“…But without a placebo group, we have no way to check whether some patients would have improved merely by being in the study. In a follow‐up paper, Monseur et al
37 describe alternative 2‐arm designs that attempt to offer some protection against the bias that a 1‐arm study may contain. Among other innovations, these authors use a parametric (linear) changepoint model where for subject
and time point j , the mean response is
where
is the time of the intervention, and the “+” superscript denotes positive part (ie,
for
, and
for
).…”