Topical Nonsteroidal Anti-inflammatory Drugs for Corneal Abrasions: Meta-analysis of Randomized Trials

ACAD EMERG MED d May 2005, Vol. 12, No. 5 d www.aemj.org 467 SPECIAL CONTRIBUTION Topical Nonsteroidal Anti-inflammatory Drugs for Corneal Abrasions: Meta-analysis of Randomized Trials LisaAnneCalder,BScH,MD,SowmyaBalasubramanian,BSc,MSc, Dean Fergusson, MHA, PhD Abstract Objectives: To determine the effectiveness of topical nonsteroidal anti-inflammatory drugs (NSAIDs) in traumatic corneal abrasions. Methods: This was a systematic literature review and meta-analysis of randomized clinical trials (RCTs). The following databases were searched: MEDLINE (1966 2004), EMBASE (1980 2004), and Cochrane Central Register of Controlled Trials and Database of Systematic Reviews (first quarter 2004). The structured search strategy included a RCT filter and the terms cornea, wounds and injuries, trauma, corneal diseases, eye injuries, anti-inflammatory agents, nonsteroidal and specific trade names of topical NSAIDs. In addition, four journals in ophthalmology and emergency medicine were hand searched. Two authors independently reviewed citations from the literature searches. To be included, studies had to be RCTs evaluating topical NSAIDs in traumatic corneal abrasions. Trials were included regardless of language or whether they were unpublished or published. Exclusion criteria were corneal ulcers, nonrandomized studies, animal studies, or perioperative setting. Outcomes were pain scale scores at 24 hours and adverse effects. Two independent reviewers assessed four trial quality components: randomization, double blinding, reporting of withdrawals, and concealment of allocation. Weighted mean difference, using a random effects model, was calculated. Results: Of the 200 citations identified, 11 RCTs met inclusion criteria. Eight trials were identified from the MEDLINE search, two from the EMBASE search, and one from conference proceedings. Seven trials enrolled fewer than 100 patients, and more than half of the studies were conducted in Europe. Five trials reported suitable data for analysis. The overall weighted mean difference for 459 patients was a reduction in pain by 1.30 points (95% confidence interval = 21.56 to 21.03) on the pain scale. Five of the trials met criteria for good quality. Transient stinging was a commonly noted adverse effect. Conclusions: Topical NSAIDs are effective analgesics for traumatic corneal abrasions. Key words: corneal diseases; eye injuries; anti-inflammatory agents; nonsteroidal. ACA- DEMIC EMERGENCY MEDICINE 2005; 12:467 473. When a patient presents with a traumatic corneal abrasion, possible therapeutic modalities include cycloplegics, patching, topical nonsteroidal anti-inflammatory drugs (NSAIDs), and oral analgesics. The amount and quality of evidence supporting the use of these modalities are varied. There are few studies supporting the use of cycloplegics in corneal abrasions. One randomized controlled trial showed no benefit in pain scores and oral analgesia use. 1 Many clinical trials 2 9 and a meta-analysis 10 have shown that patching does not improve pain or healing in traumatic corneal abrasions. One therapy that does show promise for effective pain relief is topical NSAIDs. Several small clinical trials have examined From the Clinical Epidemiology Unit, Ottawa Health Research Institute, University of Ottawa (LAC, DF), Ottawa, Ontario, Canada; and Queen s University (SB), Kingston, Ontario, Canada. Received September 15, 2004; revision received October 26, 2004; accepted October 27, 2004. Address for correspondence and reprints: Lisa Anne Calder, BScH, MD, Department of Emergency Medicine, The Ottawa Hospital, Civic Campus, 1053 Carling Avenue, Ottawa, Ontario, Canada K1Y 4E9. Fax: 613-761-5351; e-mail: lcalder@sympatico.ca. doi:10.1197/j.aem.2004.10.026 topical NSAIDs as analgesics for corneal abrasions, yet definitive evidence does not exist. A recent survey of Canadian emergency physicians assessing the management of traumatic