Interim monitoring for non-inferiority trials: minimizing patient exposure to inferior therapies.

The goal of a non-inferiority randomized trial is to demonstrate that an experimental treatment is not unacceptably worse than a standard treatment. The experimental treatment is known to have less toxicity or other quality-of-life benefits when compared with the standard treatment, so that a small decrement in efficacy would be acceptable. Interim monitoring of randomized trials is used to stop trials early if the conclusions of the trial become definitive early. In the context of a non-inferiority trial, of special interest is stopping a trial early when the experimental treatment is inferior to the standard treatment. Methods for performing interim monitoring of non-inferiority trials are reviewed for their ability to minimize patient exposure to inferior experimental treatments. Examples of trials from the literature are discussed along with a computer simulation of a simple non-inferiority monitoring rule. Interim monitoring for non-inferiority trials is shown to substantially reduce the exposure of patients to inferior therapies when, in fact, the experimental treatment is inferior to the standard treatment. Interim monitoring rules typically used in superiority trials may be sub-optimal for non-inferiority trials, and may unnecessarily expose patients to inferior therapies. Examples of trials with inferior experimental arms and trials with sub-optimal monitoring rules are given. Appropriate interim monitoring of non-inferiority trials can reduce the exposure of patients to inferior therapies when the experimental treatment is inferior to the standard treatment.

Assessing treatment efficacy in the subset of responders in a randomized clinical trial.

BACKGROUND: Durability of response is a clinically relevant dimension of the treatment effect in randomized clinical trials; it is often measured by comparing among the responders the duration of response between the treatment arms. However, since the comparison groups are defined by response (a post-randomization event), it is subject to analysis-by-responder bias, especially if the proportion of responders differs between the arms. METHODS: Two simple methods are developed that use tumor shrinkage measurements in order to lessen analysis-by-responder bias by generating more comparable patient subsets in the control and experimental arms of the trial. These subsets are then used to estimate between-arm differences in response duration. In the subtraction method, responding patients with the least tumor shrinkage in the treatment arm with more responders are removed from the patient subset for that arm. In the addition method, non-responding patients with the most tumor shrinkage in the treatment arm with fewer responders are added to the patient subset for that arm. In both methods, the numbers of patients subtracted or added are such that the proportion of patients in the modified patient subset is the same as the proportion of responders in the other treatment arm. RESULTS: The methods are demonstrated on a hypothetical dataset where they are shown to eliminate analysis-by-responder bias, and on two published analyses of randomized trials that compared the duration of response between the treatment arms. CONCLUSIONS: The proposed methods can lessen the analysis-by-responder bias. These methods to compare duration of response between treatment arms may provide a useful exploratory analysis to measure treatment efficacy among responders.

Evaluation of chemoresponse assays as predictive markers.

BACKGROUND: A chemoresponse assay that can be used to predict which patients will respond to which drugs would be useful in directing treatment. Two new analytic methods to assess the predictive ability of a chemoresponse assay have been proposed by Tian et al. METHODS: Three examples in which a hypothetical assay has no predictive ability are considered to evaluate the properties of the proposed analytic methods. RESULTS: For these specific examples, the proposed methods incorrectly suggest that the assay is predictive. CONCLUSIONS: The examples presented here demonstrate that it can be challenging to evaluate the predictive value of a chemoresponse assay.

Trends in reported incidence of primary malignant brain tumors in children in the United States.

BACKGROUND: The reported incidence of primary malignant brain tumors among children in the United States increased by 35% during the period from 1973 through 1994. The purpose of our study was twofold: 1) to determine whether the reported incidence rates for this period are better represented by a linear increase over the entire period ("linear model") or, alternatively, by a step function, with a lower rate in the years preceding 1984-1985 and a constant higher rate afterward ("jump model"); and 2) to identify the specific brain regions and histologic subtypes that have increased in incidence. METHODS: Incidence data from the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute for the period from 1973 through 1994 for primary malignant brain tumors in children were used to model the number of cases in a year as a random variable from a Poisson distribution by use of either a linear model or a jump model. RESULTS/CONCLUSIONS: The increase in reported incidence of childhood primary malignant brain tumors is best explained by the jump model, with a step increase in incidence occurring in the mid-1980s. The brain stem and the cerebrum are the primary sites for which an increase in tumor incidence has been reported. The increase in reported incidence of low-grade gliomas in the cerebrum and the brain stem (unaccompanied by an increase in mortality for these sites) supports the substantial contribution of low-grade gliomas to the overall increase in reported incidence for childhood brain tumors. IMPLICATIONS: The significantly better fit of the data to a jump model supports the hypothesis that the observed increase in incidence somehow resulted from changes in detection and/or reporting of childhood primary malignant brain tumors during the mid-1980s.

