A New Palliative Prognostic Score

Vol. 17 No. 4 April 1999

Journal of Pain and Symptom Management

231

Original Article

A New Palliative Prognostic Score: A First Step for the Staging of Terminally Ill Cancer Patients Marco Pirovano, MD, Marco Maltoni, MD, Oriana Nanni, PhD, Mauro Marinari, MD, Monica Indelli, MD, Giovanni Zaninetta, MD, Vincenzo Petrella, MD, Sandro Barni, MD, Ernesto Zecca, MD, Emanuela Scarpi, PhD, Roberto Labianca, MD, Dino Amadori, MD, and Gino Luporini, MD, for the Italian Multicenter and Study Group on Palliative Care Divisione di Oncologia Medica (M.P., R.L., G.L.), Ospedale S. Carlo Borromeo, Milan, Italy; Divisione di Oncologia Medica (M.M., D.A.), Ospedale Pierantoni, Forlì, Italy; Unità di Biostatistica (O.N., E.S.), Istituto Oncologico Romagnolo, Forlì, Italy; Servizio di Terapia del Dolore e Cure Palliative (M.M., C.O.), Ospedale S. Leopoldo Mandic, Merate, Italy; Divisione di Oncologia Medica (M.I.), Ospedale S. Anna, Ferrara, Italy; Unità di Cure Continuative (G.Z.), Hospice Domus Salutis, Brescia, Italy; Divisione di Medicina Generale (V.P., O.N.), Ospedale Civile di Arona, Arona, Italy; Divisione di Oncologia Radioterapica (S.B., M.I.), Ospedale S. Gerardo, Monza, Italy; Divisione Terapia del Dolore e Cure Palliative (E.Z.), Istituto Nazionale Tumori, Milan, Italy

Abstract In recent years, extensive research has been performed to identify prognostic factors that predict survival in terminally ill cancer patients. This study describes the construction of a simple prognostic score based on factors identified in a prospective multicenter study of 519 patients with a median survival of 32 days. An exponential multiple regression model was adopted to evaluate the joint effect of some clinico-biological variables on survival. From an initial model containing 36 variables, a final parsimonious model was obtained by means of a backward selection procedure. The Palliative Prognostic Score (PaP Score) is based on the final model and includes the following variables: Clinical Prediction of Survival (CPS), Karnofsky Performance Status (KPS), anorexia, dyspnea, total white blood count (WBC) and lymphocyte percentage. A numerical score was given to each variable, based on the relative weight of the independent prognostic significance shown by each single category in the multivariate analysis. The sum of the single scores gives the overall PaP Score for each patient and was used to subdivide the study population into three groups, each with a different probability of survival at 30 days: (1) group A: probability of survival at 30 days .70%, with patient score #5.5; (2) group B: probability of survival at 30 days 30–70%, with patient score 5.6–11.0; and (3) group C: probability of survival at 30 days ,30%, with patient score .11.0. Using this method, 178/519 (34.3%) patients were classified in risk group A, 205 (39.5%) patients were in risk group B, and 136 (26.2%) patients were in risk group C. The patients

Address reprint requests to: Marco Maltoni, MD, Divisione di Oncologia Medica, Ospedale Pierantoni, Via C. Forlanini n. 34, 47100 Forlì, Italy. Accepted for publication: July 29, 1998. © U.S. Cancer Pain Relief Committee, 1999 Published by Elsevier, New York, New York

0885-3924/99/$–see front matter PII S0885-3924(98)00145-6

232

Pirovano et al.

