Biostatistics for Non-Statisticians
A Review by Volker Schoder
Many people working in the pharmaceutical, medical device or cosmetic industry regularly get in contact with statistical issues in their daily practice. This leads to an ongoing need to deepen knowledge in statistical methods without being overwhelmed by mathematical formalism. The aim of this seminar was to provide an extensive, yet practical overview of the most relevant statistical topics with special focus on applications in dermatology/cosmetics. The seminar was held by Volker Schoder, founder of the statistics department at proDERM and currently working as senior manager at Fresenius Medical Care, supported by Friedrich-Peter Vollmer and Sylvia Zebrowski from the statistics team at proDERM.
After a short welcome at the Klövensteen Hotel accompanied by a lunch-buffet, the seminar started with Volker Schoder presenting crucial issues in designing and analyzing questionnaires. He gave general recommendations on design as well as analytical issues, showing also many positive and negative examples. The second part of day 1 focused on statistical testing and incorporated the basic concepts of inductive statistics, namely generation of null and alternative hypotheses, type-1 and type-2 error, p-value and statistical significance, statistical power and sample size calculation.
After the coffee break, Friedrich-Peter Vollmer presented the basic features of the normal distribution and explained its features and outstanding relevance for traditional statistical methods. Following that Volker Schoder gave an overview on specific issues on randomization in dermatological/cosmetic studies based on three real-life examples. Finishing day 1, he then provided an overview of decision guidelines on statistical methods, including also a short excursion on the choice of statistical software.
In the evening of day 1, many statistical and non-statistical topics were discussed among the participants during a delicious dinner at the Klövensteen Hotel.
The next morning started with Volker Schoder explaining the omnipresent and highly underestimated problem of multiplicity, employing also the results of a questionnaire all participants had been asked to fill in the day before. He explained multiplicity as from a misuse and misinterpretation of statistical methods and also gave recommendations how to avoid this issue in practice. After that, Sylvia Zebrowski explained the special features of paired data in contrast to independent data, pointing out that paired data is commonly encountered in dermatology/cosmetics. She then presented the three most frequently used statistical tests for paired data, namely the paired t-test, the Wilcoxon-Signed rank test and the McNemar test, all on a worked example.
Following the coffee break, Volker Schoder pursued the concept of paired data, explaining the use and misuse of correlation and in the second part giving short and easily comprehensible insight into more complex situations as encountered in typical research settings. After lunch, a lecture on the concept of formally proving equivalence was presented and discussed. Remarkably, although maybe being the most complex and theoretical part of the seminar, the topic was highly anticipated by some of the participants, showing its practical relevance.
To round off the seminar and also extend the focus to the industry perspective, the final 30 minutes involved the statistician’s view on ongoing and future developments, including topics as big data, international harmonization and transparency.
The positive feedback to this seminar and also the request received from prospective participants led to the decision to repeat the seminar next year in German language.