Problem 1: small target population

Uveitis has an annual incidence of 14–50 per 100 000 [46], with a prevalence of around 38–115 per 100 000 in the general population [47–49]. Since the majority of these cases are acute anterior uveitis responsive to topical therapy, the population of patients who need systemic therapy (and are eligible for clinical trials of immunosuppressants) is small. Other inflammatory diseases such as scleritis have a higher proportion requiring systemic therapy, but are overall much less common. It is partly in response to this, that increasing collaboration has emerged in recent years. One example is the SITE consortium, a collaboration of five academic ocular inflammation practices in the United States who undertook a retrospective cohort study primarily to investigate long-term adverse events occurring in patients on systemic therapies for ocular inflammatory disease; although not the primary aim, this study also provides interesting data on outcomes in the use of these drugs [50]. The study group comprised 7957 (of whom 2340 were treated with systemic immunosuppression) patients observed over 68751 visits, spanning 14.910 person years [2–6]. Whilst accepting the limitations inherent in a retrospective study, this is an important milestone in being the largest dataset of immunosuppression in ocular inflammatory disease.

Problem 2: heterogeneous disease groups

Ocular inflammatory disease is very heterogeneous group. Within uveitis, classification is generally by anatomical grouping and etiology, including the presence or absence of systemic disease. ‘Splitting’ i.e. defining a very pure cohort (e.g. Posterior uveitis of Birdshot pattern) leads to a very small target population, whereas ‘lumping’ enables easier recruitment but may be clinically meaningless due to the range of disease included. This is true both for routine clinical practice and for clinical trials. In clinical trials maximization of the signal: noise ratio is critical. An intervention may be highly effective for a disease but will fail to return a statistically significant benefit if it is trialed on too broad a group of patients (e.g. patients who have conditions that look superficially similar but differ fundamentally in etiology).

Problem 3: poorly defined phenotypes

In addition to its heterogeneity as a group, the uveitis subtypes are imperfectly defined with no diagnostic criteria with high enough sensitivity and specificity to reliably separate all cases of one uveitis entity from another. This leads to considerable variation in classification and diagnosis. To some extent such markers may only become apparent as we better understand disease etiology. This has occurred in the field of retinal dystrophies where there has been a shift from grouping by somewhat arbitrary pictorial descriptions (e.g. butterfly-shaped macular dystrophy) to defining disease by the gene responsible (e.g. a peripherin/RDS retinal dystrophy). This is important not only because it leads to a more clear-cut and objective classification of disease, but also because a classification based on real differences in etiology is more likely to translate into an appropriately targeted therapeutic approach.

Problem 4: diagnostic inconsistency

Within uveitis the definition of syndromes by pattern and the lack of clear diagnostic markers leads to significant variation between clinicians in the classification of uveitis entities. Currently a major international collaboration - the Standardization of Uveitis Nomenclature consortium - is seeking to define classification criteria for 28 different uveitis syndromes [51]. The consortium have identified 194 uveitic terms and ‘dimensions’ including 87 unique terms that are specifically used by uveitis specialists to describe signs and symptoms. These were mapped to the 28 uveitic syndromes under consideration. Currently 250 cases per uveitis syndrome are being gathered which will be used to validate the mappings and to form the basis of a classification criteria and a proposed ontology for uveitis [52]. Although this system is based on pattern rather pathogenesis, this project does have the potential to significantly improve standardization of uveitis classification with benefits for both clinical practice and trials.

Problem 5: subjective outcome measures

One of the advantages of managing inflammation in the eye (as opposed to elsewhere in the body) is that the transparent nature of the cornea and visual axis enables us to see and score inflammatory activity and titrate treatment accordingly. The flip-side is that we have accepted these same subjective activity indices in clinical trials. Non-invasive technologies which can objectively quantify the key parameters of cellular activity, flare and vitreous haze are needed. With regard to the assessment of AC flare, laser flare meters have potential but current models are relatively time-consuming and cumbersome, with some concerns over variability especially in the non-ideal patient (e.g. posterior synechiae, patient mobility). Some models have sought to also quantify cellular activity but with limited success. With regard to the posterior segment the NEI vitreous haze score described by Nussenblatt et al. (which depends on the clinician scoring the clarity of the optic disc which can be compared to a standard set of photographs) is the key outcome recognized by the FDA [53]. Although it is the gold-standard, it has a number of limitations including that it is (1) subjective [54]; (2) non-continuous, leading to very large steps in disease activity between categories; (3) poorly discriminatory at lower levels of vitreous haze, with most cases of active uveitis being scored at 0.5+ or 1+; and (4) limiting of recruitment and sensitivity in a clinical trial context (where a 2 point change is usually required to be counted as a significant change).

