Paul J. Elson,2 Robert S. Witte, and Donald L. Trump liana Farber Cancer InstÃ-lale,Boston, Massachusetts [P. J. E.]; Gundersen Clinic, LaCrosse, Wisconsin ...
[CANCER RESEARCH 48. 7310-7313, December IS, 1988]
Prognostic Factors for Survival in Patients with Recurrent or Metastatic Renal Cell Carcinoma1 Paul J. Elson,2 Robert S. Witte, and Donald L. Trump liana Farber Cancer InstÃ-lale,Boston, Massachusetts [P. J. E.]; Gundersen Clinic, LaCrosse, Wisconsin [R. W. S.J; ana University of Wisconsin Clinical Cancer Center, Madison, Wisconsin ¡D.L. T.]
ABSTRACT A data base study of 610 patients with recurrent or metastatic renal cell carcinoma was conducted in order to identify clinical characteristics that are prognostic for survival in patients with this disease. Multivariate analysis identified initial Eastern Cooperative Oncology Group perform ance status (0 versus \ versus 2 versus 3), time from initial diagnosis (>I year versus - 1 year), number of metastatic sites (0,1 versus >l ). prior cytotoxic chemotherapy (no versus yes), and recent weight loss (no versus yes) as important indicators of survival. Closer examination of the resulting model indicated that patients can easily be separated into five prognostic subgroups, the subgroups being defined by a simple function of the number of risk factors present [Eastern Cooperative Oncology Croup performance status 1, recent diagnosis (- 1 year), >1 metastatic site, recent weight loss, and prior cytotoxic chemotherapy each counting as a single risk factor; and Eastern Cooperative Oncology Group per formance status 2 and 3 counting as 2 and 3 risk factors, respectively]. Median survival for each of the five risk groups was 12.8, 7.7, 5.3, 3.4, and 2.1 months, respectively.
INTRODUCTION Advanced RCC3 is generally refractory to cytotoxic chemo therapy and hormonal therapy. Literature reviews (1, 2) as well as our own experience in the ECOG (3-9) indicate that most conventional systemic therapies have little or no activity in this disease (0-15% observed response rates). Preliminary results using biologic response modifiers, such as interferons and in terleukin 2 plus lymphokine activated killer cells are encour aging and suggest activity on the order of 15% response for this class of compounds (10-14). Broader, more extensive experi ence, however, will be needed to confirm these findings. There is a clear need to continue working towards an effective treatment plan for patients with advanced RCC. Critical to this goal is the identification of patient characteristics that are likely to influence outcome, i.e., prognostic factors. Knowledge of important prognostic factors is useful in the management of individual patients and is an important consideration in the design and analysis of clinical trials in this disease. A primary end point of most therapeutic clinical trials is survival; therefore, the present study was undertaken in an effort to identify important factors related to survival in patients with this disease. MATERIALS AND METHODS All patients entered into ECOG groupwide Phase II protocols for advanced RCC between 1975 and 1984 were considered for inclusion Received 3/7/88; revised 7/1/88, 9/6/88; accepted 9/16/88. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1This study was conducted by the Eastern Cooperative Oncology Group (Paul P. Carbone, M.D., Chairman, ÇA21115) and supported by USPHS grants from the National Cancer Institute. NIH. and the Department of Health and Human Services. : To whom requests for reprints should be addressed, at Division of Biostatistics, Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 021 IS. 3 The abbreviations used are: RCC, renal cell carcinoma; ECOG, Eastern Cooperative Oncology Group; TS, training sample; VS, validation sample.
in this study. During this period, 680 patients were accrued to 7 separate clinical trials and treated with 1 of 16 different chemotherapies. Table 1 summarizes these trials and the associated treatment regimens. The trials analyzed in this report were open to patients with a histologically confirmed diagnosis of recurrent or metastatic RCC. All patients were required to have a measurable lesion to serve as an indicator of response. Contraindications to eligibility included: ECOG performance status 4 (15) (ECOG performance status >3 for EST 1882); prior treatment with a study drug; inadequate bone marrow, renal, or hepatic function; major surgery within the 3 weeks prior to study entry; exploratory surgery or biopsy within 2 weeks; and any chemotherapy or radiotherapy within 1 month unless there was disease progression and all treatment related complications had abated. Addi tional details regarding eligibility can be found in the individual study reports (4-9). Of the 680 patients initially considered in this investigation, 56 (8%) were excluded from all analyses for the following reasons: 14 because of inadequate bone marrow or renal function; 13 because of inadequate data; 10 because the patient refused the assigned treatment; 7 because no measurable disease was present; 4 because the patient's condition deteriorated rapidly and therapy could not be started; 3 because the assigned chemotherapy was unavailable; 3 because the ECOG perform ance status was 4; 1 because the diagnosis of RCC was incorrect; and 1 because concurrent radiotherapy was given. In addition, 14 patients (2%) were treated on 2 different trials and, therefore, counted twice in the data base. For the purposes of this report, baseline data obtained from these patients' second trial were ignored. Exclusing these records, the final data set consisted of 610 patients. Table 2 gives the distribution of patient characteristics that were thought to influence survival. The study population remained fairly stable throughout the period studied and there were no significant time trends in the patient characteristics examined.
