38 C.M. Schmidt

Secondary questions should optimally be defined from the outset as well.There should be a limited number of secondary questions to avoid significantly increasing the risk of an alpha error, i.e., finding a difference when no difference exists. In any analysis of data, statistical corrections should be performed to account for multiple comparisons, e.g., Bonferonni adjustment.

After defining the principal question, the population needs to be defined. The reference population may be all patients with a given condition, e.g., pancreatic adenocarcinoma. The available population may be all patients with pancreatic adenocarcinoma at Indiana University. The eligible population may be all patients with pancreatic adenocarcinoma at Indiana University who are aware of and eligible for the study. Finally, the study population is those eligible patients who actually enroll on study.

After defining the study population, the specific intervention (what, who, when, and how) and anticipated or expected results need to be defined. Writing the anticipated/expected results section prior to initiating any study will help to guide experiments and facilitate the study design. The most appropriate outcome variables and the best method to measure these outcome variables will in this manner be defined.

Given the anticipated results of the study, defining the next question or the follow-up question to the principal question is important. This gives the study purpose within a greater framework, i.e., determines its broader applicability. Defining the next question is also important for written or oral presentation of the study and justification of a grant for funding of the study.

Study Design

Once a feasible and relevant hypothesis has been developed, an optimal study may be designed to prove or disprove your hypothesis.

Questions to address when designing a study are identical to those addressed above in the section on reviewing the literature.

An optimal study should be constructed so it is internally and externally valid. The study should be adequately powered, free of bias, and able to be conducted in a reasonable timeframe with resources available to the investigator.

In order to determine if your study has adequate power, a power analysis needs to be performed. Power is the probability of finding a significant difference when a difference does indeed exist. A power calculation is complicated and likely should be performed in consultation with a biostatistician. Nonetheless, there are free software programs available online to familiarize one with the technique:

http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/ http://sitemaker.umich.edu/group-based/home

The essential aspects of any power analysis included:

1.Type of test (e.g., T-test)

2.Alpha error (e.g., 0.05)

3.Sample size

4.Estimated effect size

Effect size is a way to standardize measurement of a treatment effect, so it may be compared to other treatments targeted at the same outcome. Effect size also allows one to estimate whether statistically significant effects are actually relevant or meaningful. Effect size is the difference in means (of a given parameter) between treatment and control divided by the standard deviation of the control. The following is a rough guide to interpretation of the effect size:

The effect size is unknown in the design of a trial, so one must estimate the effect size. By convention, a value of 0.5 (moderate effect) is used in power calculations. When all values are entered into the power formula, a value (0–1) will result. In general, if the result is <0.8, sample size is inadequate to fully power the study.

40 C.M. Schmidt

Bias can never be fully excluded in any study, but it must be minimized as much as possible. Time to complete a study, if prolonged, may result in the introduction of significant time bias as new technologies and approaches will inevitably develop. Finally, if adequate time, personnel, and monetary resources are not available to the investigator, then the trial will likely fail to achieve its objectives.

Any research investigation initiative should start by asking the question: Is a randomized controlled trial possible? Randomized controlled trials are difficult to perform logistically in rare diseases. They are also difficult to perform where strong community bias (ethical or otherwise) exists for a particular treatment or existing data is compelling for one treatment over another. Randomized controlled trials provide the highest level of data followed by prospective cohort, retrospective case control studies, and other trials. Randomized controlled trials may actually determine cause and effect relationships, whereas other studies may only determine association. Any study less rigorous than a randomized controlled trial has significant potential for bias and confounders.

In addition to all of the above, to design an optimal study, consulting with experts in the field is essential. Individuals who have studied or practiced in the field of interest will have assimilated the relevant information on multiple occasions and may have a perspective one cannot glean from reviewing the literature on one occasion in isolation. Make sure that you have adequate resources and mentorship before venturing into any study.

Research is about discovering the truth. Revel in the mystery and joy of discovery whether or not your hypothesis is correct. If done carefully, your contribution to mankind may be immeasurable.

Acknowledgments Financial Support: Indiana Genomics Initiative (INGEN) of Indiana University. INGEN is supported in part by Lilly Endowment Inc. (CMS).

I would like to thank Dr. Carl Schmidt, MD, MSCI, Assistant Professor Surgery, the Ohio State University, for his expertise and critical review of this chapter.

Selected Readings

Friedman LM, Furberg CD, DeMets DL. Fundamentals of Clinical Trials. 4th ed. New York: Springer; 1998.

Hulley SB, Cummings SR, Browner WS, Grady DG, Newman TB. Designing Clinical Research. 3rd ed. Philadelphia: Lippincott Williams & Wilkins; 2007.

Penson DF, Wei J. Clinical Research Methods for Surgeons. Totowa: Humana Press; 2006.

Piantadosi S. Clinical Trials: A Methodological Perspective. 2nd ed. New York: Wiley; 2005.