An adaptation of this technique proposed by Davis and colleagues is to score clarity on photographs (rather than on the live biomicroscopic view of the NEI technique) which can then be compared to a more extensive set of reference images leading to a potentially more discriminatory 0–9 score (vs. the 0, 0.5, 1, 2, 3 and 4 scale of the SUN modification of the Nussenblatt) [55]. Overall however this is still a subjective technique and shares most of the key limitations of the standard vitreous haze score. A truly objective measure is likely to be based on quantification of vitreous density or reflectivity using current or future imaging technologies.

Problem 6: the ‘distraction’ of visual acuity

Whilst it is absolutely right that our key aim is to retain and restore vision, we must recognize that visual acuity is a very poor marker of the efficacy of a drug in inflammatory eye disease. The impact of uveitis on visual acuity will depend on both the activity of the disease and the damage caused by the disease. Thus whilst vision may improve due to treatment-induced improvement in cystoid macular edema, vitritis, keratic precipitates and aqueous clarity, this benefit may be obscured by loss of vision due to cataract, band keratopathy, macular scarring or even glaucoma. A further complexity arises in that some of these types of damage are reversible (notably cataract and band keratopathy) leading to the slightly bizarre scenario that their correction has to be specifically prohibited within most clinical trials. Additionally it is impossible to accurately quantify the extent to which each factor is contributing to the reduced visual acuity. Although the clinician is used to making these judgment calls every day in clinic - for example when deciding the extent to which advancing cataract or persistent cystoid macular edema is responsible for a fall in vision – these estimates are very approximate and frequently shown to be incorrect. It should also be recognized that visual acuity is a subjective parameter which despite every effort towards standardization may be affected by factors such as patient mood, general health and compliance. Visual outcome remains the core purpose for clinical trials in inflammatory eye disease, but it is less clear how it should be utilized as a trial endpoint.

The increasing and important focus on patient reported outcome measures (PROMs) as a component of clinical trials may start to capture some of these aspects of the overall benefit experienced by the patient. It is encouraging that more recent clinical trials in uveitis have often incorporated measures both of quality of life and health utility. These include specific vision-related quality of life measures (such as the 25-item NEI-Visual Function Questionnaire, NEI-VFQ25 or the Vision-related Quality of Life Core Measure, VCM-1) [56–61] general health-related quality of life measure (such as the short form-36, SF-36) [58, 59, 61, 62] and health utility measures (such as the EuroQol 5-dimension, EQ-5D, and Visual Analogue Scores) [61, 63].

Problem 7: Low commercial interest

In the past pharmaceutical companies have shown little interest in ocular inflammatory disease, partly due to the relatively small market. Encouragingly the last decade has seen a surge of interest with industry-sponsored studies of both novel agents – such as the calcineurin antagonist voclosporin and the anti-IL17 agent secukinumab – and of novel routes for older drugs (e.g. intravitreal corticosteroid implants, intravitreal sirolimus) and providing ultimately robust evidence of anti-TNF blockade.

Ivacaftor: detailed understanding of pathogenesis leads to specific drug-design matched to carefully defined disease groups

