A study design that would provide best evidence for relating a possible cause to an adverse clinical outcome – e.g. tobacco smoking and implant bone loss?

• Animals having received dental implants are split randomly into two study arms. One group of animals is exposed to controlled levels of tobacco smoke. If these loose more bone than the others, there is likely a causal relationship.

• The conduct of an experiment of this nature on humans is today unthinkable for ethical reasons, grounded in the Declaration of Helsinki.

• There are long traditions in medical research to use animal experimentation to appraise possible harm of drugs, medical devices and interventions. Some are more or less controversial because of lack of biological similarity with humans.

• Causality is therefore often based on data from non-experimental (i.e. observational) clinical studies.