corneal abrasions 11 showed that the average physician saw ten cases per month and yet only 53% were using topical NSAIDs. Given this lack of consensus, we conducted a systematic review of the literature and meta-analysis to identify randomized clinical trials (RCTs) using topical NSAIDs for traumatic corneal abrasions. The objective was to determine whether topical NSAIDs provide effective analgesia for this common eye injury. METHODS Study Design. This was a systematic literature review and meta-analysis of RCTs. Ethics approval was not required for this study because it did not involve human subjects or medical records. Study Population and Setting. Only randomized controlled trials were considered for analysis. The study population was patients with traumatic corneal abrasions, but the setting could be the emergency department or an ophthalmology clinic. Topical rather

468 Calder et al. d CORNEAL ABRASION META-ANALYSIS specify the type of pain scale used a priori. Clinical heterogeneity was assessed by examining descriptive statistics of individual study elements such as study population, setting, intervention type, and outcome assessment. Figure 1. Results of literature selection process. than oral NSAIDs were included because we did not want to confound the effect of analgesia by combining different routes of administration. There were no restrictions on type of NSAID because the mechanisms of action within the class are the same. The pain outcome had to be measured on a scale; we did not Study Protocol. A structured search strategy was developed and included the medical subject heading (MeSH) terms cornea, cornea/injuries, wounds and injuries, corneal diseases, eye injuries, antiinflammatory agents, non-steroidal, and specific trade and generic names of topical NSAIDs (diclofenac, Voltaren, ketorolac, Acular, flurbiprofen, indomethacin, Indocollyre, Indocid). The following text words were included with truncation where appropriate: injury, trauma, abrasion, wound, tear, perforation, anti-inflammatory, and non steroid. This strategy was combined with a Cochrane RCT filter 12 and applied to the following databases: MEDLINE (1966 to the first week of February in 2004), EMBASE (1980 to week 10 of 2004), the Cochrane Central Register of Controlled Trials (first quarter 2004), and the Cochrane Database of Systematic Reviews (first quarter 2004) using a standard search engine (Ovid). In addition to searching the bibliographic databases, the reference lists of all identified clinical trials and reviews recovered were hand searched. Academic Emergency Medicine, American Journal of Ophthalmology, Annals of Emergency Medicine, and Archives of Ophthalmology were also hand searched for relevant trials (from 1995 to 2004). The search included conference proceedings. Pharmaceutical companies that manufacture topical NSAIDs were contacted to obtain access to industry-sponsored trials. Efforts were made to contact the authors of the original articles to confirm the reported findings and determine any unreported details. TABLE 1. Study Characteristics First Author Year Country Setting NSAID Albertiy 2001 France, Portugal Ophthalmology clinics Indomethacin 0.1% Brahma 1996 United Kingdom Emergency eye center Flurbiprofen 0.03% Donnenfeld 1995 United States ED and ophthalmology clinic Ketorolac 0.5% Goyal 2001 United Kingdom Unclear Ketorolac 0.5% Jayamanne 1997 United Kingdom Eye casualty department Diclofenac 0.1% Kaisery 1997 United States Eye emergency unit Ketorolac 0.5% Le Sagey 2004 Canada ED Diclofenac Patroney 1999 Italy Emergency eye clinic Indomethacin 0.1% Solomon 2000 Israel Ophthalmology clinic Indomethacin 1% Szucs 2000 United States ED Diclofenac 0.1% Vigasioy 1986 Italy Emergency eye clinic Piroxicam 0.5% *Cointervention indicates that another therapy was given to both the treatment and control groups. ytrial with suitable data for analysis. zcontrol group was given two drops of chloramphenicol instead of the one drop received by the treatment group.