Accelerated titration designs for phase I clinical trials in oncology.

BACKGROUND: Many cancer patients in phase I clinical trials are treated at doses of chemotherapeutic agents that are below the biologically active level, thus reducing their chances for therapeutic benefit. Current phase I trials often take a long time to complete and provide little information about interpatient variability or cumulative toxicity. PURPOSE: Our objective was to develop alternative designs for phase I trials so that fewer patients are treated at subtherapeutic dose levels, trials are of reduced duration, and important information (i.e., cumulative toxicity and maximum tolerated dose) needed to plan phase II trials is obtained. METHODS: We fit a stochastic model to data from 20 phase I trials involving the study of nine different drugs. We then simulated new data from the model with the parameters estimated from the actual trials and evaluated the performance of alternative phase I designs on this simulated data. Four designs were evaluated. Design 1 was a conventional design (similar to the commonly used modified Fibonacci method) using cohorts of three to six patients, with 40% dose-step increments and no intrapatient dose escalation. Designs 2 through 4 included only one patient per cohort until one patient experienced dose-limiting toxic effects or two patients experienced grade 2 toxic effects (during their first course of treatment for designs 2 and 3 or during any course of treatment for design 4). Designs 3 and 4 used 100% dose steps during this initial accelerated phase. After the initial accelerated phase, designs 2 through 4 resorted to standard cohorts of three to six patients, with 40% dose-step increments. Designs 2 through 4 used intrapatient dose escalation if the worst toxicity is grade 0-1 in the previous course for that patient. RESULTS: Only three of the actual trials demonstrated cumulative toxic effects of the chemotherapeutic agents in patients. The average number of patients required for a phase I trial was reduced from 39.9 for design 1 to 24.4, 20.7, and 21.2 for designs 2, 3, and 4, respectively. The average number of patients who would be expected to have grade 0-1 toxicity as their worst toxicity over three cycles of treatment is 23.3 for design 1, but only 7.9, 3.9, and 4.8 for designs 2, 3, and 4, respectively. The average number of patients with grade 3 toxicity as their worst toxicity increases from 5.5 for design 1 to 6.2, 6.8, and 6.2 for designs 2, 3, and 4, respectively. The average number of patients with grade 4 toxicity as their worst toxicity increases from 1.9 for design 1 to 3.0, 4.3, and 3.2 for designs 2, 3, and 4, respectively. CONCLUSION: Accelerated titration (i.e., rapid intrapatient drug dose escalation) designs appear to effectively reduce the number of patients who are under-treated, speed the completion of phase I trials, and provide a substantial increase in the information obtained.

Second malignancies as a consequence of nucleoside analog therapy for chronic lymphoid leukemias.

PURPOSE: The nucleoside analogs fludarabine, 2'-deoxycoformycin (DCF), and 2-chlorodeoxyadenosine (CdA), commonly used in the treatment of patients with indolent lymphoid malignancies such as chronic lymphocytic leukemia (CLL) and hairy cell leukemia (HCL), are associated with myelosuppression and profound and prolonged immunosuppression. These complications raise the possibility of an increase in secondary malignancies in patients whose disease already places them at greater risk. The purpose of the present study was to assess the frequency of second tumors in patients with CLL who are treated with fludarabine and in patients with HCL who are treated with DCF and CdA. PATIENTS AND METHODS: We reviewed the long-term follow-up data for 2,014 patients treated on National Cancer Institute Group C protocols with fludarabine for relapsed and refractory CLL and with DCF and CdA for HCL using a Second Cancer Report. The numbers of observed and expected secondary tumors were compared. RESULTS: Median follow-up periods for the DCF (n = 409), fludarabine (n = 724), and CdA (n = 979) studies were 6.9, 7.4, and 5.1 years, respectively. The 111 malignancies were most commonly lymphoma (25 patients), prostate (19), lung (15), colorectal (nine), bladder (six), and breast (six), but also CNS, stomach, ovary, head and neck, melanoma, sarcoma, testicular, and myeloid leukemias. Compared with age-adjusted 1994 Surveillance and Epidemiology End-Results rates for the general population, the observed/expected frequencies for DCF, fludarabine, and CdA were 1.43 (95% confidence interval [CI], 0.93 to 2.10), 1.65 (95% CI, 1.04 to 2.47), and 1.50 (95% CI, 1.14 to 1.93), respectively, indicating a significant (at P =.05) increase in risk for patients treated on the latter two protocols compared with a normal population. However, these values are consistent with the increase already associated with these diseases. CONCLUSION: Despite their immunosuppression, nucleoside analogs can be safely administered to patients with CLL or HCL without a significantly increased risk of secondary malignancies.