Vol. 17 No. 4 April 1999

classified in the three risk groups had a very different survival experience (logrank 5 294.8, P , 0.001), with a median survival of 64 days for group A, 32 days for group B, and 11 days for group C. The PaP Score based on simple clinical and biohumoral variables proved to be statistically significant in a multivariate analysis. The score is valid in this population (training set). An independent validation on another patient series (testing set) is required and is the object of a companion paper. J. Pain Symptom Manage 1999;17:231–239. © U.S. Cancer Relief Committee, 1999. Key Words Advanced cancer, prognostic factors, palliative care, predictors of survival

Introduction The classification of patients with very advanced cancer into homogeneous prognostic classes would undoubtedly lead to an improvement in therapeutic and care strategies, at the same time minimizing risks of undertreatment or overtreatment. This stratification would also constitute a first step towards an effective staging of terminally ill cancer patients, thus enabling the matching and evaluation of homogeneous and reliable data. In the past, several works have pointed out the high bias in predicting prognoses, which very often tend to overestimate expected survival.1–3 More recent work seems to suggest that assessment of patient experience may reduce the number of incorrect predictions.4–6 Specifically, some authors have highlighted the prognostic importance of factors linked to quality of life and to psychological and social status.7 However, factors associated with the disease would seem to be of greatest significance.8,9 The most widely used performance status (PS) scores (Karnofsky and Eastern Cooperative Oncology Group [ECOG]) have proved to be reliable prognostic parameters mainly in the middle to low range of scores.10,11 Some authors have suggested that the prognostic capacity of the PS could be increased by the integration of some clinical symptoms.12,13 Biohumoral prognostic factors, on the other hand, have been investigated to a lesser degree, even though some have confirmed predictive value.14–23 Prognostic scores have been described for postsurgical,24,25 advanced,26 and geriatric27 cancer patients; moreover, “intensity” and “acuity” scores have been elaborated for different stages of the disease.28,29 Our group planned30 a prospective multicenter study on more than 500 terminally ill cancer patients (training set)

to identify clinical13 and biological23 prognostic factors. The main objective of the present paper was to integrate them into a score. The score has been validated on a completely independent prospective population of consecutive patients (testing set) in a companion paper.31

Methods From October 1992 to November 1993, 540 consecutive eligible patients from 22 Italian centers were enrolled in the study. Information relating to at least two of the variables under investigation was missing for 21 patients, who were excluded from the study. The analysis was thus conducted on 519 eligible patients. Some centers only took part in the project for a limited period. More details on the methods and on the study sample, including the results of all univariate analyses, can be found in previous papers.13,23 All patients had advanced solid tumors which were no longer considered suitable for primary treatment. Palliative radiotherapy and anabolic hormonal treatment were allowed, whereas myelomas, renal tumors, and hematological neoplasms were excluded because of the possible interference with some blood values. On study entry, the following information was gathered: personal data and data relative to the patient’s state of illness; data relating to the terminal phase from a clinical viewpoint, such as prediction of survival, performance status, and symptoms present; and data relating to certain biologic parameters (Table 1). All the information was recorded by the physician on a form specifically designed for the study. The Clinical Prediction of Survival (CPS) (stated in weeks of life expectancy) was based on the clinical experience of the physician. Patient

Vol. 17 No. 4 April 1999

PaP Score for Prognosis in Terminally Ill Cancer Patients

Table 1 Clinical and Biological Parameters Evaluated Clinical parameters Sex Age Primary site of neoplasia Metastatic sites:

Locally advanced disease Viscera Bone Soft tissue Central nervous system Palliative hormonal treatment: Progestinal Corticosteroid Karnofsky Performance Status Clinical Prediction of Survival (wk) Hospitalization Blood transfusion in the last 15 days Weight loss Symptomatology: Fever Anorexia Dry mouth Dysphagia Dyspnea Pain Analgesic treatment Biological parameters Serum albumin level Serum prealbumin level Proteinuria 24 hr Hemoglobin Transport iron Pseudocholinesterase Transferrin Total WBC Leukocyte count: Neutrophil percentage Lymphocyte percentage Basophil percentage Monocyte percentage Eosinophil percentage

performance status was assessed according to the Karnofsky scale (KPS). As the study was multicenter, ranges of normal values for the biological parameters varied among different laboratories. Therefore, conversion formulas which permitted standardization were used. Normal, altered, and very altered ranges of values were established “a priori” for each parameter. Median values within the altered value group were chosen as cutoff to distinguish between “altered values” and “very altered values.” The biological prognostic factors of statistical relevance were classified as follows: • White blood count (WBC). The WBC value was considered normal between 4800 and 8500 cells/mm3. Leukocytosis was defined as mild when the total WBC was .8500 cells/