STATISTICAL
METHODS
Survival was the primary end point considered in this study and was measured from the date of registration on-study to the date of death or the date last known to be alive. Of the 610 patients analyzed, 597 are known to have died and 13 had censored survival times. Table 2 summarizes the survival data for the clinical characteristics that were considered to be potentially important prognostic factors. Medians were computed using the Kaplan-Meier estimator (16) and univariate tests of significance were performed using the logrank test (17). Considering each factor independently, initial performance status, time from diagnosis to study entry, weight loss, prior nephrectomy, prior cytotoxic chemotherapy, and clinically evident or biopsy proved hepatic, osseous, or brain métastaseswere all seen to be associated with decreased survival (P < 0.10). Since many factors appear to influence survival, a Cox proportional hazards regression model, stratified by clinical trial (17, 18), was used to analyze simultaneously the effects of the factors listed in Table 2. This approach models survival, in the form of the hazard function, as
h(t; Z„Zi...,
ZPJ) =
(A)
where ho¿(t),j= 1 ... s, is the underlying baseline hazard at time t for the y'th level of the stratification variable, the Zi, ¡= l ... p, are the values of thep covariates included in the model, and the. ¡, are regression coefficients associated with the covariates and estimated from the data. A stratified model was used in order to control for any inherent survival
7310
Downloaded from cancerres.aacrjournals.org on October 18, 2017. © 1988 American Association for Cancer Research.
PROGNOSTIC
FACTORS FOR SURVIVAL IN RECURRENT OR METASTATIC RCC
Table 2 Patient characteristics
Table 1 Studies analyzed AccrualStudy Ref.EST 1874
period 1/75-12/76
Treatment arms VP4Megestrol 16 41Cyclophosphamide acetate 40Galactitol
N 40
Vinblastine + CCNU"
47
40EST 5(Generation 3877
11/77-10/78 I)
48Actinomycin Triazinate D
48EST 6(Generation 3877 51EST II)
5/79-3/80
PALA m-AMSA
49
8(Generation 1880 I)EST
2/81-7/81
PCNU
47
8(Generation 1880 II)EST
9/81-12/81
Bisantrene
33
9(Generation 1882 37EST I)
5/82-5/83
7,9(Generation 1882 5/83-7/84 II) 38CyclophosphamideTotal
Vinblastine (infusion) L-Alanosine
40
Acivicin Aminothiadiazole 46Misonidazole +
35
680"CCNU, A'-phosphona-cetyl-L-aspartic l-(2-chloroethyl)-3-cyclohexyl-I-nitrosourea; PALA, l-(2-chIoroethyl)-3-(2,6-dioxo-3-piperidyl)-l-nitrosourea.differences acid; m-AMSA, amsacrine: PCNU,
thatdo that might exist between trials, such as a time trend, model.With not fit the proportional hazards listedin the exception of initial performance status, all the factors 0being Table 2 were transformed to (0,1) indicator variables, a value of usedforcoded for the first level of the factor and a value of 1 being usedwithout the second level. The ECOG performance status score was 3had any receding; thus in the analysis a performance status of 1.The 3 times the effect of a performance status of werefirstmultivariate analysis was conducted in four stages. Patients twogroups, stratified by treatment and randomly allocated to one of wasdesigned a training sample or validation sample. The randomization tothe so that two-thirds (n = 408) of all patients were assigned model(Equation TS and one-third (n = 202) to the VS. The stratified Cox dataobtainedA) was then used to fit a multivariate model to the buildthe from the TS. A step-up regression procedure was used to partiallikelihood model; the variable producing the largest change in the log ofdecreasingwas entered first with subsequent variables added in order producea changes in the likelihood. Variables which did not inthe significant change in the likelihood (/' > 0.10) were not included anindependent model. The resulting model was then validated by refitting it to wasapplied group of patients, the VS. As a last step, the model estimatesof to the entire data set in order to obtain more precise proportionalhazards the regression coefficients. Graphical checks of the performedduring assumption and goodness of fit testing ( 18) were also analysis.RESULTSAll this stage of the
LogrankÕ ll-tr-11'llTÕsfll' rGenderMale/"i ll:ir:u li TISI u
survival /A/ \JV —¿ odillilu/
1 Till 11 ^nioj
6.0Female 0.32RaceWhite
69 31
4.9
5.8Other 0.71Age
90 10
5.0
57 43
5.6
24 44 24 8
2.0
36 64
4.6
(yr)60 0.71Mean 56.6Median 58.0Range 21-82ECOG status0performance 10.21 6.72 3.33