Twenty-two years of phase III trials for patients with advanced non-small-cell lung cancer: sobering results.

PURPOSE: To determine the changes in clinical trials and outcomes of patients with advanced-stage non-small-cell lung cancer (NSCLC) treated on phase III randomized trials initiated in North America from 1973 to 1994. PATIENTS AND METHODS: Phase III trials for patients with advanced-stage NSCLC were identified through a search of the National Cancer Institute's Cancer Therapy Evaluation Program database from 1973 to 1994, contact with Cooperative Groups, and by literature search of MEDLINE. Patients with advanced NSCLC treated during a similar time interval were also examined in the SEER database. Trends were tested in the number of trials, in the number and sex of patients entered on the trials, and in survival over time. RESULTS: Thirty-three phase III trials were initiated between 1973 and 1994. Twenty-four trials (73%) were initiated within the first half of this period (1973 to 1983) and accounted for 5,359 (64%) of the 8,434 eligible patients. The median number of patients treated per arm of the trials rose from 77 (1973 to 1983) to 121 (1984 to 1994) (P <.001). Five trials (15%) showed a statistically significant difference in survival between treatment arms, with a median prolongation of the median survival of 2 months (range, 0.7 to 2.7 months). CONCLUSION: Analysis of past trials in North America shows that the prolongation in median survival between two arms of a randomized study was rarely in excess of 2 months. Techniques for improved use of patient resources and appropriate trial design for phase III randomized therapeutic trials with patients with advanced NSCLC need to be developed.

Secondary leukemia or myelodysplastic syndrome after treatment with epipodophyllotoxins.

PURPOSE: The incidence of secondary leukemia after epipodophyllotoxin treatment and the relationship between epipodophyllotoxin cumulative dose and risk are not well characterized. The Cancer Therapy Evaluation Program (CTEP) of the National Cancer Institute (NCI) has developed a monitoring plan to obtain reliable estimates of the risk of secondary leukemia after epipodophyllotoxin treatment. METHODS: Twelve NCI-supported cooperative group clinical trials were identified that use epipodophyllotoxins at low (<1.5 g/m2 etoposide), moderate (1.5 to 2.99 g/m2 etoposide), or higher (> or =3.0 g/m2 etoposide) cumulative doses. Cases of secondary leukemia (including treatment-related myelodysplastic syndrome) occurring on these trials have been reported to CTEP, as has duration of follow-up for all patients, thereby allowing calculation of cumulative 6-year incidence rates of secondary leukemia for each etoposide dose group. RESULTS: The calculated cumulative 6-year risks for development of secondary leukemia for the low, moderate, and higher cumulative dose groups were 3.3%, (95% upper confidence bound of 5.9%), 0.7% (95% upper confidence bound of 1.6%), and 2.2%, (95% upper confidence bound of 4.6%), respectively. CONCLUSION: Within the context of the epipodophyllotoxin cumulative dose range and schedules of administration encompassed by the monitoring plan regimens, and within the context of multiagent chemotherapy regimens that include alkylating agents, doxorubicin, and other agents, factors other than epipodophyllotoxin cumulative dose seem to be of primary importance in determining the risk of secondary leukemia. Data obtained by the CTEP secondary leukemia monitoring plan support the relative safety of using epipodophyllotoxins according to the therapeutic plans outlined in the monitored protocols.

Eligibility and response guidelines for phase II clinical trials in androgen-independent prostate cancer: recommendations from the Prostate-Specific Antigen Working Group.

PURPOSE: Prostate-specific antigen (PSA) is a glycoprotein that is found almost exclusively in normal and neoplastic prostate cells. For patients with metastatic disease, changes in PSA will often antedate changes in bone scan. Furthermore, many but not all investigators have observed an association between a decline in PSA levels of 50% or greater and survival. Since the majority of phase II clinical trials for patients with androgen-independent prostate cancer (AIPC) have used PSA as a marker, we believed it was important for investigators to agree on definitions and values for a minimum set of parameters for eligibility and PSA declines and to develop a common approach to outcome analysis and reporting. We held a consensus conference with 26 leading investigators in the field of AIPC to define these parameters. RESULT: We defined four patient groups: (1) progressive measurable disease, (2) progressive bone metastasis, (3) stable metastases and a rising PSA, and (4) rising PSA and no other evidence of metastatic disease. The purpose of determining the number of patients whose PSA level drops in a phase II trial of AIPC is to guide the selection of agents for further testing and phase III trials. We propose that investigators report at a minimum a PSA decline of at least 50% and this must be confirmed by a second PSA value 4 or more weeks later. Patients may not demonstrate clinical or radiographic evidence of disease progression during this time period. Some investigators may want to report additional measures of PSA changes (ie, 75% decline, 90% decline). Response duration and the time to PSA progression may also be important clinical end point. CONCLUSION: Through this consensus conference, we believe we have developed practical guidelines for using PSA as a measurement of outcome. Furthermore, the use of common standards is important as we determine which agents should progress to randomized trials which will use survival as an end point.

