Study Population

All patients with noninfectious ocular inflammatory disease who had started receiving treatment with mycophenolate mofetil during the study as the sole noncorticosteroid immunosuppressive agent at 3 ocular immunology and uveitis clinics and an approximate 40% random sample of such patients from a fourth center were selected to be the study population. This group of patients was nested within a larger study, the Systemic Immunosuppressive Therapy for Eye Diseases (SITE) Cohort Study, the methods of which have been reported previously. Patients from a fifth center participating in the parent study were excluded, because the co-management approach used by the center—wherein many follow-up visits were conducted outside the center—markedly biased ascertainment of the time-to-treatment response.

Data Collection

Trained, certified, expert reviewers reviewed the medical records of all patients and entered information for each eye of every patient at every visit into a custom Microsoft Access database (Microsoft Corporation, Redmond, Washington, USA). Demographic information obtained during the initial visit and details of all medications in use at every clinic visit were recorded. Details of ocular inflammation activity status—based on clinical evaluation using external, slit-lamp, and dilated fundus examination—also were recorded. Sequelae of ocular inflammation were noted when present.

Follow-up and Main Outcome Measures

Patients were followed until they stopped using mycophenolate mofetil or started using an additional noncorticosteroid immunosuppressive agent; had an event of interest; or stopped attending a study clinic; or until the end of data collection was reached—whichever occurred first. The primary outcome measures were: success in gaining control of inflammation; success in maintaining control of inflammation after tapering systemic corticosteroids to specified thresholds (“corticosteroid-sparing effects”); and discontinuation of treatment. Control of inflammation was categorized at each visit as “active,” “slightly active,” or “inactive” for every eye at every visit. Descriptors such as “active,” “uncontrolled,” “worsening inflammation,” or “disease progression” in the case notes were counted as indicating “active inflammation.” Inflammatory activity was scored as “slightly active” when inflammation was minimally present, described in the patient notes using descriptors such as “mild,” “few,” “trace cells,” or “trace activity.” Inflammation was defined as “inactive” when descriptors such as “quiet,” “quiescent,” “no cells,” and “no active inflammation” were used. When data regarding activity could not be ascertained from case notes, data regarding activity at the visit in question were entered as missing. Corticosteroid-sparing effects were evaluated based on the time to reduction of the prednisone dose to 10 mg, 5 mg, or 0 mg while maintaining inactivity of the ocular inflammation, among those not meeting each respective criterion for success at the outset. Prednisone-equivalent doses of alternative corticosteroids were included in these evaluations when required. When applicable, dates of discontinuation of mycophenolate mofetil and the reasons for the discontinuation documented in the charts were noted.

Statistical Methods

SAS version 9.1 (SAS Corporation, Cary, North Carolina, USA) was used for all analyses. The frequencies of demographic and clinical characteristics at enrollment among patients treated with mycophenolate mofetil were tabulated. The benefits of therapy were evaluated based on time to success (control of inflammation and corticosteroid-sparing effects). A time-to-failure approach was used for time to discontinuation of mycophenolate mofetil. In the primary analyses, outcomes were accepted only when observed at 2 or more visits spanning at least 28 days, to avoid the error of including transient improvements and brief interruptions in therapy as successes or failures.

The Kaplan-Meier method was used to estimate the proportion of patients who achieved success or experienced failure at or before 6 and 12 months of therapy. Incidence rates, by person and/or by eye, were calculated as the number of events divided by the person- or eye-years of observation. Cox regression was used to calculate crude and adjusted hazard ratios (HRs) for occurrence of major outcome events, in relation to potentially explanatory covariates.