ACAD EMERG MED d May 2005, Vol. 12, No. 5 d www.aemj.org 469 Selection. The citations and abstracts identified by the literature search were screened independently by two reviewers (LAC, SB) using predefined selection criteria. Any interobserver discrepancies were resolved after further discussion and reaching mutual agreement. The goal of the citation review was to find all RCTs that studied topical NSAIDs in the setting of traumatic corneal abrasions. We considered published and unpublished trials conducted in humans and also considered abstracts and letters. We included all languages. Corneal ulcers were excluded because this was considered to be a different pathophysiologic process. Likewise, we excluded trials that were conducted in the perioperative setting, because this is a different population of patients. Validity Assessment. The quality of the trials was assessed by each reviewer independently, examining four main components: randomization, double blinding, reporting of withdrawals, and concealment of allocation. A Jadad score 13 was assigned to each trial; this was deemed appropriate given that it has been validated in pain studies. High-quality trials were defined as a score of 3 or greater on the Jadad scale. 13 No attempt was made to blind the reviewers to the authors, journal, or results. Data Abstraction. Two trained reviewers (LAC, SB) independently abstracted data from the retrieved trials onto a standardized and piloted data collection form. Disagreement was once again resolved by consensus. Abstracted variables included study identifying information, population, treatment groups, outcomes, follow-up period, and relevant results. The primary outcome was mean pain scale score at 24 hours, and the secondary outcome was rate of adverse effects. Data Analysis. Pain scales were standardized to a 0 10 scale. For example, to convert scores on a threepoint scale to those on a ten-point scale, scores were multiplied by a factor of 3.33. The implicit assumption is that the relative difference between scores remains the same. The overall summary statistic used (standardized or weighted mean difference) was chosen depending on the uniformity of type of pain scales used in the individual trials. For the primary outcome, a weighted mean difference and 95% confidence interval (CI) was calculated using Review Manager software. 14 A random effects model was used to account for variability among trials. Trials that contained adequate data and were homogeneous in terms of population, design, and outcome were pooled to produce a summary effect. The Cochrane Q statistic was used as a measure of heterogeneity and a guide to selecting the model. A negative mean difference (and hence pain reduction) with a 95% CI that did not cross zero was considered to favor topical NSAIDs. Sensitivity analysis was planned to test the impact of the following variables: type of pain scale, trial quality, type of control, language, country of origin, type of topical NSAID, and any other sources of clinical or methodological heterogeneity. To assess publication bias, a funnel plot was to be constructed if a sufficient number of trials were identified. RESULTS Trial Flow. The results of the literature search are summarized in Figure 1. Of the 200 citations identified, eight RCTs were identified from the MEDLINE search, two from the EMBASE search, and one from conference proceedings. The citation reviewers (LAC, SB) agreed on citation selection 95% of the time. Discrepancies for the remaining 5% were resolved through mutual agreement. Study Characteristics. Study characteristics are displayed in Table 1. The sample sizes ranged from 22 to TABLE 1. (Continued). Cointervention* Control Pain Scale Used Times of Measurement Gentamicin Cointervention only 10-cm VAS 1 hr after first dose, 1 hr after second dose, 24 hr, day 4 Eye patch, chloramphenicol Liquifilm tears 10-cm VAS Every 6 hr for 24 hr Bandage contact lens, Polytrim, Placebo drops Unclear 24 hr cyclopentolate Cyclopentolate, chloramphenicol Liquifilm tears VAS 0 5 24 hr Chloramphenicol Placebo drops 10-cm VAS 24 hr, 48 hr Cyclopentolate, phenylephrine, Placebo drops Verbal pain scale 0 10 24 hr, 48 hr Polysporin None Placebo drops 10-cm VAS 24 hr, 48 hr, 72 hr Bandage contact lens, netilmicin Cointervention only Verbal pain scale 1 10 30 minutes, 12 hr, 24 hr Cyclopentolate, chloramphenicol Cointerventionz Verbal pain scale 0 10 6 9 hr, 18 24 hr and eye patch Gentamicin, oral analgesia Placebo drops 10-cm VAS 2 hr, 4 hr, 24 30 hr, 10 days Chloramphenicol, eye patch Placebo drops Verbal pain scale 0 3 72 hr

