Joseph G. Ibrahim

A Pooled Analysis of Eastern Cooperative Oncology Group and Intergroup Trials of Adjuvant High-Dose Interferon for Melanoma

Purpose: Nearly 2000 patients with stage IIB and III melanoma have participated in four multicenter, randomized trials, conducted by the Eastern Cooperative Oncology Group and the Intergroup, investigating adjuvant high-dose IFN-α 2b therapy. The objectives of this study were to update the analyses of each individual trial and to analyze prognostic factors and treatment effects based on pooled data. Experimental Design: Survival and disease status were updated to April 2001. Analysis of prognostic factors using optimized statistical models was based on data from patients in E1684, E1690, E1694, and E2696. Analysis of treatment effects versus observation (Obs) was based on data from 713 patients randomized to high-dose IFN-α 2b (HDI) or Obs in Trials E1684 and E1690. Results: Updated analysis of E1684, E1690, and E1694 confirmed their original conclusions, now at median follow-up intervals of 2.1–12.6 years. Based on two-sided univariate log-rank analysis of pooled data from E1684 and E1690 (median follow-up, 7.2 years), relapse-free survival (RFS)—but not overall survival (OS)—was significantly prolonged (two-sided log-rank P value = 0.006) for patients treated with HDI versus Obs. Among all patients, prognostic factors that significantly negatively impacted RFS and OS included ulceration, recurrent disease at entry, enrollment in E1684, and age > 49 years. Multivariate statistical models adjusting for these factors confirmed the statistically significant RFS benefit of HDI versus Obs but did not demonstrate a significant OS benefit in the pooled populations. Conclusions: In patients with high-risk resected melanoma, HDI is effective adjuvant therapy with strong evidence for improved RFS and evidence for moderate improvement in OS based on two prospective randomized studies but not the pooled analysis. Analyses of predictors of relapse and response are now needed to improve the therapeutic value of this modality.

Missing-Data Methods for Generalized Linear Models: A Comparative Review

Missing data is a major issue in many applied problems, especially in the biomedical sciences. We review four common approaches for inference in generalized linear models (GLMs) with missing covariate data: maximum likelihood (ML), multiple imputation (MI), fully Bayesian (FB), and weighted estimating equations (WEEs). There is considerable interest in how these four methodologies are related, the properties of each approach, the advantages and disadvantages of each methodology, and computational implementation. We examine data that are missing at random and nonignorable missing. For ML, we focus on techniques using the EM algorithm, and in particular, discuss the EM by the method of weights and related procedures as discussed by Ibrahim. For MI, we examine the techniques developed by Rubin. For FB, we review approaches considered by Ibrahim et al. For WEE, we focus on the techniques developed by Robins et al. We use a real dataset and a detailed simulation study to compare the four methods.

A New Bayesian Model for Survival Data with a Surviving Fraction

Abstract We consider Bayesian methods for right-censored survival data for populations with a surviving (cure) fraction. We propose a model that is quite different from the standard mixture model for cure rates. We provide a natural motivation and interpretation of the model and derive several novel properties of it. First, we show that the model has a proportional hazards structure, with the covariates depending naturally on the cure rate. Second, we derive several properties of the hazard function for the proposed model and establish mathematical relationships with the mixture model for cure rates. Prior elicitation is discussed in detail, and classes of noninformative and informative prior distributions are proposed. Several theoretical properties of the proposed priors and resulting posteriors are derived, and comparisons are made to the standard mixture model. A real dataset from a melanoma clinical trial is discussed in detail.

Incomplete Data in Generalized Linear Models

Abstract This article examines incomplete data for the class of generalized linear models, in which incompleteness is due to partially missing covariates on some observations. Under the assumption that the missing data are missing at random, it is shown that the E step of the EM algorithm for any generalized linear model can be expressed as a weighted complete data log-likelihood when the unobserved covariates are assumed to come from a discrete distribution with finite range. Expressing the E step in this manner allows for a straightforward maximization in the M step, thus leading to maximum likelihood estimates (MLEs) for the parameters. Asymptotic variances of the MLEs are also derived, and results are illustrated with two examples.

