Ethics and statistics

Incidence, R value, efficacy studies: since the coronavirus emerged, the Germans have been bandying about all kinds of specialist terms from medical statistics. And yet this discipline is often misunderstood.

Graphical representation of placebo studies
Graphic: pur.pur

"A lot of people think we're constantly occupied with tables and graphs," said Professor Michael Krawczak, Director of the Institute of Medical Informatics and Statistics (IMIS). However, in his words, at its core medical statistics is "a mathematical discipline that is about drawing solid conclusions from data". And so it is no wonder that Krawczak, like his fellow institute professor Astrid Dempfle, originally studied mathematics.

"Does that have any effect at all?" That is probably the most important question posed by medical statistics. No matter whether in oncology, cardiology, psychology or any other medical field, there are always new treatments whose effects and side-effects must be proven.

"Randomised double-blind studies are our most important tool for this," explained Professor Dempfle. In principle, these studies always have in common that some of the patients receive the new treatment to be tested, while the rest receive the conventional treatment or a placebo. The allocation between the two groups is randomised, i.e. based on chance. The addition of "double-blind" means that neither the study participants nor the doctors know who will receive the new drug and who will receive the comparative substance. This method aims to avoid that someone believes in the benefits of a new drug, and thus believes that they feel an improvement to their health.

However, the exact structure of a study depends on the respective scientific question to be addressed. "This starts with the number of people involved, which can range from 200 to several thousand or even tens of thousands," explained Professor Krawczak, citing a rule of thumb: "If the difference between the standard treatment or placebo and the new treatment is large, a small group is sufficient, and the smaller the difference, the bigger the number of patients required."

The importance of careful planning is shown by the purported coronavirus treatment hydroxychloroquine. Initial data showed that some people affected by the coronavirus seemed to be doing better than expected. However, solid studies on a larger scale showed that the medication has no effect. "Mathematically speaking, it is always a matter of whether we are dealing with randomness or with systematic effects," said Krawczak to describe the principle.

High ethical standards are applied to the studies. According to Krawczak and Dempfle, an absolute must is comprehensive information for the participants, who have the right to stop their participation at any time during the course of the study. Another unbreakable rule is that placebos are only administered if it is completely unclear whether a new medicinal substance helps. No one should deliberately receive a worse treatment for research purposes. At the very beginning, with the participation of the Ethics Committee at the CAU’s Faculty of Medicine, it is also determined under which circumstances a study must be terminated prematurely, and which success criteria are set.

Even if the data is correctly obtained both ethically and professionally, this cannot prevent errors from occurring sometimes in the data interpretation. "If up to 50 percent of the coronavirus patients who are given mechanical ventilation die, it could be incorrectly concluded that ventilation is highly dangerous," said Astrid Dempfle by way of example. "But the point is that without ventilation, 95 percent would probably die."

Such cases are also addressed in the teaching of medical statistics, which Professor Krawczak believes should convey "a basic scientific understanding" to young medical professionals. Because it is clear that the subject of statistics will also play a role in later careers. At the UKSH, several dozen efficacy studies are often run in parallel, partly for own innovations, and partly also on behalf of the pharmaceutical industry. Krawczak and Dempfle emphasise that both are not reprehensible, but in fact are downright necessary in order to make therapeutic advances for many diseases that are still fatal.

Author: Martin Geist

It all started with the punch card

The current Institute of Medical Informatics and Statistics (IMIS) was founded in 1964 as a university institute for medical statistics and documentation at the University Medical Center Schleswig-Holstein (UKSH) in Kiel. Initially, it was dedicated to the processing of data for the production of statistics, using punch cards at the time. In addition to medical statistics, which is the focus of Professor Michael Krawczak and his colleague Professor Astrid Dempfle, medical informatics, which had slipped into the background, once again plays a central role at the IMIS. The subject is represented in Kiel by professors Björn Bergh and Björn Schreiweis, and deals with the provision of data and algorithms in medical care and research. By now, the interaction of medical informatics and statistics has become indispensable for medical progress. (mag)