470 Calder et al. d CORNEAL ABRASION META-ANALYSIS 347 with a median of 85. Seven of the trials enrolled fewer than 100 patients. One trial 15 was in Italian, and the other ten were published in English. Eight trials were reported in ophthalmology journals, and three trials were reported in emergency medicine journals. 1,16,17 Ten trials were reported as full-text articles and one as an abstract 17 from a conference proceeding. The demographics of the trials were similar; subjects were mostly men (median proportion, 82.5%; range, 73% 100%) in their late 30s (mean [6SD] age, 37 [63.22] years). Two trials 15,18 did not report baseline characteristics of their study population. Eight of the 11 trials were placebo controlled. The source of funding was not stated in six trials; four trials 1,17,19,20 were funded by pharmaceutical companies, with three stating that these were unrestricted grants. One trial 21 was funded by a laboratory. All pain assessments were patient-reported levels of pain, and no trials indicated that their pain measurement tools were validated. Primary Outcome Analysis. Suitable data for analysis were reported in five of the 11 identified trials (Table 1). The requirements for analysis were pain outcome data at 24 hours with raw or mean pain scores and standard deviation or standard errors reported. The reasons for excluding the six trials from the analysis are listed in Figure 1. The primary analysis consisted of three of the five eligible trials because these shared similar methods of pain measurement (verbal pain scales). As such, a weighted mean difference was calculated for a total of 459 patients. The overall weighted mean difference was 21.30 (95% CI = 21.56 to 21.03), suggesting a reduction in pain and favoring treatment with topical NSAIDs. This result was statistically significant (p, 0.00001). The Q statistic was 1.57 (p = 0.46); thus, heterogeneity was unlikely. A visual inspection of the Forest plot in Figure 2 confirms the lack of statistical heterogeneity among the trials. When the two trials using visual analog scales with adequate data were included, the standardized mean difference was 20.52 (95% CI = 20.91 to 20.13). The standardized mean difference summary statistic was used because of the differing pain scales among the pooled trials. Secondary Outcome Analysis. Eight of the 11 trials reported adverse effects. Two trials 18,22 had no adverse effects occur, while four trials 1,15,16,23 recorded that some patients experienced transient stinging with instillation of the drops. One trial 15 provided the actual number of patients (two) with these symptoms. One trial 20 reported three recurrent corneal erosions: two in the control group and one in the NSAID group. One patient developed a corneal abscess in the control group of one trial, 21 while three developed urticaria in the NSAID group. In this latter trial, the NSAID group also received gentamicin drops; therefore, the cause of the urticaria was unclear. Insufficient data were provided to conduct a meta-analysis of adverse effects. Trial Quality. Five of the trials 15 18,21 met Jadad s criteria for high quality. All trials were randomized; however, nine trials 1,15,17 20,22 24 did not describe the methods of randomization. Four trials 1,19,23,24 were not double blinded. Four trials 15,18,21,22 were described as double blinded but did not outline the methods of double blinding. Withdrawals or losses to follow-up were not described for each group by six trials. 16,17,19,20,22,24 Two trials had large losses to follow-up. Brahma et al. 1 conducted their follow-up via postal survey and thus lost 56% of the study group to follow-up. Solomon et al. 23 lost 21% of the control group and 29% of the NSAID group to follow-up. Neither of these trials was included in the metaanalysis. None of the 11 trials reported that their methods of allocation were concealed. Two trial authors 17,21 responded to correspondence and indicated that their method of allocation was concealed, although it was not reported in their published article or abstract. Sensitivity Analyses. When only trials of high quality based on the Jadad score were analyzed, this resulted in a standardized mean difference of 20.22 (95% CI = 20.45 to 0.00). This represents three trials with a total of 299 patients. In terms of quality components, only one trial 21 fulfilled all criteria for randomization, two trials did so for double blinding, 17,20 one did so for description of withdrawals, 15 and none did so for concealment of allocation. Thus, no sensitivity analysis was performed for individual components of trial quality. Sensitivity analyses for placebo-controlled trials, language, and country of origin did not change the statistical significance of the overall estimate. A funnel plot is not reported because few of the trials were suitable for analysis. DISCUSSION Based on the pooling of data from three randomized controlled trials, topical NSAIDs appear to reduce selfreported pain in traumatic corneal abrasions. These trials shared similar populations, study settings, and intervention characteristics. They all measured pain using verbal pain scales. The weighted mean difference of 21.30 is statistically significant and has a narrow CI, but the question of clinical significance requires examination of the pain literature. Previous work 25 has determined that a difference of 1.3 (with a 95% CI of 1.0 to 1.5) on the verbally administered numerical rating scale is considered clinically significant among emergency department patients with acute pain. This was found to be highly correlated with the clinically

