Accounting for disagreements on average cone loss rates in retinitis pigmentosa with a new kinetic model: Its relevance for clinical trials

Since 1985, at least nine studies of the average rate of cone loss in retinitis pigmentosa (RP) populations have yielded conflicting average rate constant values (−k), differing by 90-160%. This is surprising, since, except for the first two investigations, the Harvard or Johns Hopkins' protocols used in these studies were identical with respect to: use of the same exponential decline model, calculation of average −k from individual patient k values, monitoring patients over similarly large time frames, and excluding data exhibiting floor and ceiling effects.A detailed analysis of Harvard's and Hopkins' protocols and data revealed two subtle differences: (i) Hopkins' use of half-life t0.5 (or t1/e) for expressing patient cone-loss rates rather than k as used by Harvard; (ii) Harvard obtaining substantially more +k from improving fields due to dormant-cone recovery effects and “small −k” values than Hopkins' (“small −k” is defined as less than −0.040 year−1), e.g., 16% +k, 31% small −k, vs. Hopkins' 3% and 6% respectively. Since t0.5 = 0.693/k, it follows that when k = 0, or is very small, t0.5 (or t1/e) is respectively infinity or a very large number. This unfortunate mathematical property (which also prevents t0.5 (t1/e) histogram construction corresponding to −k to +k) caused Hopkins' to delete all “small −k” and all +k due to “strong leverage”. Naturally this contributed to Hopkins' larger average −k. Difference (ii) led us to re-evaluate the Harvard/Hopkins' exponential unchanging −k model. In its place we propose a model of increasing biochemical stresses from dying rods on cones during RP progression: increasing oxidative stresses and trophic factor deficiencies (e.g., RdCVF), and RPE malfunction. Our kinetic analysis showed rod loss to follow exponential kinetics with unchanging −k due to constant genetic stresses, thereby providing a theoretical basis for Clarke et al.'s empirical observation of such kinetics with eleven animal models of RP. In contrast to this, we show that cone loss occurs in patients with increasing −k values during RP progression. And as the Hopkins' protocol selects more advanced RP cases than Harvard's to assure avoidance of ceiling effects (Harvard does this by kinetic monitoring), we show increasing −k kinetics to be the reason Harvard obtains more +k and small −k values. Thus the combined effects of (i) and (ii) produce Harvard's smaller average −k value. The relevance of the increasing biochemical stress model for optimizing clinical trials is discussed.