%) was the most significant factor (Log-rank: 29.65, p=0.0000). The best Cox’s model for survival prediction, constructed using all the available clinical information, included, in order of importance, the following three factors: stage of disease classification, performance status and total WL. Contrary to one may expect WL speedy was less predictive than WL quantity. Evidence from this study suggests that, while the loss of body weight is confirmed an important prognostic factor in NSCLC, the value of this factor is strongly dependent on its definition. Key words: Body weight, cancer cachexia, lung cancer, prognostic factors, pre-treatment evaluation, and prognosis INTRODUCTION Several years ago, the Veterans Administration Lung Cancer Group (VALCG) showed that patients experiencing pre-treatment weight losses of at least 10 pounds have significantly shorter survival that those with weight losses less than 10 pounds 1; 2. The Eastern Co-operative Oncology Group (ECOG) records pre-treatment weight loss (WL) as percent loss of body weight in the six months preceding treatment. Using this measure, they demonstrated that lung cancer patients, whose weight losses exceeded 5 percent in the six months preceding treatment, had significantly worse prognoses than those whose weight losses were less than 5 percent 3. More recently, hundreds of clinical investigations, trying to identify useful prognostic factors for lung cancer, were reviewed also in relationship to a possible WL effect 4. It was concluded that, while WL had not been sufficiently studied in small-cell lung cancer, evidence concerning non-small cell lung cancer (NSCLC) was enough to suggest it as a significant and independent predictor of survival 4. A few years later, official statements 5; 6 and a consensus conference report 7 recommended the WL determination as an essential part of the pre-treatment evaluation of NSCLC. Despite its clinical utility, definitions and methods to measure WL differ remarkably across studies, lacking any standard approach. Methods of WL evaluation range from simple yes/no qualitative assessment 8 to a quantitative estimate based on the percent difference between the actual and the "usual" or "ideal" weight. In such cases, the 5% 9; 10, 10% 9; 11; 12, or the 11% or more 9 thresholds are used. Following the ECOG definition 3, the WL assessment period is often limited to the six months preceding treatment 9; 13; 14. In other studies 8; 10-12; 15, no backward limit to the history of WL is set, in analogy with the classic VALCG studies 1; 2. Occasionally, the 10 pounds cut-off for dichotomization is utilised 15. Given the situation, we decided to initiate a new prospective study to determine the most clinically helpful method of WL recording. Our ultimate ambition was stimulating a process of standardisation, which we believe useful and needed. METHODS Patients' database and study design Since 1982, all new lung cancer patients, referred to the Pulmonary Medicine Unit of the "S. Croce e Carle" hospital, in the city of Cuneo, Piedmont, Italy, are managed uniformly. Description of the diagnostic and staging methods, treatment protocols and data recording system has been already reported 16. All patients, seen after January 1995, were eligible for this study, if they had a cytologically or pathologically documented diagnosis of NSCLC 17 and a specific set of clinical variables. This set included five WL-related variables, besides the following anthropometric and clinical variables: age, sex, performance status (PS) measured on the ECOG scale 18, T, N, and M descriptors, stage of disease classification, and cell type. WL-related variables were the following: 1.) The patient’s awareness of WL (PA); 2.) The body weight measured at diagnosis (WD); 3.) The weight measured at six months before diagnosis (W6); 4.) The last weight while enjoying good health, or usual weight (UW); and, 5.) The weight losing time (WLT), which was the time, needed to reach the maximum difference between UW and WD. Additional WL-related variables were then calculated from the above five. Since TNM definitions changed throughout the study period, patients’ charts were reviewed and TNM variables upgraded. This work was done as soon as the revision of the International Staging System for Lung Cancer was formalised 19. Starting from the revised T, N, and M descriptors, the 1997 stage of disease was calculated for each patient 19. Three hundred and eighty-eight patients met the eligibility criteria. Their anthropometric and clinical characteristics are shown in Table 1. WL evaluation The doctor who, first, recorded the clinical history and examined a highly suspected lung cancer patient conducted personally the WL evaluation. He posed the WL-related questions, first, to the patient and, then, separately, to the relatives. Only consistent, believable answers were accepted. Regarding WD, this was measured in the morning of a following day, while the patient was in the hospital for further diagnostic and