To assess the efficacy of psychiatric treatments, rating scales are essential. In this presentation, methodological and statistical problems associated with the use of rating scales in psychiatry are discussed, by taking the anxiety disorders as an example.

Not only standard symptom-specific rating scales, such as the Hamilton Anxiety Scale (HAMA), but also global measures such as the Clinical Global Impression Scale (CGI) and Quality of Life scales are used, as not only improvement of single symptoms, but also an overall increase of a patient's well-being should be the focus of a clinical trial. However, due to high placebo response rates in anxiety disorders, the use of rating scales is determined by the need for an instrument, which is sensitive enough to detect differences between active drugs and placebo (or between a certain psychological treatment and a psychological placebo). Quality of Life scales often do not detect such differences, as domains such as partnership or employment do not show rapid changes within the 8-12 weeks of a clinical study. Moreover, when using too many different scales, problems of multiple testing occur.

The outcome of a clinical study is often described by the number of patients who responded or remitted, as these endpoints are easily understood by patients. “Response” is commonly defined as a ≥ 50% reduction on these standard scales. However, this definition is arbitrary, and cut-off points should rather be founded on empirical data than on a thumb rule. The definition of “remission” on standard scale scores varies from study to study and is also very subjective. An analysis of available treatment studies in patients with anxiety disorders revealed that these definitions do not necessarily reflect clinical reality. Also, “number needed to treat” is an endpoint, which may lead to inaccurate assessments from a statistical point of view, because it is based on a subjective definition of remission. The adequate use of rating scales also touches ethical questions.