233

mm3, but #11,000 cells/mm3 (this is equivalent to a #30% alteration of normal values) and severe when the total WBC was found to be over 11,000 cells/mm3 (.30% alteration of normal values) . • Lymphocyte percentage. The lymphocyte percentage was considered normal between 20% and 40% of the total WBC and low for values ,20% but $12% (this is equivalent to a #40% alteration of normal values). The lymphocyte percentage was considered very low for values ,12% (.40% alteration of normal values). • Pseudocholinesterase (CHE). The CHE value was considered normal between 2750 and 3800 UI/lt and low between 2450 and 2749 UI/lt (this is equivalent to a #10% alteration of normal values); a value #2449 UI/lt was considered very low (.10% alteration of normal values). CHE is a liver a-glycoprotein widely assessed in Europe, and is usually assumed to be an index of hepatic synthetic activity. The PaP Score was constructed by combining two previous analyses: one on the clinical parameters and the other on the laboratory results. When clinical parameters alone were evaluated by multivariate analysis, the following were found to have an independent prognostic effect: CPS, anorexia, dyspnea, palliative steroid treatment, KPS, hospitalization.13 The second multivariate analysis identified leukocytosis, lymphocytopenia, and pseudocholinesterase levels23 as significant and independent predictors of survival. The differing prognostic results from the two multivariate analyses were then inserted into a multiple regression model which finally produced an integrated prognostic score (PaP Score).

Statistical Analysis Survival times were measured from the date of enrollment in the study, and death from all causes was taken as outcome. Survival curves were traced by the Kaplan-Meier method and comparison of survival curves were based on log-rank test.32 As described in our previous papers,13,23 the death rate can be considered as constant over time. This assumption was checked by an empirical point of view. In our series, the plot of

234

Pirovano et al.

ln{2ln[S(t)]} [where S(t) is the Kaplan-Meier estimate of the survival curves] against ln time for each level of the factors studied approximated a straight line through the origin. Such a plot suggests that the parametric exponential model might be valid to evaluate the joint effect of patient variables measured at the time of admission.33 For this model the cumulative survival probability of the jth patient at time t was: S ( t, x j ) = exp ( – r j t ) where r j = exp [ – ( β 0 + β 1 x 1 j + … + β p x pj ) ] This model allowed us to consider several variables simultaneously. Their effect on survival was investigated by the linear predictor: g ( x j ) = β 0 + β 1 x lj + … + β p x pj where x ij ( i = 1, 2, . . . , p ) ( j = 1, 2, . . . , n ) was the value assumed by each of the p variables and bi was the pertinent regression coefficient. As the prognostic variables were categorical, one or more dummies were built for each of them.34 The maximum likelihood method was used to estimate the regression coefficients (bi). In this way, exp (bi) can be interpreted in terms of hazard of death from a given category relating to that of the reference category. For each biological variable, the range that categorized normal values was chosen as the reference category; for each clinical variable, the best survival was chosen as reference. Furthermore, it may be shown that the ratio of the estimate bi to its standard error [SE(bi), i.e. z 5 bi/SE(bi)] is approximatively distributed as a Gaussian standardized random variable. Therefore, this statistic is used to test the null hypothesis: H0: bi 5 0. From an initial model containing all clinical and biological factors, which resulted independently in the two previous analyses listed in Table 2,13,23 a final parsimonious model was obtained by means of a backward selection procedure. We assumed that the contemporaneous inclusion of such variables in the same model is possible without any concern for overlap. The quantity g(xj) was used to compute a score suitable for classifying each patient in groups with different prognoses at a given time. The choice of time and number of groups with

Vol. 17 No. 4 April 1999

Table 2 Main Clinical–Biological Characteristics of 519 Patients Variables KPS $50 30–40 10–20 CPS week .12 11–12 9–10 7–8 5–6 3–4 1–2 missing Anorexia no yes Dyspnea no yes Palliative steroid treatment no yes Palliative progestinic treatment no yes Hospitalization no yes Blood transfusion in the last 15 days no yes Pseudocholinesterase Normal (2750–3800 UI/L) Low (2450–2749 UI/L) Very low (,2450 UI/L) Missing Total WBC Normal (4800–8500 cell/mm3) High (8501–11,000 cell/mm3) Very high (.11,000 cell/mm3) Lymphocyte percentage Normal (20–40%) Low (12–19.9%) Very low (0–11.9%)

