For randomized clinical and pre-clinical trials, we are familiar with CONSORT guidelines for analysis and reporting, and our statistical consultants can help you conduct and report your methods and results. We can also help at the planning stage, writing statistical methods and analysis section for grants, and conducting power analyses to determine sample size. Every randomized trial is unique, and we will consider the methods, design, and goals of each trial to select the most accurate, robust analyses for your trial. Below are some examples of methods and analyses we can do.
If you are planning a randomized trial, we can help discuss methods and analytic plans. Based on your research questions and expected results, we can help you optimize your study design and sample size as well as the proportions in each group to maximize the power your study has to detect the effect you are exploring.
To optimize the potency of an intervention, we can help you implement a multiphase strategy, where fractional factorial designs are used to screen individual intervention components to empirically select maximally effective component combinations for the randomized trial, while making the most efficient use of a small sample size1, 2, 3.
We can help with intent to treat (ITT) analyses for primary outcomes or analyses for secondary outcomes and subgroups. Example analytical techniques include ANOVA for mean comparisons between conditions at post-test and growth models to test the change from pre- to post-test, or longer follow up.
Categorical outcomes such as disease presence, or count outcomes such as number of cigarettes smoked can be analyzed using logistic regression and Poisson regression. For survival outcomes such as how long do patients live after a new heart surgery, or how long before clinically significant pain recurs with a new anti pain medication, we can use survival analysis including helping you report Kaplan-Meier curves and cox proportional hazards models.
Even with best practices, participants may drop out of a trial, particularly at follow up assessments. Dropout and missing data can bias the results. Our statistical consultants can address missing data and dropout with a variety of models including multiple imputation, full information maximum likelihood, and Bayesian analysis. Even if there is systematic bias, modern analytic techniques such as joint models and pattern mixture models can be used to recover unbiased estimates of the intervention effect in some cases. We can also conduct sensitivity analyses to tell you how different the participants who dropped out of your study would have to be to render the significant intervention effect null.
Whereas most trials randomize individual people, for some trials, randomization can occur at the level of a site or group. For example all patients from one medical center or all mice in one litter may be randomized to either treatment or control. We can easily incorporate a multilevel data structure and account for clustering within sites and groups to give you accurate, unbiased treatment estimates.
At Elkhart Group Ltd., we know that analyses are only one piece. Research is not finished until the results are reported whether that is in a scientific publication or a presentation to the board. We can help calculating effect sizes, creating polished tables, converting p-values and statistical significance into terms and numbers that stakeholders can understand. We can also create custom graphics to highlight your novel results and clearly show your take away message.
Premier: Our most bespoke offering. Let our consultants create solutions and reports that turn your data into strategic insight.
Academic: For accredited, not-for-profit universities. From methods write up assistance to advanced model analysis, we are experienced researchers who enjoy scientific discovery.
© 2013 - 2017 by Elkhart Group LLC