staging tests. Patients should be in the fasted state, after voiding and without heavy clothing and shoes. Additional variables were calculated from the basic five WL variables. They included total WL (i.e., the difference between WD and UW, expressed in percent of UW), WL per month (i.e., the total WL divided by the number of months of WLT) and WL in six months (i.e., the difference between WD and W6, in percent of W6). Categorisation of both total WL and WL in six months was made using the quartile range and median of their distribution, along with the threshold levels of 15%, 12.5%, 11%, 10%, 7.5%, and 5%. Data analysis and statistical considerations Statistical analysis was performed using the SPSS package for Windows, Version 9.0 (SPSS Inc., Chicago, IL, USA). Medians and quartile ranges described continuous variables and provided the cut-off points for their categorisation, unless otherwise specified. Survivals were recorded from the time of the histologic diagnosis to death or to December 1999, when a telephone interview with the patient, the family, or the house doctor confirmed his being alive. Univariate analyses of survival were based on the Kaplan-Meier method 20. Multivariate analyses of survival were made using the Cox's proportional hazards regression model, and the stepwise backward selection procedure 21. The significance of each factor was calculated by the maximum likelihood ratio. A p value of 0.1 was set to enter, while a value of 0.15 was set to remove a variable from the model. A p value of less than 0.05 was regarded as statistically significant. All tests were two-sided. RESULTS Descriptive Statistics Table 1 shows summary statistics that describe the anthropometric and clinical characteristics of the study population. The pathological diagnosis included 142 squamous cell carcinomas, 141 adenocarcinomas, 57 large cell carcinomas, and 48 mixed histology or unclassified carcinomas. Median of the ECOG PS scores was 1. The study cohort included three in situ cancers and 100 locally limited carcinomas, plus other 285 locally advanced or metastatic tumours. As foreseeable, only a small fraction of patients could be operated (19% of the cohort). Chemotherapy was the most frequent treatment, either alone or in combination (53% of the population). In December 1999, 105 patients (27%) were still alive, after a median follow-up time of 31 weeks (range: 1-255, quartile range: 14-65). Median body weight at diagnosis was 66 kg. Two hundred eighty-eight patients had some knowledge of recent weight loss. About one third of the patients recorded a remarkable WL (equal to or greater than 10% of the UW). The median WL per month was nearly 2% of the UW. Total WL (median value: 5.71%) and WL in six months (median value: 5.6%) were similarly distributed. Univariate Survival Analysis Table 2 summarises the results of diverse univariate analyses of survival. In the first part, the four main ways to describe WL, i.e., PA, total WL, WL per month and WL in six months are considered. Since PA is, by definition, a dichotomous variable, also the other 3 continuous variables were divided into two strata (i.e., up to the median or greater than the median value). In this type of analysis, all WL definitions were prognostically significant, but with important differences of predictability. The best method of WL recording was total WL (log-rank statistic 16.46 vs. 6.84-9.08 of the other methods). Figure 1 shows the Kaplan-Meier analysis based on the quartiles of its distribution. Table 2 demonstrates that the survival predictability of total WL is greatly influenced by the cut-off level used for dichotomization. In this type of analysis, the log-rank statistic was 29.65 for the 11% threshold, 25.13 for the 5% threshold and only 18.31 for the 7.5% threshold. Multivariate analysis Table 3 summarises the results of different multivariate models. When only WL-related variables were tested, total WL alone was included in the model (overall X-squared: 25.070, p=0.0000). This observation is in line with the results of the previously described univariate analyses. Analogously, when age, sex, ECOG PS and stage of disease were tested separately with each WL variable, the model that contained total WL was the most predictive. In this type of analysis, however, the predictability of WL variables was similar, given the modest additional contribution offered by WL to the two powerful survival predictors, stage of disease and ECOG PS. Things did not change when all the WL variables and the TNM descriptors were added to the starting equation. DISCUSSION The main message of this study is that "while the loss of body weight is confirmed an important prognostic factor in NSCLC, the value of this factor is strongly dependent on its definition." Once this message is accepted and the existing variety of the WL definitions recognised, it becomes evident that a consensus agreement on this issue would be highly desirable. Evidence from this study suggests defining WL as the percent