No. of patients

%

248 217 54

47.8 41.8 10.4

69 51 41 77 74 109 81 17

13.3 9.8 7.9 14.8 14.3 21.0 15.6 3.3

191 328

36.8 63.2

340 179

65.5 34.5

185 334

35.6 64.4

484 35

93.3 6.7

380 139

73.2 26.8

464 55

89.4 10.6

150 146 160 63

28.9 28.1 30.8 12.1

256 120 143

49.3 23.1 27.6

150 198 171

28.9 38.2 32.9

different prognoses was made in terms of clinical considerations, such as t 5 30 days. The three groups had the following 30 day-survivals: .70% (group A), 30–70% (group B), ,30% (group C). Finally, in order to obtain an easy-to-handle score for the prognostic factors retained in the final integrated model, the value of each regression coefficient was divided by the smallest regression coefficient and the results were rounded to the nearest integer, or to the nearest integer 10.5. The total score for a given patient was ob-

Vol. 17 No. 4 April 1999

PaP Score for Prognosis in Terminally Ill Cancer Patients

tained by adding together his appropriate partial scores. All analyses were carried out using SAS Software.35

235

0.05. Table 3 reports the maximum likelihood estimate of the regression coefficient, its standard error, and Wald’s statistics for each regressor retained in the final model. Only KPS, CPS, dyspnea, anorexia, total WBC, and lymphocytopenia maintained independent prognostic value. Moreover, Table 3 shows the partial score value for each category, which was obtained by subdividing each regression coefficient by the smallest one as described in statistical analysis. The PaP Score for a given patient was obtained by adding together his/her partial scores. The PaP Score ranged from 0.0 (no altered variables) to 17.5 (all values with maximum alterations). When the PaP Score was #5.5, patients were classified in group A (survival at 30 days .70%); when the PaP Score was between 5.6 and 11.0, patients were classified in group B (survival at 30 days 30–70%) and patients with

Results The median survival of the 519 patients who took part in the study was 32 days (range 1– 355). Clinical and biological characteristics associated with survival are reported in Table 2. An exponential regression model was used to investigate the independent effect of each putative prognostic factor, adjusted for all factors included in the model. The integrated model contained all the clinical and biological parameters that had previously influenced prognosis in two multivariate analyses. Palliative steroid treatment, hospitalization, and pseudocholinesterase were removed because the likelihood ratio test for each of these regressors had P .

Table 3 Maximum Likelihood Estimates of Regression Coefficients, Their Standard Errors, Wald’s Statistics, Partial Scores for Categories of Prognostic Factors, and Classification of Patients in Three Risk Groups

Dyspnea no yes Anorexia no yes KPS $50 30–40 10–20 CPS (weeks) .12 11–12 9–10 7–8 5–6 3–4 1–2 Total WBC Normal (4800–8500 cell/mm3) High (8501–11,000 cell/mm3) Very high (.11,000 cell/mm3) Lymphocyte percentage Normal (20.0–40.0%) Low (12.0–19.9%) Very low (0–11.9%) Intercept Risk groups A 30-day survival probability .70% B 30-day survival probability 30–70% C 30-day survival probability ,30%

b

SE (b)

Z

p

0.00 20.19

0.10

1.94

0.05

0 1

0.00 20.25

0.10

2.54

0.01

0 1.5

0.00 0.03 20.44

0.11 0.20

0.27 2.18

0.79 0.03

0 0 2.5

0.00 20.33 20.56 20.49 20.83 21.10 21.61

0.18 0.19 0.16 0.17 0.17 0.20

1.88 2.90 3.00 4.94 6.46 7.95

0.06 0.004 0.003 ,0.001 ,0.001 ,0.001

0 2.0 2.5 2.5 4.5 6.0 8.5

0.00 20.14 20.28

0.12 0.12

1.18 2.37

0.24 0.02

0 0.5 1.5

0.00 20.19 20.49 5.42

0.11 0.13 0.21

1.65 3.72

0.10 ,0.001

0 1.0 2.5

Partial score

Total score 0–5.5 5.6–11.0 11.1–17.5

PaP Score 5 Dyspnea score 1 Anorexia score 1 KPS score 1 CPS score 1 total WBC score 1 Lymphocyte percentage score.

