When compared with the compendial requirements , BAK-containing latanoprost exceeded the EP-A criteria at all time points . Travoprost with sofZia, however, while meeting USP standards [6, 14], did not meet the EP-A criteria for either bacteria or fungi, exhibiting only modest reductions at 6 and 24 hours, nor did it meet EP-B criteria due to its limited effectiveness against Staphylococcus aureus. Staphylococcus infections are frequently associated with both primary and recurrent bleb infections following trabeculectomy  and endophthalmitis subsequent to postoperative procedures such as lens replacement . Pseudomonas aeruginosa is also a common cause of endophthalmitis, occurring postoperatively or subsequent to corneal ulcers, and is often associated with poor visual outcomes .
The early time points, which assess the rate of kill of the challenge organisms, revealed the most significant differences between the two preservative systems. Latanoprost with 0.02% BAK exhibited complete reduction of a large microbial insult (bacterial and fungal) within the first 6 hours of exposure while travoprost with sofZia showed only modest reductions. These results are especially important as the early time points simulate microbial contamination that may occur upon use and be present over the next 24 hours after use. In addition, the fungal/yeast challenge never reached a point of "no recovery" in the travoprost samples during the 28-day test.
Benzalkonium chloride also has been shown to be more effective than other preservatives when measured against the EP-A criteria . When artificial tears containing BAK (0.01%)/ethylenediaminetetraacetic acid (EDTA; 0.05%), chlorobutanol (0.5%), stabilized oxychloro complex (50 parts per million), sodium silver chloride complex (0.001%) or methyl-, ethyl - and propylparaben (undeclared concentration) were compared , the product containing BAK/EDTA alone satisfied the criteria for all test microorganisms. The majority of products failed the criteria for one or more bacteria, notably with the 6- and 24-hour samples. An agar diffusion test also was performed, with only the BAK/EDTA sample showing a zone of inhibition; the effect was shown to be due to BAK only since other products without EDTA gave similar results .
Recent studies in which patients were videotaped to assess their success at instillation of topical ocular hypotensive medications highlight the concerns about bottle contamination [3, 4]. In the first of these studies, 92.8% of patients with a diagnosis of glaucoma or ocular hypertension who used 1 or more glaucoma medications for at least 6 months reported no problems administering their eye medications; yet, less than a third of patients were successful at instilling a single drop with touching the bottle to the eye . In a subsequent study in patients with visual impairment or moderate to severe visual field loss, only 39% were able to instill a single drop without touching the eye; age (<70 vs ≥70 years) was found to be a significant predictor for less successful instillation . These studies demonstrate that bottle contamination is a more important issue than previously believed.
There has been an ongoing controversy about the contributions of BAK-containing ophthalmic solutions to ocular toxicity, particularly using in vitro studies and rabbit models, many with exaggerated-use protocols (for reviews, see Kaur et al  and Furrer et al ). The relevance of these studies to the clinical setting is not well established given the various methodologies, models, exposure times and concentrations. Several recent studies, all sponsored by Alcon, have compared travoprost to latanoprost or bimatoprost with respect to ocular tolerability in glaucoma patients [20–23]. In the first of these studies, patients (n = 691) who required alternate therapy due to tolerability issues were switched from either latanoprost or bimatoprost to travoprost with sofZia . While there was no significant difference in the reported OSD index (OSDI) scores between patients who were classified as normal at baseline (n = 456), patients who were symptomatic at baseline (n = 235) reported significant improvements in their scores 3 months after switching to travoprost. However, as the authors indicated, the study design (nonrandomized, nonmasked) was limited so that expectation of improvement may have resulted in patients subjectively reporting a more favorable outcome.