Cystic Fibrosis is a life-limiting multisystem disorder arising from abnormal function or complete loss of function of the protein, Cystic Fibrosis Transmembrane Conductance Regulator (CFTR). The CFTR is an ABC transporter-class ion channel in epithelial cell membranes which regulates the constituents of sweat, mucus and digestive fluids (reviewed [64]). Although non-specialists often regard CF as a single condition, there is significant variation in the severity and organ systems preferentially affected. Since identification of the CFTR gene in 1989, it has become increasingly clear that a key part of this variation in phenotype is dependent on the specific mutation in the CFTR gene. G551D is a mutation which affects around 5% of CF patients. Critically, although the G551D mutation does cause impaired ion transport of the CFTR, the CFTR protein is still expressed on the surface of the epithelial cell [65]. It was therefore postulated that it might be possible to intervene in this subgroup of CF patients by using drugs that potentiate the existing CFTR function. Increasing interest led to the Cystic Fibrosis Foundation contributing an estimated $75 million to support the development of one such drug Ivacaftor (Kalydeco) by Vertex Pharmaceuticals. In 2011 two placebo-controlled clinical studies comprising a total of 213 patients showed significant and sustained improvement in lung function [66]. In early 2012 Ivacaftor was approved by the FDA for use in patients with the G551D mutation. Thus for the first time there is a therapy which actually targets and restores function within the abnormal protein in CF, albeit in a carefully defined subgroup of the disease and with an estimated annual cost of $294, 000 per patient.

Although posterior segment uveitis is probably more heterogeneous than CF (at least in terms of etiology and pathogenesis), one of the key lessons here is that the development of drugs such as Ivacaftor is only possible when there is accurate classification of subtypes of disease based on real differences at the pathogenic level. Within ophthalmology we see this approach showing promise in the field of retinal dystrophies where accurate definition of genotype opens the possibility of targeted gene therapy [67]. As our understanding of the pathogenesis of the ocular inflammatory diseases improves, therapeutically useful classification of disease should follow.

Vemurafenib: established drug-development pipeline enables rapid translation for patient benefit

Metastatic melanoma has a very poor prognosis with a median survival of 8 to 18 months after diagnosis for stage IV melanoma. In 2002 Davis et al. noted that around half of melanomas have an activating mutation in the B-RAF oncogene, a regulator of the MAP kinase pathway that controls cell proliferation [68]. Subsequently it was shown that 90% of these mutations are of the V600E type. In 2008 Tsai et al. published their results documenting the targeted development of a specific inhibitor of the mutant V600E BRAF and its effects on both melanoma cell lines and tumor xenograft models [69]. In 2010 the first Phase I trial defined the maximal tolerated dose and noted frequent tumor responses [70]. In 2011 the first phase 2 and first phase 3 reported efficacy data, the latter demonstrating significant benefits with a six month survival of 84% (vs. 64% for standard treatment) and a 74% relative reduction in death or disease progression compared to standard treatment [71, 72]. Later that same year the drug was approved by the FDA for the treatment of late stage melanoma, a development time of just nine years since the original Davis paper.

Within uveitis, the recent development of the anti-IL17 therapy secukinumab and rapid translation into clinical trials for non-infectious posterior segment disease is an encouraging example of how ocular inflammatory disease can start to benefit from established drug development pipelines [73]. Although it is too early to comment on whether secukinumab will be a major chapter or only a foot-note in the management of uveitis, its targeted development from animal model to clinical trial has been a significant milestone within the field.

I-SPY-2: adaptive trial design and public-private partnership underpin innovative rolling drug-testing program

I-SPY-2 is a phase II rolling drug screening program to test new therapies for breast cancer [74]. The design has six treatment arms, including an arm for standard therapy. A key feature is that randomization is adaptive (within biomarker subsets) whereby the probability of being assigned to a particular treatment arm increases if the outcome of the prior patients in the group was good [75, 76]. As the trial evolves the treatment arms are replaced either because they show sufficient success to graduate to a smaller, focused phase III study or because they are terminated due to lack of benefit (based on Bayesian predictions of success or futility). This design means that the more successful a drug, the faster that it will move through the screening process. It also means that trial participants will tend to receive the more effective treatments. Fewer patients are required (per drug outcome) and it is predicted to be significantly cheaper than standard single drug non-adaptive designs. Up to 12 drugs will be tested through the I-SPY-2 program. This study is a collaboration across the USA between the National Cancer Institute, the FDA and around 20 major cancer centers, with active engagement by pharmaceutical companies who are providing their drugs for free. There is a commitment to publish all key trial results and the data arising from the trial will be managed by the independent non-profit organization the Foundation for the NIH. It is probably too early to judge but if successful, I-SPY-2 could revolutionize our approach to clinical trials [75, 76].