236

Pirovano et al.

scores .11.0 were classified in group C (survival at 30 days ,30%). For example, the PaP Score for a patient with dyspnea, KPS 5 30, CPS 5 12 weeks, total WBC normal, and normal lymphocyte percentage was: 1.0 1 0.0 1 2.0 1 0.0 1 0.0 5 3.0 (group A). The PaP Score for a patient with KPS 5 20, CPS 5 6 weeks, high total WBC, and low lymphocyte rate was 2.5 1 4.5 1 0.5 1 1.0 5 8.5 (group B). Another patient with dyspnea, anorexia, KPS 5 30, CPS 5 3 weeks, very high total WBC, and very low lymphocyte rate, had a PaP Score of 1.0 1 1.5 1 0.0 1 6.0 1 1.5 1 2.5 5 12.5 and was classified in group C. With this procedure 178/519 (34.3%) patients were classified in risk group A, 205/519 (39.5%) patients were in risk group B, and 136/519 (26.2%) patients were in risk group C. Median survival and relative 95% confidence interval (95% CI) for each group were: group A 5 64 days (95% CI 5 55–73 days), group B 5 32 days (95% CI 5 28–36 days), and group C 5 11 days (7–14 days). Figure 1 demonstrates that patients classified in the three risk groups had a very different survival experience. Their 30 days survival

Vol. 17 No. 4 April 1999

probability was 82.0% for group A, 52.7% for group B, and 9.6% for group C (logrank 5 294.8 [2 df], P , 0.001). It also shows the goodness of fitting survival curves estimated by the exponential model and those estimated by the Kaplan and Meier method.

Discussion Previously published data on prognostic factors in terminally ill cancer patients prompted our group to evaluate clinical13 and biological23 factors prospectively. The aim of this work was to insert known predictive factors into an integrated prognostic score (PaP Score), which could be used in clinical practice. Although there have been contradictory reports in the literature, there seems to be a general consensus that factors such as primary and metastatic site of disease, psychosocial factors, and other quality-of-life factors remain secondary to organic factors closely linked to the final stages of disease.7–9,12 In recent years, CPS, defined as “the ability of the physician to estimate the probable survival potentially remaining to the patient,”

Fig. 1. Survival experience of the three groups of patients identified by the PaP Score. Survival probabilities estimated by the exponential model (dotted lines) and by the Kaplan-Meier method (continuous lines). Logrank 5 294.8 (2 df), P , 0.001. Risk groups A (178 patients) B (205 patients) C (136 patients)

30-day survival probability .70% 30-day survival probability 30–70% 30-day survival probability ,30%

