Pipeline GARY D. NOVACK, PHD,
SECTION EDITOR
Some Are More Equal Than Others GARY D. NOVACK, PHD
I
’ve always found delight in the brilliant satire of George Orwell about revisionism in Animal Farm, wherein the ruling pigs change their egalitarian statement of “All animals are equal” to include the elitist “but some are more equal than others.”1 The concept of equality is considered frequently in statistical tests of trials of comparative therapeutics. More recently, we began to hear the concept of “non-inferiority.” In an editorial in 2006, Musch and Gillespie presented a straightforward discussion of equivalency and non-inferority.2 This editorial was a companion to the report of Diestelhorst et al of a controlled study of fixed-combination of latanoprost and timolol to each individual component e what I call a “one-bottle” vs “two-bottle” study. The authors found that mean diurnal reductions in intraocular pressure were 8.7 mm Hg and 9.0 mm Hg for the fixed dose and unfixed dose combination, respectively. This led to a between-treatment difference of 0.3 mm Hg with 95% confidence interval of 0.1 to 0.7 mm Hg. As this was within an acceptable noninferiority margin of 1.5 mm Hg, they concluded that the “one-bottle” treatment was “non-inferior” to the “twobottle” treatment.3 Musch and Gillespie cited a guidance from the European Accepted for publication January 2014. PharmaLogic Development, Inc., San Rafael CA Disclosure: Gary D. Novack, PhD consults with numerous pharmaceutical firms. © 2014 Elsevier Inc. All rights reserved.
Medicines Agency (EMA) issued in 2000 and one from the International Conference on Harmonisation (ICH) E10 issued in 2001. Four key criteria are noted that must be met in a noninferiority trial: 1) the active control must be effective; 2) there must be an acceptable non-inferiority margin– that is, “the extent to which the test treatment could be worse in its effect than the standard treatment and yet be considered essentially no different”; 3) the design of the comparative trial of the new treatment must be similar to that of the original studies; and 4) the conduct of the trial must be of high quality. After publication of Musch and Gillespie’s article, the FDA issued a draft guidance on non-inferiority in 2010. While consistent with the previous paper, this guidance is more explicit than previously available documents. FDA cites 21 CFR 314.126 to describe the four kinds of concurrently controlled trials that provide evidence of effectiveness: placebo, no treatment, dose response, and active-controlled trials. Clearly, the first three are superiority trials (i.e., the new treatment must beat the other treatments). Examples of possible outcomes of a placebo-controlled trial are shown in Figure 1. The developer of the new medicine would hope for outcome #1 e the test medicine is more effective than the placebo with the confidence intervals not including zero. The fourth design may be a superiority treatment (i.e., the new treatment beats the old treatment). However, it may be a study “.to show that the difference between the new and active control
treatment is small, small enough to allow the known effectiveness of the active control to support the conclusion that the new test drug is also effective.” The explicit definition of these types of trials is very meticulous. At the risk of over-simplification (and my own limitations in statistical theory), let me try to simplify the options as either equivalence (neither better nor worse, two-tailed) or noninferiority (i.e., no worse than, onetailed). Numerous possible outcomes of a positive-controlled trial are shown in Figure 2. Outcome #1 would be a judgment of non-inferiority e the difference in means between treatments is 0, and the upper bound of the confidence interval is less than the non-inferiority margin. Outcome #5 would be a judgment of superiority e the difference in means between treatments favors the test group, and the upper bound of the confidence interval is less than the non-inferiority margin and less than 0. How do these publications impact therapy of ocular surface diseases? As I write this column (January 2014), there are relatively few pharmacological treatments approved for the treatment of dry eye.4 RestasisÒ (cyclosporine ophthalmic emulsion) is approved in the US, Canada, and seven countries in Latin America, eight countries in Asia/ Pac and nine countries in the Middle East/Eastern Europe. Diquafasol (DiquasÒ), hyaluronic acid (HyaleinÒ), and rebamipide (MucostaÒ) are approved in Japan. None is approved in Europe. At this time, of four topical ocular small-molecule therapies publicly
THE OCULAR SURFACE / APRIL 2014, VOL. 12 NO. 2 / www.theocularsurface.com
155
PIPELINE / Novack
Figure 1. Three possible results of a placebo-controlled superiority study (point estimate, 95% confidence intervals). Assume that the response variable is positive and a larger value indicates a better response. 1. Point estimate of effect is 2; 95% CI lower bound is 1. Conclusion: Drug is effective and appears to have an effect of at least 1. 2. Point estimate of effect is 2; 95% CI lower bound is 2, well above MI; NI is not demonstrated. 3. Point estimate of C-T is zero, suggesting equal effect; but upper bound of the 95% CI for C-T is >2 (i.e., above MI), so that NI is not demonstrated. 4. Point estimate favors T; NI is demonstrated, but superiority is not demonstrated. 5. Point estimate favors T; superiority and NI are demonstrated. 6. Point estimate of C-T favors C and C is statistically significantly superior to T. Nonetheless, upper bound of the 95% CI for C-T is
SECTION EDITOR
Some Are More Equal Than Others GARY D. NOVACK, PHD
I
’ve always found delight in the brilliant satire of George Orwell about revisionism in Animal Farm, wherein the ruling pigs change their egalitarian statement of “All animals are equal” to include the elitist “but some are more equal than others.”1 The concept of equality is considered frequently in statistical tests of trials of comparative therapeutics. More recently, we began to hear the concept of “non-inferiority.” In an editorial in 2006, Musch and Gillespie presented a straightforward discussion of equivalency and non-inferority.2 This editorial was a companion to the report of Diestelhorst et al of a controlled study of fixed-combination of latanoprost and timolol to each individual component e what I call a “one-bottle” vs “two-bottle” study. The authors found that mean diurnal reductions in intraocular pressure were 8.7 mm Hg and 9.0 mm Hg for the fixed dose and unfixed dose combination, respectively. This led to a between-treatment difference of 0.3 mm Hg with 95% confidence interval of 0.1 to 0.7 mm Hg. As this was within an acceptable noninferiority margin of 1.5 mm Hg, they concluded that the “one-bottle” treatment was “non-inferior” to the “twobottle” treatment.3 Musch and Gillespie cited a guidance from the European Accepted for publication January 2014. PharmaLogic Development, Inc., San Rafael CA Disclosure: Gary D. Novack, PhD consults with numerous pharmaceutical firms. © 2014 Elsevier Inc. All rights reserved.