Assessing causal relationships between treatments and clinical outcomes: always read the fine print.

Changes in clinical practice should be driven by relevant and reliable evidence. Hence, adoption of a new therapy requires demonstrating that it provides (causes) benefit. Such evidence is generally obtained from intent-to-treat analyses of randomized clinical trials (RCTs). In this paper, we review other approaches to assessing the causal relationship between treatments and outcomes: (1) inference from non-randomized (observational) studies, (2) analysis of randomized studies where patients received treatments other than those to which they were randomized and (3) analysis of studies where the outcome of interest is sometimes unobservable because of a competing event (competing risks). We conclude that for the practice-changing demonstration of a favorable benefit-to-risk ratio, the gold standard is the intent-to-treat analysis of RCTs. At the same time, we illustrate how careful application of special statistical methods for assessment of treatment-outcome causation can be instrumental in complementing existing randomized evidence and guiding design of future research.

Unconditional versions of several tests commonly used in the analysis of contingency tables.

By focusing on a confidence interval for a nuisance parameter, Berger and Boos (1994, Journal of the American Statistical Association 89, 1012-1016) proposed new unconditional tests. In particular, they showed that, for a 2 x 2 table, this procedure generally was more powerful than Fisher's exact test. This paper utilizes and extends their approach to obtain unconditional tests for combining several 2 x 2 tables and testing for trend and homogeneity in a 2 x K table. The unconditional procedures are compared to the conditional ones by reanalyzing some published biomedical data.

On power and efficiency robust linkage tests for affected sibs.

For diseases that do not follow a clear Mendelian pattern of inheritance nonparametric tests applied to affected sibs have been shown to be robust to the inherent uncertainty about the precise underlying genetic model. It is known that the weights optimizing the power of tests using IBD alleles shared by affected sib pairs or triples depend on the underlying model. We show how efficiency robustness techniques, used in other areas of statistics, provide a systematic approach for constructing a robust linear combination of the statistics that are optimal for the individual members of a family of plausible genetic models. The method depends on the correlation matrix of the optimal tests as these correlations reflect how different the models are. When the minimal correlation is less than 0.5, an alternate robust procedure is proposed. The methods apply to combining data from sibships of different sizes.

Trend tests for case-control studies of genetic markers: power, sample size and robustness.

The Cochran-Armitage trend test is commonly used as a genotype-based test for candidate gene association. Corresponding to each underlying genetic model there is a particular set of scores assigned to the genotypes that maximizes its power. When the variance of the test statistic is known, the formulas for approximate power and associated sample size are readily obtained. In practice, however, the variance of the test statistic needs to be estimated. We present formulas for the required sample size to achieve a prespecified power that account for the need to estimate the variance of the test statistic. When the underlying genetic model is unknown one can incur a substantial loss of power when a test suitable for one mode of inheritance is used where another mode is the true one. Thus, tests having good power properties relative to the optimal tests for each model are useful. These tests are called efficiency robust and we study two of them: the maximin efficiency robust test is a linear combination of the standardized optimal tests that has high efficiency and the MAX test, the maximum of the standardized optimal tests. Simulation results of the robustness of these two tests indicate that the more computationally involved MAX test is preferable.

Robust TDT-type candidate-gene association tests.

In studies of association between genetic markers and a disease, the transmission disequilibrium test (TDT) has become a standard procedure. It was introduced originally as a test for linkage in the presence of association and can be used as a test for association under appropriate assumptions. The power of the TDT test for association between a candidate gene and disease depends on the underlying genetic model and the TDT is the optimal test if the additive model holds. Related methods have been obtained for a given mode of inheritance (e.g. dominant or recessive). Quite often, however, the true model is unknown and selection of a single method of analysis is problematic, since use of a test optimal for one genetic model usually leads to a substantial loss of power if another genetic model is the true one. The general approach of efficiency robustness has suggested two types of robust procedures, which we apply to TDT-type association tests. When the plausible range of alternative models is wide (e.g. dominant through recessive) our results indicate that the maximum (MAX) of several test statistics, each of which is optimal for quite different models, has good power under all genetic models. In situations where the set of possible models can be narrowed (e.g. dominant through additive) a simple linear combination also performs well. In general, the MAX has better power properties than the TDT for the study of candidate genes when the mode of inheritance is unknown.