In other studies involving patients with preexisting OSD, tolerability findings have been inconsistent. In one double-blind study, patients who were receiving latanoprost and reported ocular dryness and irritation (n = 33) were randomized to receive latanoprost in one eye and travoprost with sofZia in the other eye; eyes were assessed by a single examiner every 3 to 4 weeks for 3 months, and patients completed an OSDI survey . Significant increases in corneal staining occurred in the travoprost-treated eyes compared to the latanoprost-treated eyes, with OSDI surveys also showing a trend toward more dryness and irritation symptoms in the travoprost eyes. There were no differences in tear breakup times (TBUT), intraocular pressure, visual acuity or Schirmer testing between the two groups . In an open-label, prospective study of patients (n = 20) with a baseline TBUT of less than 6 seconds, significant increases in mean TBUT and decreases in mean OSDI scores and corneal staining were reported 8 weeks after switching from latanoprost to travoprost with sofZia .
In contrast, in a prospective, double-masked, randomized comparative study of 54 subjects, there were no significant differences between latanoprost and travoprost with sofZia with respect to reported discomfort scores following a single instillation of either agent . In a small, prospective, observational cohort study with masked examiners supported by Merck , patients with glaucoma or ocular hypertension (naive to treatment, n = 10; previously on latanoprost, n = 8) were instructed to use latanoprost in the right eye and travoprost with sofZia in the left eye. There was statistically significantly less conjunctival hyperemia in eyes treated with latanoprost (both in the naive and previously treated patients) and in corneal staining in eyes previously treated with latanoprost but no statistical difference in TBUT, change in intraocular pressure from baseline, or impression cytology between the treatment groups . None of these studies specifically assessed the incidence of ocular infections or rates of bottle contamination; additional studies are warranted.
Patients receiving ocular hypotensive medications are reported to have a high prevalence of OSD, with 59% of patients reporting OSDI symptoms in at least one eye; Schirmer testing was abnormal in 61% of patients and TBUT was decreased in 78% of patients . After adjustment for age and sex, factors considered to influence the prevalence of OSD, multivariate logistic regression found that the use of BAK-containing agents was associated with a two-fold increase risk of lissamine green staining in the 22% of patients with positive results (none had severe staining based on a scale of 0 to I, normal; II to III, mild to moderate; and IV to V, severe). These rates are higher than those reported in population-based studies, which likely reflects the fact that some of these patients may have been referred due to OSD symptoms or may have been treated with multiple preservative-containing eye drops . This hypothesis is consistent with the results of a retrospective analysis of three large prescription databases sponsored by Pfizer . Patients newly treated with latanoprost or travoprost with sofZia and without a diagnosis of dry eye or ocular infection in the prior 6 months had no significant differences in the rates of dry eye or ocular infections (identified by International Classification of Diseases, Ninth Revision, Clinical Manifestation code or by prescription for cyclosporine ophthalmic emulsion or ocular antibiotics) at 1 year .
Of importance in considering the findings of the present study, patients with OSD have an increased risk of microbial keratitis [27–29], with OSD found to be a predisposing factor in 21% of cases of bacterial keratitis in one study  and in 15% of case in another study . Staphylococcus species were found to be the most commonly isolated organisms in OSD-associated bacterial keratitis [27, 29]. Moreover, a history of OSD was found to be significantly associated with a "very poor" visual outcome following bacterial keratitis . Thus, the failure of travoprost to satisfy even EP-B requirements due to the limited effectiveness against Staphylococcus aureus at 24 hours raises concerns about the adequacy of its preservation.
To date, in clinical usage and in observational studies, the findings are mixed with regard to ocular tolerability and may depend upon the study design. While some switch studies found improvements in tolerability when switching from bimatoprost/latanoprost to travoprost with sofZia [20, 21], one study found increases in corneal staining and irritation when switching from travoprost with sofZia to latanoprost . Still, in the large, retrospective study sponsored by Pfizer , there were no significant differences in dry eye or infection between patients receiving latanoprost versus travoprost with sofZia. The randomized, masked, clinical registration study sponsored by Alcon comparing travoprost with BAK to travoprost with sofZia found no differences in adverse events or safety endpoints . Therefore, the presumed benefits of BAK-free or other alternative preservative systems in terms of ocular tolerability remain to be clearly established.