Analysis of survival curve configuration is relevant for determining pathogenesis and causation

Abstract

Improving technology helps us to identify more and more defects at the level of genes or proteins (event) as potential sources of a disease (effect), hopefully allowing more targeted cures with a "magic bullet". However, the complex interference of genes by the environment hinders the detection of strict causal relationships between defect and disease. We consider causality as temporal relationship between event and effect, thus causation is reflected by the configuration of "survival" curves. This is indicated by several survival curves of diseases with known causal relation. Furthermore, we discuss three theoretical models: a causal chain model, a causal field concept and a causal chain model with variable order, and present three assumptions about the specific consequences for configuration of outcome curves. Clinical examples of diseases that are caused by single hits reveal an S-shaped curve of cumulative incidence. In contrast, for diseases with numerous interacting pathogenetic effectors the superposition of all contributions results in widely linear cumulative incidence curves. The rare S-shaped deformation in the survival curves in patients with recurrent cancer is in conflict with our current view of recurrent cancer as mainly being a consequence of residual tumour cell load. The assumption of a "web of causation" instead of a "causal chain" reflects a more real situation for many clinical problems and can explain the widely seen absence of decisive, causally relevant conditions. As consequences for our current treatment of cancer is not insignificant, a careful analysis of the configuration of outcome curves with recognition of an S-shape may either help to identify causal therapies or may encourage more comprehensive approaches that consider the complexity of the disease.