ACAD EMERG MED d May 2005, Vol. 12, No. 5 d www.aemj.org 471 Figure 2. Pain scores at 24 hours. significant difference of 1.4 cm (95% CI = 1.1 to 1.7) on the visual analog scale. The authors of the study concluded that a verbally administered numerical rating scale (0 10) can be substituted for the visual analog scale. When the two trials 21 eligible for analysis that used a visual analog scale were included, the statistical significance remained. More than half of the 11 trials identified in our review used visual analog scales, but unfortunately the majority did not report adequate data for meta-analysis. Corneal tissue recovers quickly from abrasions, often in 48 72 hours. Thus, measuring pain outcomes at 24 hours could be considered late in the recovery process. None of the trials examined the immediate impact of the topical NSAIDs. We chose the 24-hour outcome because all of the clinical trials took measurements at this time point. Only four of the 11 trials measured pain at less than 24 hours. None of these were consistent in the time points of pain measurement (i.e., measurements varied from one to 12 hours after presentation). Therefore, even a subgroup metaanalysis was not possible. Because none of the trials reported a mean of less than 1/10 in their placebo groups, we can assume that these patients were still having pain. It is possible that a greater analgesic effect could have been observed if pain outcomes were measured earlier among the trials. Variability in the timing of pain scores was an issue in this analysis, but we also found that the number and type of cointerventions used was variable. Two trials 19,24 used bandage contact lenses that may help reduce pain in corneal abrasions. Both trials distributed this intervention equally among the treatment and control groups; therefore, a differential impact on the outcome is unlikely. However, the overall pain scores may have been reduced compared with other trials. Three trials 1,15,23 had eye patches as cointerventions; two 1,15 gave the patch to both groups, and one 23 provided the patch to the control group and not the treatment group. All of the trials except one 17 used topical antibiotics as a cointervention. The use of these agents is unlikely to have influenced the primary outcome of pain relief but may have influenced the rates of adverse effects in the one trial 21 that identified urticaria among some participants. Overall, the use of cointerventions was varied but, for the most part, applied equally to treatment groups. Thus, these effects are nondifferential. One key finding from this systematic review is the issue of adequate trial reporting. Previous investigators have highlighted this as a concern for the ophthalmology literature. 26 28 In this analysis, six of the 11 identified RCTs presented inadequate data for pooled analysis. These problems ranged from not describing the anchors of the pain scale to not reporting measures of variance. Scherer and Crawley 26 analyzed the use of the CONSORT 29 descriptors in reporting of RCTs among three ophthalmology journals (1991 1994). They found that the mean number of descriptors included by investigators was 15.8 out of a possible 56. In particular, poor reporting was noted in the areas of allocation and masking. This is consistent with our findings. Scherer and Crawley s work was replicated with the journal Ophthalmology for all of the issues in 1999. 27 Improvements in reporting of CONSORT descriptors were found, although several items were still underreported. Similar studies were not found for emergency medicine journals. Our review confirms that efforts to improve trial reporting in both fields are worthy of continued pursuit.

472 Calder et al. d CORNEAL ABRASION META-ANALYSIS LIMITATIONS The limitations of this meta-analysis stem from the exclusion of six trials (473 patients) that could have contributed meaningful data if adequately reported. The use of different pain scales (visual analog, verbal, differing anchors) and different cointerventions contributed to heterogeneity across the 11 trials. The variable levels of trial quality were of concern, especially when the sensitivity analysis demonstrated a near loss of statistically significant effect when trials of lower quality were excluded. An underlying assumption of using the mean to describe the pain data is that the distribution of the pain scores among the study populations was normal. We were unable to confirm that this was indeed the case; it is possible that some of the populations were skewed in terms of their pain scores. There also exists the possibility of publication bias because there were few negative trials identified and unpublished literature was not discovered. The pharmaceutical companies contacted indicated they were unaware of such data. Unfortunately, there were insufficient data to assess this via a funnel plot. The strengths of this meta-analysis include the systematic way in which the literature was searched and the independent review and quality assessment by two reviewers. The homogeneity of the analyzed trials in terms of study population, intervention, and randomized design also contributes to the strength of the results. There are two nonsystematic reviews 30,31 previously published on this topic. These articles were written by the same group, the second as an update of the first. The first article utilized a search of MED- LINE for RCTs, and the second article expanded the search to include the Cochrane Library and ACP Journal Club. The first review identified four of the 11 trials included in the present systematic literature review, and these authors of the review 30 evaluated two of them