Medicines Agency (EMA) issued in 2000 and one from the International Conference on Harmonisation (ICH) E10 issued in 2001. Four key criteria are noted that must be met in a noninferiority trial: 1) the active control must be effective; 2) there must be an acceptable non-inferiority margin– that is, “the extent to which the test treatment could be worse in its effect than the standard treatment and yet be considered essentially no different”; 3) the design of the comparative trial of the new treatment must be similar to that of the original studies; and 4) the conduct of the trial must be of high quality. After publication of Musch and Gillespie’s article, the FDA issued a draft guidance on non-inferiority in 2010. While consistent with the previous paper, this guidance is more explicit than previously available documents. FDA cites 21 CFR 314.126 to describe the four kinds of concurrently controlled trials that provide evidence of effectiveness: placebo, no treatment, dose response, and active-controlled trials. Clearly, the first three are superiority trials (i.e., the new treatment must beat the other treatments). Examples of possible outcomes of a placebo-controlled trial are shown in Figure 1. The developer of the new medicine would hope for outcome #1 e the test medicine is more effective than the placebo with the confidence intervals not including zero. The fourth design may be a superiority treatment (i.e., the new treatment beats the old treatment). However, it may be a study “.to show that the difference between the new and active control
treatment is small, small enough to allow the known effectiveness of the active control to support the conclusion that the new test drug is also effective.” The explicit definition of these types of trials is very meticulous. At the risk of over-simplification (and my own limitations in statistical theory), let me try to simplify the options as either equivalence (neither better nor worse, two-tailed) or noninferiority (i.e., no worse than, onetailed). Numerous possible outcomes of a positive-controlled trial are shown in Figure 2. Outcome #1 would be a judgment of non-inferiority e the difference in means between treatments is 0, and the upper bound of the confidence interval is less than the non-inferiority margin. Outcome #5 would be a judgment of superiority e the difference in means between treatments favors the test group, and the upper bound of the confidence interval is less than the non-inferiority margin and less than 0. How do these publications impact therapy of ocular surface diseases? As I write this column (January 2014), there are relatively few pharmacological treatments approved for the treatment of dry eye.4 RestasisÒ (cyclosporine ophthalmic emulsion) is approved in the US, Canada, and seven countries in Latin America, eight countries in Asia/ Pac and nine countries in the Middle East/Eastern Europe. Diquafasol (DiquasÒ), hyaluronic acid (HyaleinÒ), and rebamipide (MucostaÒ) are approved in Japan. None is approved in Europe. At this time, of four topical ocular small-molecule therapies publicly
THE OCULAR SURFACE / APRIL 2014, VOL. 12 NO. 2 / www.theocularsurface.com
155
PIPELINE / Novack
Figure 1. Three possible results of a placebo-controlled superiority study (point estimate, 95% confidence intervals). Assume that the response variable is positive and a larger value indicates a better response. 1. Point estimate of effect is 2; 95% CI lower bound is 1. Conclusion: Drug is effective and appears to have an effect of at least 1. 2. Point estimate of effect is 2; 95% CI lower bound is 2, well above MI; NI is not demonstrated. 3. Point estimate of C-T is zero, suggesting equal effect; but upper bound of the 95% CI for C-T is >2 (i.e., above MI), so that NI is not demonstrated. 4. Point estimate favors T; NI is demonstrated, but superiority is not demonstrated. 5. Point estimate favors T; superiority and NI are demonstrated. 6. Point estimate of C-T favors C and C is statistically significantly superior to T. Nonetheless, upper bound of the 95% CI for C-T is