Data monitoring committees and interim monitoring guidelines.

Most large randomized clinical trials have a data monitoring committee that periodically examines efficacy and safety results. A typical data monitoring committee meets every 6 months, but the interim monitoring guidelines for many trials specify formal analyses that are years apart. In this article we argue that study protocols should include monitoring guidelines with formal looks at each data monitoring committee meeting. Such guidelines are shown to reduce the average duration of a trial with negligible effect on power and estimation bias. Some of the common statistical monitoring guidelines require extreme evidence to stop a trial early and do not distinguish between stopping a trial during active accrual and follow-up stages. We propose practical solutions for these issues.

Efficiency robust tests for survival or ordered categorical data.

The selection of a single method of analysis is problematic when the data could have been generated by one of several possible models. We examine the properties of two tests designed to have high power over a range of models. The first one, the maximum efficiency robust test (MERT), uses the linear combination of the optimal statistics for each model that maximizes the minimum efficiency. The second procedure, called the MX, uses the maximum of the optimal statistics. Both approaches yield efficiency robust procedures for survival analysis and ordinal categorical data. Guidelines for choosing between them are provided.

Representation of older patients in cancer treatment trials.

In 1990, the five leading causes of cancer death in men aged 65 and older were carcinomas of the lung, prostate, colon and rectum, and pancreas, and leukemia. For women in this age group, the five leading causes of cancer death were carcinomas of the lung, breast, colon and rectum, pancreas, and ovary. To determine the representation of the elderly in clinical trials, the 1992 accrual of the National Cancer Institute (NCI)-sponsored Clinical Cooperative Group treatment trials (which included more than 8000 elderly patients) for the aforementioned sites was compared with the 1990 incidence data from the NCI's Surveillance, Epidemiology, and End Results program. Of the male patients enrolled in the trials, an average of 39% were older than 65 (47.3% lung, 79.5% prostate, 47.5% colorectal, 45.6% pancreas, and 9.6% leukemia); whereas 25.9% of all women enrolled in trials were 65 or older (43.6% lung, 17.3% breast, 46.2% colorectal, 59.6% pancreas, and 35.4% ovary). With respect to incidence, older patients generally are underrepresented in cancer treatment trials. With the exception of the data on prostate cancer, each of the comparisons using the Z statistic gave probability values of less than 0.01. The most significant discrepancies between incidence and participation in cancer treatment protocols were noted for leukemia in males and breast cancer in females. Possible explanations for these findings include (1) a research focus on aggressive therapy, which may be unacceptably toxic to the elderly; (2) presence of comorbidity in the elderly; (3) fewer trials available specifically aimed at older patients; (4) limited expectations for long term benefits on the part of physicians, relatives, and the patients themselves; and (5) a lack of financial, logistic, and social support for the participation of elderly patients in clinical trials. Recognizing this situation, NCI recently sponsored a number of trials that specifically target the elderly. This paper describes the status of all major Phase II and III clinical trials that recently were closed, still are active, or now are in review that address the clinical care of this important segment of the U.S. population.

Clustering of non-major histocompatibility complex susceptibility candidate loci in human autoimmune diseases.

Human autoimmune diseases are thought to develop through a complex combination of genetic and environmental factors. Genome-wide linkage searches of autoimmune and inflammatory/immune disorders have identified a large number of non-major histocompatibility complex loci that collectively contribute to disease susceptibility. A comparison was made of the linkage results from 23 published autoimmune or immune-mediated disease genome-wide scans. Human diseases included multiple sclerosis, Crohn's disease, familial psoriasis, asthma, and type-I diabetes (IDDM). Experimental animal disease studies included murine experimental autoimmune encephalomyelitis, rat inflammatory arthritis, rat and murine IDDM, histamine sensitization, immunity to exogenous antigens, and murine lupus (systemic lupus erythematosus; SLE). A majority (approximately 65%) of the human positive linkages map nonrandomly into 18 distinct clusters. Overlapping of susceptibility loci occurs between different human immune diseases and by comparing conserved regions with experimental autoimmune/immune disease models. This nonrandom clustering supports a hypothesis that, in some cases, clinically distinct autoimmune diseases may be controlled by a common set of susceptibility genes.