CINJ home page link

Biometrics

Home | Services | Equipment | Price List | Location | Contact Us

Services

Research Project Design and Development: Biostatisticians are available to CINJ investigators to discuss relevant statistical considerations during all aspects of study design. The shared resource assists investigators in defining study objectives, realistic endpoints, study population, randomization schema (when appropriate), quality assurance, sample size, and power analysis required to effectively answer research questions. The biostatistician then develops the statistical methods for data analyses.

For laboratory experiments, biostatisticians help the investigator review experiments, the hypothesis to be tested or generated, the methodology to be used, and the characteristics of the data to be collected. Different methodologies for analyses may depend on whether the data collected will be discrete or continuous, clustered in time or space, and univariate or multivariate, or whether there are other variables. Recommendations are made regarding organization of data into appropriate databases. This approach is undertaken predominately for experiments involving cell lines or research animals.

Clinical investigators are required to utilize the Biometrics Shared Resource for all CINJ investigator-initiated clinical trials. During concept development, a biostatistician and the clinical investigator discuss the study objectives and endpoints. Depending on the type of study, the following information is reviewed: (1) for phase I studies, where the primary objective is to determine the maximum-tolerated dose, the dose-sequence levels, assumptions of the dose-toxicity relationships, dose escalation and/or de-escalation strategies, and statistical properties of sequential dose-adjustment methods; (2) for phase II studies, the magnitude of treatment effect that the protocol seeks to exclude and with which to power the study, and the associated type-I and type-II error rates; (3) the correlation to be drawn from assays performed on tissues, serum, or other samples with the observed response; (4) for correlative clinical and laboratory studies, appropriate regression or association analyses to be planned; (5) for all studies, patient characteristics that could affect response, survival, or other biological effects; (6) the sources of patients that are available for the study, including a realistic appraisal of sample heterogeneity and projected accrual rates and length of the study; and (7) whenever applicable, appropriate models of pharmacokinetic and pharmacodynamic data.

After reviewing the study with the principal investigator, the biostatistician writes the “Statistical Considerations Section” of the research protocol. This section includes the sample-size justification and other statistical design aspects, such as randomization and stratification (when appropriate) as well as the statistical methods to be used for data analyses. Frequently, a biostatistician from the shared resource is a co-investigator on CINJ investigator-initiated clinical studies.,. Because translational research within the Cancer Pharmacology/Developmental Therapeutics and other programs resulted in multiple early phase trials, Shih and Lin developed standard operating procedures during the current grant period to ensure a systematic approach to CINJ Phase I/II studies.

Biostatistical support of the Division of Prevention, Control and Population Science includes a wide range of needs and support. For example, biostatisticians collaborated with the Carcinogenesis and Chemoprevention Program to help with study design and data analysis for experiments involving laboratory animals. Researchers proposed to test multiple chemopreventive compounds for their synergistic effects in animal models.

Biostatisticians are also available to discuss study objectives, endpoints, correlative markers and/or surrogate endpoints used for human chemoprevention clinical studies. The following information is defined: (1) an effect size that the protocol seeks to detect; (2) the magnitude of the effect size expected in the study; (3) the control group needed for comparison of the effect; (4) the correlations with the observed response to be drawn from assays performed on tissues, serum, or other samples; appropriate regression or association analyses for correlative clinical and laboratory studies; (5) the randomization procedures and tables; (6) subject or patient characteristics that could affect time-to-disease progression or survival (if sought), or other biological effects; (7) the sources of subjects that are available for the study, including a realistic appraisal of sample heterogeneity; and (8) the duration of the study. Following this evaluation, the biostatistician estimates the sample size and projected duration needed to complete the study and derives the power calculations. This information is discussed with the investigator and a design strategy and sample-size target is then finalized.

Biometrics also provides statistical support in the design, collection and analysis of survey-data. Survey projects include those initiated from the Population Science Program’s interest in behavioral psychology, smoking cessation, and community outreach.

Development of Methodology: In some instances, consultation on project design identifies the need to develop unique analytical methods to fully capitalize on the data, to conserve resources, or to optimize experiments. An example of this need is the large quantity of gene expression data generated using DNA microarray technology, as well as RNA and proteomics data. Biometrics collaborated with investigators and jointly published several papers using microarray technology. An innovative statistical methodology paper regarding this technology entitled “ A mixture model for multiple testing in DNA microarray” was published ( Liao, Lin and Shih, Bioinformatics, 2004). Collaboration between Biometrics and the Transcriptional Profiling Shared Resource is essential.

