Friday, January 13, 2012

Checklist for reviewing pilot studies

Pilot Study Checklist

At least one of the following reasons to conduct the study applies
·      Study administration
·      Data management
·      Scientific

Aims and objectives are clearly stated

Collected data are consistent with goals

No statistical hypothesis is tested

Sample size is justified (not necessarily in a statistical sense)

The way in which collected will be used in the design of a larger study has been addressed

This study will answer the question on whether a full scale trial/experiment is worth pursuing

Criteria that will determine continuation of a larger study are specified



Other recommendations and thoughts

Distinguish between
·      Toy problems for learning experience
·      Identifying feasibility answers
·      Getting “preliminary data” (why?)

Always identify
·      What the goal of the study is
·      How it will be determined whether the goal is met
·      What steps need to be taken to meet the goal

What is innovative or novel.

A poorly designed study can be used as an example to train investigators in designing experiments.

Example of a badly designed study

Prior to a larger clinical trial, a pilot study involving 10 patients is proposed. The goal of the project is to investigate how well patients will tolerate wearing a new ambulatory heart monitor while receiving an experimental medication. Data will be downloaded from the monitors and will be analyzed using a t-test procedure for comparison of post-treatment heart rate to pre-treatment heart rate.

·      The main goal was to assess tolerability but there is no measure defined for it. No data to address the main goal is collected.
·      There is an implicit hypothesis tested.
·      Sample size is not justified and may be driven by the desire to detect a difference.


If the goal is to estimate tolerability, where tolerability is defined as a dichotomous variable. Then the investigator needs to describe a criterion for unacceptable tolerability. The closer the intolerability is to zero, the larger the sample size needs to be.

Suggest a two tiered review. Statistician need only review protocols that have scientific merit.


References

http://www.ctspedia.org/do/view/CTSpedia/SuccessStory041

Moore CG, Carter R, Nietert PJ, Stewart PW (2011). Recommendations for planning pilot studies in clinical and translational research. Clinical and Translational Science, 4(5): 332-337.

Monday, January 9, 2012

SAS: Exporting data

In the past, I have only used PROC EXPORT to export SAS data in non-SAS format (e.g. in Excel).  For example:  

proc export data=data_out
 outfile='C:\Clients\data_out.xls'
 dbms=excel
 replace;
run;

However, to my knowledge, value labels cannot be exported.  An easy work-around is to use ODS and save the data as a .csv file: 

ODS CSV file="C:\Clients\data_out.csv";
  proc print data=data_out label noobs;
  run;
ODS CSV close;

Additional ways of exporting SAS data can be found here:

Friday, January 6, 2012

latexdiff - mark up differences between latex files

Determine and mark up significant differences between latex files.
http://www.ctan.org/tex-archive/support/latexdiff


This perl script will generate a latex file with changes marked up (similar to Word) from two latex files

latexdiff-so fileold.tex filenew.tex > dif.tex

Subscribe via email

Enter your email address:

Delivered by FeedBurner

Followers

google analytics