Total score 0–5.5 5.6–11.0 11.1–17.5

Vol. 17 No. 4 April 1999

PaP Score for Prognosis in Terminally Ill Cancer Patients

seems to have become more reliable than in the past,4–7 although it remains heavily influenced by the operator’s experience and is poorly reproducible. Considering erroneous a prediction of survival resulting in more than double or less than half the actual survival, about 30% error can be expected in expert hands. Two-thirds of this error can be ascribed to optimism and one-third to pessimism.5 Performance status is an important predictor of survival, especially for low score values;7,10,11,36,37 its predictive capacity increases when integrated with certain clinical symptoms.12,13,36 A prognostic role has also been assigned to some indices of daily life activity38,39 and some authors have highlighted the predictive capacity of cognitive functions.40 Nutritional clinicalbiological parameters, which are extremely significant in geriatric patients,17,19,27 maintain an independent prognostic value in terminally ill cancer patients, but are somewhat limited because they share a great amount of common variance.12 Parameters such as serum levels of albumin, prealbumin, total proteins, and pseudocholinesterase are either lost to univariate analysis or indeed shown as linked to reduced food intake (anorexia).13,23 The prognostic importance of some biological parameters, such as leukocytosis and lymphocytopenia, has already been observed in small case series,18,21 and was recently confirmed on larger numbers with multivariate analysis.23 Other parameters such as proteinuria could not be confirmed as prognostic factors.15,23 There was also no impact on survival from treatment with opioids;23,41 this finding might contribute to dispel current myths and prejudices concerning the use of opioids for cancer pain.42 CPS, KPS, clinical symptoms (dyspnea 1 nutritional factors), leukocytosis, and lymphocytopenia thus emerge from all available information13,23,43–45 as factors of prime importance in predicting survival. Two studies have integrated several factors into a weighed score which is capable of predicting prognosis in advanced cancer patients.28,29 The contribution of the present study consists in having identified prognostically significant parameters in a prospective multicenter study based on a reasonably high number of consecutive cases. These factors have been integrated into a score that predicts survival. In a companion paper,31 this score is validated in an independent, prospective series of consecutive cases

237

in accordance with the training/testing procedure. The PaP Score can help clinicians classify patients in prognostically homogeneous classes, which may help in planning treatment and providing patient consultation. The role of CPS and the ability of other less subjective variables to complement CPS in the prognostic assessment should be more widely investigated. The original symptom selection evaluated for the PaP Score was partly determined from previous evaluations in the literature.12 An evaluation of cognitive failure, delirium and distress has been carried out in a subgroup of patients in order to verify their possible impact as predictive variables on the PaP Score. From the results of this and other studies, it might be possible to improve the present model by including cognitive failure as a weighed predictive parameter.

Acknowledgments The co-authorship of all the following group members is gratefully acknowledged: Maria Paola Innocenti, MD (Divisione di Oncologia Medica, Ospedale Pierantoni, Forlì), Maria Vinci, MD and Gianfranco Giaccon, MD (Divisione di Oncologia Medica, Ospedale S. Carlo Borromeo, Milano), Gaetano Centrone, MD (Servizio di Terapia del Dolore e Cure Palliative, Ospedale S. Leopoldo Mandic, Merate [CO]), Raffaella Indelli, MD and Marina Marzola, MD (Divisione di Oncologia Medica, Ospedale S. Anna, Ferrara), Edmondo Terzoli, MD and Italo Cardamone, MD (Divisione di Oncologia Medica Complementare, Istituto Regina Elena, Roma), Attilio Gramazio, MD (Cattedra di Oncologia Medica, Università degli Studi di Ancona, Ancona), Massimo Luzzani, MD (Servizio Terapia Antalgica, Istituto Scientifico Tumori, Genova), Filippo De Marinis, MD (III Divisione Pneumologia, Ospedale Forlanini, Roma), Gianni Beretta, MD, Massimo Monti, MD (Divisione di Oncologia Medica, Pio Albergo Trivulzio, Milano), Augusto Caraceni, MD and Liliana Groff, MD (Divisione Terapia del Dolore e Cure Palliative, Istituto Nazionale Tumori, Milano), Ermenegildo Arnoldi, MD (Divisione di Oncologia Medica, AUSL 30, Ospedale di Trescore, Seriate [BG]), Michele Gallucci, MD (Unità di Cure Palliative e Terapia del Dolore, Ospedale di Desio [MI]), Luciano Frontini, MD and Anna Calcagno, MD (Servizio

238

Pirovano et al.

Ambulatoriale di Oncologia, Ospedale S. Paolo, Milano), Laura Piva, MD (Istituti Clinici di Perfezionamento, Milano), Costanza Calia, MD (Fondazione FARO, Ospedale S. Giovanni Antica Sede, Torino), Gregorio Moro, MD (Divisione di Radioterapia, Ospedale degli Infermi, Biella [VC]). We are indebted to the Serono Company for their assistance in the use of their material distribution and data collection network. The collaboration of Drs. Monica Taverna, Elena Verdi, Paola Trogu, and Pietro Aconito was particularly valuable. We are grateful to Ms. Grainne Tierney for the English translation of the manuscript. We also thank for their help in data collection: Laura Fabbri, Rosanna Tedaldi, Luisa Guanella, Giuseppe Agazzi, Gianluigi Testa, Giovanna Zaninetta, and Roberto Pendola.

