Common use of Statistical Analysis Clause in Contracts

Statistical Analysis. After analyzing price premiums descriptively, we estimated six regression models. The dependent variable (i.e., the relation between a pharmaceutical’s costs and those of its comparator) appeared to be gamma-distributed (many values at or close to 0), which we confirmed by applying the Modified Park Test [43]. Therefore, we could not use a simple ordinary least squares (OLS) regression. To avoid retransformation problems we preferred a generalized linear model (GLM) over a log OLS model [43]. Applying the Pregibon Goodness of Link Test [44] we confirmed the usage of a log-link. Thus, we used a GLM with a log-link function to analyze the impact of added benefit on price pre- miums. To allow for meaningful interpretation, we calculated marginal effects for each variable, with all other variables set to their means. We conducted several sensitivity analyses. First, because single observations may have a large impact on results based on such a small sample, we excluded substances with a ▇▇▇▇’▇ D greater than the conventional cutoff point of 4/n [45]. Second, we controlled for whether pharmaceuticals had been assessed dur- ing the first 7 months after the AMNOG legislation came into effect. During this transitional period, manufacturers were advised on the completeness of their dossiers by the G-BA and, if required, were granted an additional 3 months to complete them [46]. Third, we included a variable that controlled for whether the final price had been set by the arbitration board. Last, instead of using average comparator costs when several interchangeable comparators were eligible for the same patient subgroup, we reran our models using the least and the most costly comparators. All analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC).