Although we hypothesized that the rate of index therapy change would be lower with travoprost-Z preserved with SofZia than with latanoprost preserved with BAK, we found change rates to be statistically significantly lower in latanoprost-treated patients both within six months and across the full study period. As in studies of persistency with ocular hypotensive agents generally, rates of uninterrupted use were lower than desirable for both latanoprost and travoprost-Z, but patients initially prescribed travoprost-Z were approximately 50% more likely to change therapy both within the first 6 months and across the full study period (p < 0.01 for both comparisons with latanoprost). Combining proportions of patients who either switched or discontinued to estimate how many patients completely stopped taking the initial therapy revealed that, within the first six months, 12.5% (79/632) of latanoprost-treated patients and 15.3% (41/268) of those treated with travoprost-Z stopped the initial therapy and that 19.8% (125/632) and 23.5% (63/268), respectively, did so across the full study period. Similar proportions of patients in the two treatment groups changed due to a glaucoma-related surgery or procedure (approximately 14% at each measurement time point).
Our finding of greater persistence with latanoprost monotherapy than with travoprost-Z monotherapy contrasts with results of retrospective claims database analyses of prostaglandin analog treatment patterns [34–36]. In those studies, rates of adjunctive IOP-lowering therapy use favored travoprost-Z, with differences between latanoprost and travoprost-Z ranging from 7.6% over 12 months (16.5% vs 8.9%, respectively)  to 2% over two years (37% vs 35%, respectively) . Directly comparing findings of the present study and the published claims database analyses is difficult given differences in research questions, definitions of outcomes, and statistical analyses. In particular, the studies by Schmier et al [34–36] included only newly initiating patients who remained on the index prostaglandin for least 12 or 24 months (depending on the study), while the present research identified patients newly prescribed an ocular prostaglandin and tracked therapy changes throughout the follow-up period. Nevertheless, such contrasting results suggest that additional research is needed to evaluate whether the low levels of BAK contained in ophthalmic solutions may cause significant adverse corneal effects that impact patient persistence.
Changes in ocular hypotensive therapy may be precipitated by a number of factors. In the GAPS  with an average duration of chart review of 4.1 years, lack of efficacy was the most common reason cited by physicians for switching medication followed by adverse events (43% vs 19%, respectively). Among patients who changed therapy in the present study, uncontrolled IOP was noted as a reason for change within the first six months in >60% charts and as a reason for nearly 55% of changes from latanoprost and 63% of changes from travoprost-Z across the full study period. As in the GAPS , adverse events were the second most frequently cited reason for change, noted in between 11% and 21% of cases; changes were attributed to adverse events in somewhat greater proportions of those treated with travoprost-Z.
The most frequently recorded adverse event was hyperemia, which was noted significantly more frequently with travoprost-Z than with latanoprost at both measurement time points. In the GAPS , hyperemia also was the most common adverse event. In that study, of the 195 patients with charted adverse events, hyperemia was noted in 135 (69%) records, and an episode of hyperemia occurred in a significantly greater proportion of those exposed to travoprost (35%) than to latanoprost (22%; p = 0.0123). It has been suggested that BAK exposure is related to an increased incidence of dry eye , but therapy withdrawals from dry eye were not obvious in the latanoprost with BAK group.. This lack of association between these agents and the incidence of dry eye parallel the results of a retrospective analysis of three large prescription databases that found no significant difference in the incidence of ocular surface disease (defined as dry eye or ocular infection) between patients prescribed latanoprost versus travoprost-Z (14.0% vs 14.4%, respectively; p = 0.45) .
The American Academy of Ophthalmology [38, 39] and the European Glaucoma Society  have produced guidelines for the evaluation and treatment of patients with glaucoma or ocular hypertension, but eye care professionals may not routinely follow these recommendations [41–44]. In the GAPS , IOP and results of disc evaluations and imaging and of visual field tests were recorded in charts of 90% of open-angle glaucoma patients; in contrast, chart notations of central corneal thickness measurement, gonioscopy, and establishment of an IOP target level were present for about half of patients. In the present study, we tabulated frequencies of ocular procedures and ocular surface tests performed at baseline. An IOP measurement was recorded for nearly all (>98%) patients in both treatment groups and about half of charts included findings of perimetry, ophthalmoscopy, gonioscopy, and central corneal thickness measurement. Other procedures and ocular surface tests were performed less frequently.
A major strength of the current study was the large number of charts reviewed (N = 900). However, the follow-up time frame was too short to support assessments of changes in parameters such as visual field. Rates of therapy "gaps" and restarts, which have been documented using large medical/pharmacy databases [10, 11], cannot be reliably inferred from medical records. Although the 14 sites from which data were abstracted were geographically diverse, they may not represent the full population of ophthalmology practices. Every effort was made to ensure the consistency of chart abstraction across sites, but no formal tests of reliability were undertaken. In addition, time to and reasons for changing medication were not cross-validated by an independent committee and were not evaluated in a masked fashion. Insufficient IOP reduction was the most commonly reported reason for change, but IOP changes at the patient level were not assessed and could be considered in future analyses. Finally, because latanoprost and the original formulation of travoprost with BAK have similar tolerability profiles [32, 33], we hypothesized that any improvement in the tolerability of travoprost with SofZia would lead to lower rates of therapy change in comparison to latanoprost with BAK. Although we did not find lower rates of therapy change among those treated with travoprost-Z, future research might expand the study by comparing persistence with BAK-preserved travoprost, bimatoprost, and latanoprost versus ocular hypotensive formulations without BAK - including travoprost with SofZia