Clinical usefulness was mapped to the nature of innovation, so that an effective drug for a condition with no current treatment was inevitably classified as highly innovative; improvement in the treatment of

condition with no satisfactory existing treatment was classified as moderately or highly innovative depending on the nature of the innovation, Everolimus molecular weight but a more convenient treatment only could not be classified as highly innovative (table 1). Two analysts (AS and TG) independently applied these criteria to determine whether a drug was highly innovative, moderately innovative or slightly innovative. Inter-rater agreement between the two analysts was assessed using Cohen’s κ statistic. Where the analysts

disagreed, a third individual (DJW) acted as arbiter and made a determination based on discussion and further independent research (if necessary). All authors were able to review the final list of drugs and degree of innovativeness, and propose changes, which were then resolved by discussion between all authors. Analysis The proportion of new drugs categorised as highly innovative, moderately innovative or slightly innovative was calculated for the entire study period and for separate 4-year time intervals. Plots showing the numbers of new drugs categorised by degree of innovativeness (as absolute numbers and percentage of total new drugs launched that year) against year of launch were first visually inspected to identify potential time trends. Any potential trends in these data from 2004 onwards (taken as the end of the predefined dip in new drug launches) were analysed using linear regression (SPSS V.21, IBM), taking year as a continuous variable. Results There were 290 new drugs listed in relevant editions of the BNF for the 12 years from 2001 to 2012 (inclusive), a mean of 24.2/year (full list in online supplementary file 2). In the initial coding for degree of innovativeness, two analysts independently agreed on 210 drugs (72.4%,

inter-rater agreement κ=0.56 (SE=0.039, p<0.001)), after which agreement was reached on all remaining drugs through discussion involving a third arbiter. For the entire study period, 75 (25.9%) drugs were coded as highly innovative, 53 (18.3%) as moderately innovative and 162 (55.9%) as slightly innovative (table 2). Total annual numbers of new drug introductions fell from 27 in 2001 to 18 in 2006, before increasing Entinostat to a highpoint of 29 in 2010 (figure 1). Visual inspection of the line graph showing numbers of new drugs assigned to different degrees of innovativeness by year (figure 1) suggested that there were no discernible time trends in the highly innovative and moderately innovative categories, but the annual numbers of drugs categorised as only slightly innovative had risen since 2004, broadly mirroring the overall increase in numbers of new drugs.