Prognostic Factors in Metastatic Melanoma: A Pooled Analysis of Eastern Cooperative Oncology Group Trials

PURPOSE To identify factors that are prognostic for survival in patients with metastatic melanoma treated in eight Eastern Cooperative Oncology Group (ECOG) trials conducted over the past 25 years. METHODS We identified common, significant patient characteristics collected at baseline on 1,362 eligible patients for inclusion in a pooled analysis. Proportional hazards models were used to examine simultaneously the effects of multiple covariates on survival. RESULTS Median survival was 6.4 months (95% confidence interval, 6.1 to 6.9 months.) Factors conferring the greatest increased risk of death included number of metastatic sites (relative risk [RR] = 1.12), ECOG performance status of 1 or more (RR = 1.49), or metastatic disease in the gastrointestinal (GI) tract (RR = 1.49), liver (RR = 1.44), pleura (RR = 1.35), or lung (RR = 1.19). Prior immunotherapy (RR = 0.84) and female sex (RR = 0. 87) were associated with prolonged survival. Although only 12% of patients responded to protocol treatment, landmark analysis showed this to be a significant prognostic factor (RR = 0.57). A model based on three recent studies in which baseline values for alkaline phosphatase, lactate dehydrogenase (LDH), and platelets were available identified an increased number of sites of metastasis (RR = 1.30), abnormal LDH (RR = 1.89), abnormal alkaline phosphatase (RR = 1.76), abnormal platelets (RR = 1.63), and GI metastases (RR = 1. 66) as prognostic for poorer survival. Response to treatment, when examined by landmark analysis of studies with laboratory parameters, was associated with decreased risk of death (RR = 0.47). CONCLUSION This study demonstrates the importance and utility of laboratory parameters as prognostic factors for survival and confirmed the deleterious effects of multiple metastatic sites. Prior immunotherapy and female sex were associated with improved prognosis. Prognostic factors identified in this analysis are consistent with the findings of prior published studies and argue for the adoption of laboratory findings in the staging systems that are used for entry and stratification of clinical trials in the future.

Phase III trial of dacarbazine versus dacarbazine with interferon alpha-2b versus dacarbazine with tamoxifen versus dacarbazine with interferon alpha-2b and tamoxifen in patients with metastatic malignant melanoma: an Eastern Cooperative Oncology Group study.

PURPOSE To investigate the response rate, time to treatment failure (TTF), overall survival, and toxicity in patients with metastatic melanoma treated with dacarbazine alone, dacarbazine plus interferon (IFN), dacarbazine plus tamoxifen (TMX), or dacarbazine plus IFN plus TMX. MATERIALS AND METHODS Two hundred seventy-one patients (258 were eligible) were randomized in a 2 x 2 factorial design to receive one of the above treatments. The trial was designed to detect a 50% improvement in survival with 83% power. RESULTS Nine complete (CRs) and 18 partial responses (PRs) were observed in the patients who received treatments that contained IFN compared with four CRs and 18 PRs in the patients who received treatments that did not contain IFN. Five CRs and 20 PRs occurred in patients treated with TMX compared with eight CRs and 16 PRs in those treated without TMX. Response differences were nonsignificant. The overall median TTF was 2.6 months, and the overall median survival was 8.9 months. There was no significant difference in TTF or survival among any of the different treatments. Poor performance status (PS), hepatic metastases, and weight loss were significant adverse prognostic factors. Twenty-three patients had a TTF greater than 20 months, and these durable responses were evenly distributed among the treatment arms. Significantly more severe and life-threatening toxic events occurred with treatments that contained IFN. CONCLUSION Neither IFN, TMX, nor the combination significantly improved the response rate, TTF, or survival when added to dacarbazine, but IFN significantly increased toxicity.

Estimating Cure Rates From Survival Data: An Alternative to Two-Component Mixture Models.

This article considers the utility of the bounded cumulative hazard model in cure rate estimation, which is an appealing alternative to the widely used two-component mixture model. This approach has the following distinct advantages: (1) It allows for a natural way to extend the proportional hazards regression model, leading to a wide class of extended hazard regression models. (2) In some settings the model can be interpreted in terms of biologically meaningful parameters. (3) The model structure is particularly suitable for semiparametric and Bayesian methods of statistical inference. Notwithstanding the fact that the model has been around for less than a decade, a large body of theoretical results and applications has been reported to date. This review article is intended to give a big picture of these modeling techniques and associated statistical problems. These issues are discussed in the context of survival data in cancer.

INK4/ARF transcript expression is associated with chromosome 9p21 variants linked to atherosclerosis.

Background Genome-wide association studies (GWAS) have linked common single nucleotide polymorphisms (SNPs) on chromosome 9p21 near the INK4/ARF (CDKN2A/B) tumor suppressor locus with risk of atherosclerotic diseases and type 2 diabetes mellitus. To explore the mechanism of this association, we investigated whether expression of proximate transcripts (p16INK4a, p15INK4b, ARF, ANRIL and MTAP) correlate with genotype of representative 9p21 SNPs. Methodology/Principal Findings We analyzed expression of 9p21 transcripts in purified peripheral blood T-cells (PBTL) from 170 healthy donors. Samples were genotyped for six selected disease-related SNPs spanning the INK4/ARF locus. Correlations among these variables were determined by univariate and multivariate analysis. Significantly reduced expression of all INK4/ARF transcripts (p15INK4b, p16INK4a, ARF and ANRIL) was found in PBTL of individuals harboring a common SNP (rs10757278) associated with increased risk of coronary artery disease, stroke and aortic aneurysm. Expression of MTAP was not influenced by rs10757278 genotype. No association of any these transcripts was noted with five other tested 9p21 SNPs. Conclusions/Significance Genotypes of rs10757278 linked to increased risk of atherosclerotic diseases are also associated with decreased expression in PBTL of the INK4/ARF locus, which encodes three related anti-proliferative transcripts of known importance in tumor suppression and aging.

Expression of p16INK4a in peripheral blood T-cells is a biomarker of human aging

Expression of the p16INK4a tumor suppressor sharply increases with age in most mammalian tissues, and contributes to an age‐induced functional decline of certain self‐renewing compartments. These observations have suggested that p16INK4a expression could be a biomarker of mammalian aging. To translate this notion to human use, we determined p16INK4a expression in cellular fractions of human whole blood, and found highest expression in peripheral blood T‐lymphocytes (PBTL). We then measured INK4/ARF transcript expression in PBTL from two independent cohorts of healthy humans (170 donors total), and analyzed their relationship with donor characteristics. Expression of p16INK4a, but not other INK4/ARF transcripts, appeared to exponentially increase with donor chronologic age. Importantly, p16INK4a expression did not independently correlate with gender or body‐mass index, but was significantly associated with tobacco use and physical inactivity. In addition, p16INK4a expression was associated with plasma interleukin‐6 concentration, a marker of human frailty. These data suggest that p16INK4a expression in PBTL is an easily measured, peripheral blood biomarker of molecular age.

Genotype-Guided Tamoxifen Dosing Increases Active Metabolite Exposure in Women With Reduced CYP2D6 Metabolism: A Multicenter Study

PURPOSE We examined the feasibility of using CYP2D6 genotyping to determine optimal tamoxifen dose and investigated whether the key active tamoxifen metabolite, endoxifen, could be increased by genotype-guided tamoxifen dosing in patients with intermediate CYP2D6 metabolism. PATIENTS AND METHODS One hundred nineteen patients on tamoxifen 20 mg daily ≥ 4 months and not on any strong CYP2D6 inhibiting medications were assayed for CYP2D6 genotype and plasma tamoxifen metabolite concentrations. Patients found to be CYP2D6 extensive metabolizers (EM) remained on 20 mg and those found to be intermediate (IM) or poor (PM) metabolizers were increased to 40 mg daily. Eighty-nine evaluable patients had tamoxifen metabolite measurements repeated 4 months later. RESULTS As expected, the median baseline endoxifen concentration was higher in EM (34.3 ng/mL) compared with either IM (18.5 ng/mL; P = .0045) or PM (4.2 ng/mL; P < .001). When the dose was increased from 20 mg to 40 mg in IM and PM patients, the endoxifen concentration rose significantly; in IM there was a median intrapatient change from baseline of +7.6 ng/mL (-0.6 to 23.9; P < .001), and in PM there was a change of +6.1 ng/mL (2.6 to 12.5; P = .020). After the dose increase, there was no longer a significant difference in endoxifen concentrations between EM and IM patients (P = .84); however, the PM endoxifen concentration was still significantly lower. CONCLUSION This study demonstrates the feasibility of genotype-driven tamoxifen dosing and demonstrates that doubling the tamoxifen dose can increase endoxifen concentrations in IM and PM patients.