difference between the actual and the usual weight (i.e., the last weight recorded by the patient while enjoying good health) and using the cut-off level of 11%, for dichotomization. As always in clinical medicine, pre-treatment assessment of the lung cancer patient begins with a careful history and physical examination. A detailed history should include the information on the time passed from the first symptom of disease and the weight recorded at that time, while a thorough physical examination cannot omit the measurement of body weight. The above information is all is needed for WL recording in a way that, according to our data, possesses the greatest prognostic significance. Weight loss is the hallmark of cachexia. Cachexia is the clinical syndrome characterised by a variety of metabolic derangements and manifested by anorexia, weight loss, and a progressive wasting diathesis. Mechanisms mediating cancer cachexia are multiple 22. Local tumour effect, altered taste and smell, hypothalamic dysfunction, psychological factors, and antineoplastic therapy may reduce food intake. However, an adequate caloric intake does not abrogate the wasting diathesis, because of multiple alterations in the host intermediary metabolism. The consequences of cancer cachexia are considerable for the patient 23. It may provoke immunologic abnormalities of T and natural killer cell responsiveness, T cell and macrophage suppressor cell function, and lymphokine production. It may also be associated with impaired ability to tolerate antineoplastic therapy, increased treatment-related morbidity and mortality, impaired quality of life and may cause, by itself, the death in a significant number of cancer patients. In addition, dietary and pharmacologic interventions to prevent or reverse cancer-associated cachexia have shown limited or null efficacy 24. The severity of the consequences of cancer cachexia, added to the lack of effective treatment, explains why WL, the signal of cachexia, is an important prognostic factor 23. In NSCLC, experimental evidence 4, guidelines from medical societies 6, and conclusions of expert panels 5; 7 support this idea. It must be noted, however, that the experimental evidence is not univocal 4; 5. In a review of prognostic factors for lung cancer, for example, it was observed that WL was prognostically significant in 10 univariate and 11 multivariate tests of survival 4. However, it was insignificant in other three univariate and six multivariate tests 4. Even among the 10 clinical studies listed by Asamura H. and Naruke T. 5, three did not confirm WL prognostically influent. According to our evidence, a reason for this might be the use of the different WL definitions. In conclusion, we think that, while new biologic prognosticators continue to recall the attention in lung cancer arena4, the value of classic clinical indicators should not be forgotten. Among them, body weight loss remains an important predictor of prognosis and an impetus toward optimising and standardising its use should derive from the evidence herein presented. REFERENCES
Fig. 1 Survival probability based on the quartile distribution of the variable "total weight loss," defined as the percent difference between the weight at diagnosis and the usual weight in normal health conditions (we apologize for the bad quality of the graphic reproduction).
Abbreviations: ECOG= Eastern Cooperative Oncology Group; WL= weight loss. * Percent difference between the weight at diagnosis and the "usual" body weight (see text for definition). ** Total body weight change divided by the number of months of weight loosing. *** Percent body weight change in the 6 months preceding the diagnosis. ç Scale as described by Zubrod C.G. et al. (18). # Legend: E=epidermoid or squamous cell carcinoma; A=adenocarcinoma; L=large cell carcinoma; M=mixed or undefined histology (WHO classification) (17). ° Stage and TNM classifications, as defined by Clifton Mountain in his original paper (19). & Legend: S=supportive care alone; C=chemotherapy; R=radiation therapy; C+R=chemo-radiotherapy; O=operation.
Abbreviations: ECOG= Eastern Cooperative Oncology Group; WL= weight loss. * Percent difference between the weight at diagnosis and the "usual" body weight (see text for definition). ** Total body weight change divided by the number of months of weight losing. *** Percent body weight change in the 6 months preceding the diagnosis.
Abbreviations: ECOG= Eastern Cooperative Oncology Group; WL= weight loss: PFs=prognostic factors. # Stepwise forward regression (likelihood ratio), stratified by histology. * Percent difference between the weight at diagnosis and the "usual" body weight (see text for definition). ** Total body weight change divided by the number of months of weight loosing. *** Percent body weight change in the 6 months preceding the diagnosis. ° Sex, age, ECOG performance status (ECOG-PS), 1997 Stage classification | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
.htm_cmp_axis110_vbtn.gif){kind=small}