The collaborative input of the biostatisticians and their involvement in translational trials from Cancer Pharmacology/Developmental Therapeutics led the statistician to evaluate new approaches to early clinical trials. While the traditional algorithm-based designs (such as the ‘3+3 design’) are still commonly used by many phase I investigators, as a result of collaborations with CINJ investigators, Shih and Lin developed a specific methodology for these early trials. This methodology is described in a paper entitled “Statistical Properties of the Traditional Algorithm-based Designs for Phase-I Cancer Clinical Trials” (Biostatistics, 2001). Using this method, the clinical investigators provide an estimation of toxicity for the first and the last dose levels. The statistical approach then generates three scenarios (low, median, and high) of dose-toxicity relationships using linear and logistic curves. The method uses a dose escalation and/or de-escalation scheme to calculate the maximum-tolerated dose, the predicted toxicity rate, the proportion of patients at each dose level, the proportion overall with dose-limiting toxicity, and the total expected number of patients. The statistical method of this paper has now been automated in SAS. The Bayesian Continual Reassessment Method (CRM) is a technique frequently employed for phase I cancer studies. Through a seminar provided by the shared resource, Shih introduced CRM to CINJ researchers. Shih and Lin subsequently presented a paper entitled “A Hybrid Approach of Improved CRM and EWOC for Phase I Cancer Studies” at the International Biometric Society Meeting (1999) and a manuscript has been submitted for publication.

In addition, Lin and Shih presented a methodology paper entitled “Adaptive Designs for Single-arm Phase II Cancer Clinical Trials”, which adds to the Simon’s 2-stage design, at the 2001 International Biometrics Society ENAR meeting; this paper has been published in Biometrics (2004). This adaptive design has been used in three CINJ-initiated phase II studies, including one planned for the Children’s Oncology Group.

Protocol and Grant Proposal Review: Members from Biometrics review grant proposals from basic, clinical, and population science researchers. This includes grant applications for human and non-human studies, follow-up studies, and cross-sectional surveys.

It is a CINJ policy that all investigator-initiated clinical trials must include careful statistical input and all clinical trials undergo statistical review. Biostatistical input on clinical studies includes protocols originated from CINJ investigators and CINJ Oncology Group co-investigators. All investigators involved in developing clinical trials must consult with Biometrics before submitting a full protocol to the SRB. No protocol is considered by the SRB until Biometrics has formally approved the statistical content and Shih signs off on the routing slip. Shih coordinates this effort so that frequently one biostatistician is involved in the trial design and writing the “Statistical Consideration Section” of the protocol, and a different biostatistician reviews the trial for the SRB. One study concept may require several reviews by Biometrics due to revisions in the study design made in response to SRB reviews. Biometrics also reviews industry trials for statistical validation and overall scientific merit for Scientific Review Board purposes.

Ongoing Study Monitoring: The biostatistician assigned to a specific trial periodically meets with the study investigator to provide ongoing assistance in the conduct of the trial. Interim analysis is provided as specified in the protocol. For laboratory investigators these ongoing meetings are at the request of the investigator. For clinical trials, a more defined process of ongoing monitoring has been established. The weekly OHRS meetings provide Biometrics a means to learn about the status of each approved protocol, in terms of patient recruitment, adverse events, and possible protocol violations. Through participation in the SRB, the OHRS meetings, and the HROC bimonthly meetings, the Biometrics representatives get an overview of the clinical research portfolio. In addition, biostatisticians (Moore and others) attend meetings of the Carcinogenesis and Chemoprevention subgroup on clinical chemoprevention trials, which is designed for ongoing study monitoring.

Computation and Statistical Analyses: Methods for appropriate data analyses depend on the objectives and design of the study. An outline of the data analysis plan (DAP) should be defined in the study proposal. To ensure consistently and efficiency, Biometrics established a standard operating procedure for developing the DAP. In general, to conduct and interpret the results of the statistical analysis used to test the hypotheses of a study, Biometrics undertakes the following: (1) provides a preliminary review of the data to ensure that the assumptions required for the analysis are satisfied; (2) completes all statistical analyses required to test the hypotheses considered in the study (including interim analyses when required); and (3) creates tables and/or graphs describing the demographic characteristics of the subjects in the study, the characteristics of the data, and the range of values associated with measurements that could affect the response of interest. The biostatistician involved with producing the preliminary analyses, the statistical analysis, and the tables and graphs uses SAS and/or S-plus software on an IBM/Dell PC, or the Biometrics UNIX workstation system, managed by Lin. These activities are also supported with the assistance of John Oleynick, a Master’s level data analyst.

Data Reports/Publications: The Biometrics Shared Resource offers assistance in the production of data reports. To enhance efficiency and consistency, standard operating procedures were developed for data spread sheet formats and statistical reports. In collaboration with CINJ investigators, members of this shared resource meet the specific requirements for each study or protocol and for the submission of data for presentations, audits, and publication.

Education and Training Activities: Biometrics faculty supports the education and training activities coordinated by the CINJ Associate Director of Education and Training, Edmund Lattime, who successfully applied for a Training Program in Translational Research in Cancer (T32 CA099946). Biometrics helped to develop curriculum for this training program. Biometrics also plays an important role in the ongoing educational activities for other training (e.g., with CETS Shared Resource) and mentoring grants. Additional shared resource activities include the teaching of credit-earning courses, seminar series, guest lectures, and individual educational assistance in study design and data interpretation. Biometrics faculty participate each year in CINJ’s “Annual Retreat on Cancer Research in New Jersey”, present workshops for various research programs and provide assistance for several CINJ nursing research activities as a part of their continuing education program. These education and training activities are important for communications between the biostatisticians and their collaborators as well as for the overall quality of the science this shared resource supports.