References 1. Parkes CM. Accuracy of predictions of survival in later stages of cancer. Br Med J 1972;2:29–31. 2. Heyse-Moore DH, Johnson-Bell VE. Can doctors accurately predict the life expectancy of patients with terminal cancer? Palliat Med 1987;1:165–166. 3. Forster LE, Lynn J. Predicting life span for applicants to inpatients hospice. Arch Intern Med 1988; 148:2540–2543. 4. Maltoni M, Nanni O, Derni S, et al. Clinical prediction of survival is more accurate than the Karnofsky Performance Status in estimating life span of terminallyill cancer patients. Eur J Cancer 1994;30A:764–766. 5. Hardy JR, Turner R, Saunders M, et al. Prediction of survival in a hospital-based continuing care unit. Eur J Cancer 1994;30A:284–288. 6. Rosenthal MA, Gebsky VJ, Kefford RF, et al. Prediction of life expectancy in hospice patients: identification of novel prognostic factors. Palliat Med 1993;7:199–204. 7. Tamburini M, Brunelli C, Rosso S, et al. Prognostic value of quality of life scores in terminal cancer patients. J Pain Symptom Manage 1996;1:32–41. 8. Addington-Hall JM, Mac Donald LD, Anderson HR. Can the Spitzer Quality of Life Index help to reduce prognostic uncertainly in terminal care? Br J Cancer 1990;62:695–699. 9. Cassileth BR, Lusk EJ, Miller DS, et al. Psychosocial correlates of survival in advanced malignant disease? N Engl J Med 1985;312:1551–1555. 10. Yates JW, Chalmer B, Mc Kegner FP. Evaluation of patients with advanced cancer using the Karnofsky Performance Status. Cancer 1980;45:2220–2224. 11. Miller FJ. Predicting survival in the advanced cancer patient. Henry Ford Hosp Med 1991;391:81–84.

Vol. 17 No. 4 April 1999

12. Reuben DB, Mor V, Hiris J. Clinical symptoms and length of survival in patients with terminal cancer. Arch Intern Med 1988;148:1586–1591. 13. Maltoni M, Pirovano M, Scarpi E, et al. Prediction of survival of patients terminally ill with cancer. Results of an Italian Prospective Multicentric Study. Cancer 1995;75:2613–2622. 14. Braga M, Gianotti L, Radaelli G, et al. Evaluation of the predictive performance of nutritional indicators by receiver–operating characteristic curve analysis. J Parenter Enter Nutr 1991;15:619–624. 15. Sawyer N, Wadsworth J, Wijnen M, et al. Prevalence, concentration and prognostic importance of proteinuria in patients with malignancies. Br Med J 1988;296:1295–1298. 16. Herrmann FR, Safran C, Levkoff SE, et al. Serum albumin level on admission as a predictor of death, length of stay and readmission. Arch Intern Med 1992;152:125–130. 17. Fulop T, Herrmann F, Rapin CH. Prognostic role of serum albumin and prealbumin levels in elderly patients at administration to a geriatric hospital. Arch Geront Geriatr 1991;12:31–39. 18. Shoenfeld Y, Tal A, Berliner S, et al. Leukocytosis in non hematological malignancies. A possible tumor-associated marker. J Cancer Res Clin Oncol 1986;111:54–58. 19. Romagnoli A, Rapin CH. Valeur prognostique de certains paramètres biologiques chez des sujets agés hospitalisés. Age Nutrition 1991;2:130–136. 20. Cohen MH, Makuch R, Johnston-Early A, et al. Laboratory parameters as an alternative to performance status in prognostic stratification of patients with small cell lung cancer. Cancer Treat Rep 1981; 65:187–195. 21. Ventafridda V, De Conno F, Saita L, et al. Leukocyte–lymphocyte ratio as prognostic indicator of survival in cachectic cancer patients. Ann Oncol 1991;2:196. 22. Ralston ST, Gallacher SJ, Patel U, et al. Cancer associated hypercalcemia: morbidity and mortality clinical experience in 126 treated patients. Ann Intern Med 1990;112:499–504. 23. Maltoni M, Pirovano M, Nanni O, et al. Biological indexes predictive of survival in 519 Italian terminally-ill cancer patients. J Pain Symptom Manage 1997;13:1–9. 24. McGuire WT, Tandon AK, Albert DC, et al. How to use prognostic factors in axillary node-negative breast cancer patients. J Natl Cancer Inst 1990;82: 1006–1015. 25. Marubini E, Bonfanti G, Bozzetti F, et al. A prognostic score for patients resected for gastric cancer. Eur J Cancer 1993;29A:845–850. 26. Graf W, Bergstrom R, Pahlman L, et al. Appraisal of a model for prediction of prognosis in advanced colo-rectal cancer. Eur J Cancer 1994;30A: 453–457.

Vol. 17 No. 4 April 1999

PaP Score for Prognosis in Terminally Ill Cancer Patients

27. Constans T, Bruyere A, Grab B, et al. PINI as mortality index in hospitalized elderly patient: research note. Int J Vitam Nutr Res 1992;62:191. 28. Headley J, Theriault R, Smith TJ. Independent validation of Apache II severity of illness score for predicting mortality in patients with breast cancer admitted to the intensive care unit. Cancer 1992;70: 497–503. 29. Strause J, Herbst J, Ryndes T, et al. A severity index designed as an indicator of acuity in palliative care. J Palliat Care 1993;9:11–15. 30. Maltoni M, Pirovano M, Nanni O, et al. Prognostic factors in terminal cancer patients. Eur J Palliat Care 1994;1:122–125. 31. Maltoni M, Nanni O, Pirovano M, et al. Successful validation of the Palliative Prognostic Score (PaP Score) in terminally ill cancer patients. J Pain Symptom Manage 17:240–247. 32. Kaplan EL, Meier P. Non parametric estimation from incomplete observations. J Am Stat Assoc 1958; 53:457–481. 33. Lawless JS. Statistical models and methods for life-time data. New York: John Wiley, 1982. 34. Marubini E, Valsecchi MG. Analysing survival data from clinical trials and observational studies. Chichester: John Wiley and Sons, 1995. 35. SAS Institute Inc. SAS/STAT User’s Guide, version 6, 4th ed., vol. 1. Cary, NC: SAS Institute, 1989; 943.

239

36. Loprinzi CL, Laurie JA, Wiesand HS, et al. Prospective evaluation of prognostic variables from patient-completed questionnaires. J Clin Oncol 1994; 12:601–607. 37. Allard P, Dionne A, Potvin D. Factors associated with length of survival among 1081 terminally ill cancer patients. J Palliat Care 1995;11:20–24. 38. Schonwetter RS, Teasdale TA, Storey P, et al. Estimation of survival time in terminal cancer an impedance to hospice admission? Hospice J 1990;6:65–79. 39. Schonwetter RS, Robinson BE, Ramirez G. Prognostic factors for survival in terminal lung cancer patients. J Gen Intern Med 1990;9:366–371. 40. Bruera E, Miller MJ, Kuehn N, et al. Estimate survival of patients admitted to a palliative care unit: a prospective study. J Pain Symptom Manage 1992;7: 82–86. 41. Ventafridda V, Ripamonti C, De Conno F, et al. Symptom prevalence and control during cancer patients’ last days of life. J Palliat Care 1990;6:7–11. 42. Wall PD. The generation of yet another myth or the use of narcotics. Pain 1997;73:121–122. 43. Lassauniere JM, Vinant P. Prognostic factors, survival and advanced cancer. J Pall Care 1992;8:52–54. 44. den Daas N. Estimating length of survival in end-stage cancer: a review of the literature. J Pain Symptom Manage 1995;10:548–555. 45. Hoy AM. Clinical pointers to prognosis in terminal disease. Dev Oncol 1987;48:79–88.