using the User s Guide to the Literature series 32 to assess validity. There was no pooling of data. The second review identified three additional trials and included five of the trials used in the present systematic literature review. The data were presented in a qualitative manner because these investigators believed that the timing and means of assessing pain reduction prevented meta-analysis. The report 31 identified only two of the five trials ultimately included in the present meta-analysis. Thus, we believe that our systematic literature review and meta-analysis represents an improvement on previous work. CONCLUSIONS While traumatic corneal abrasions are not considered severe injuries, the pain can be quite debilitating. Patients are often unable to drive, read, or perform any work requiring depth perception for 48 72 hours. If this pain were better controlled, patients could return to work sooner and have less disruption of their daily activities. Rather than prescribing several different potentially ineffective treatments, if topical NSAIDs were indeed shown to be effective, this could provide a simpler solution to the management of this common condition. Previous meta-analysis of patching 10 as a treatment modality for corneal abrasions showed no benefit, and our recent survey 11 confirmed that a shift in practice from common to sparing use of this therapy had occurred. Our current meta-analysis demonstrates a positive effect of topical NSAIDs, and we conclude that this therapy can provide effective analgesia for patients with traumatic corneal abrasions. Future areas of investigation could explore the effectiveness of topical analgesia versus oral analgesia. The authors thank Dr. Lorenzo Moja for his assistance with Italian translation and Dr. Christian Vaillancourt for his suggestions in revising the manuscript. References 1. Brahma AK, Shah S, Hillier VF, et al. Topical analgesia for superficial corneal injuries. J Accid Emerg Med. 1996; 13:186 8. 2. Le Sage N, Verreault R, Rochette L. Efficacy of eye patching for traumatic corneal abrasions: a controlled clinical trial. Ann Emerg Med. 2001; 38:129 34. 3. Hulbert MFG. Efficacy of eyepad in corneal healing after corneal foreign body removal. Lancet. 1991; 337:643. 4. Kirkpatrick JNP, Hoh HB, Cook SD. No eye pad for corneal abrasion. Eye. 1993; 7:468 71. 5. Patterson J, Fetzer D, Krall J, Wright E, Heller M. Eye patch treatment for the pain of corneal abrasion. South Med J. 1996; 89:227 9. 6. Kaiser PK, the Corneal Abrasion Patching Study Group. A comparison of pressure patching versus no patching for corneal abrasions due to trauma or foreign body removal. Ophthalmology. 1995; 102:1936 42. 7. Campanile TM, St. Clair DA, Benaim M. The evaluation of eye patching in the treatment of traumatic corneal epithelial defects. J Emerg Med. 1997; 15:769 74. 8. Arbour JD, Brunette I, Boisjoly HM, Shi ZH, Dumas J, Guertin C. Should we patch corneal erosions? Arch Ophthalmol. 1997; 115:313 7. 9. Michael JG, Hug D, Dowd MD. Management of corneal abrasion in children: a randomized clinical trial. Ann Emerg Med. 2002; 40:67 72. 10. Flynn CA, D Amico F, Smith G. Should we patch corneal abrasions? A meta-analysis. Fam Pract. 1998; 47:270. 11. Calder L, Balasubramanian S, Stiell I. Lack of consensus on how to treat corneal abrasions: results of a national survey. Can J Emerg Med. 2004; 6:402 7. 12. Cochrane Reviewers Handbook. In: Alderson P, Green S, Higgins JPT (eds). The Cochrane Library. Chichester, England: Wiley, 2004. 13. Jadad AR, Moore A, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Controlled Clin Trials. 1996; 17:1 12. 14. Review Manager (version 4.2). Cochrane Collaboration, 2002. 15. Vigasio F, Giroletti G. Piroxicam 0,5% topico e corpi estranei corneali. Minerva Oftalmologica. 1986; 28:59 62. 16. Szucs PA, Nashed AH, Allegra JR, Eskin B. Safety and efficacy of diclofenac ophthalmic solution in the treatment of corneal abrasions. Ann Emerg Med. 2000; 35:131 7.

ACAD EMERG MED d May 2005, Vol. 12, No. 5 d www.aemj.org 473 17. Le Sage N, Rochette L, Verreault R. Non-steroidal antiinflammatory and mydriatic in the treatment of traumatic corneal abrasion: a randomized clinical study. Can J Emerg Med. 2004; 6:177. 18. Jayamanne DGR, Fitt AWD, Dayan M, Andrews RM, Mitchell KW, Griffiths PG. The effectiveness of topical diclofenac in relieving discomfort following traumatic corneal abrasions. Eye. 1997; 11:79 83. 19. Donnenfeld ED, Selkin BA, Perry HD, et al. Controlled evaluation of a bandage contact lens and a topical nonsteroidal anti-inflammatory drug in treating traumatic corneal abrasions. Ophthalmology. 1995; 102:979 84. 20. Kaiser PK, Pineda R. The Corneal Abrasion Patching Study Group. A study of topical nonsteroidal anti-inflammatory drops and no pressure patching in the treatment of corneal abrasions. Ophthalmology. 1997; 104:1353 9. 21. Alberti MM, Bouat CG, Allaire CM, Trinquand CJ, The Study Group. Combined indomethacin/gentamicin eyedrops to reduce pain after traumatic corneal abrasion. Eur J Ophthalmol. 2001; 11:233 9. 22. Goyal R, Shankar J, Fone DL, Hughes DS. Randomised controlled trial of ketorolac in the management of corneal abrasions. Acta Ophthalmol Scand. 2001; 79:177 9. 23. Solomon A, Halpert M, Frucht-Pery J. Comparison of topical indomethacin and eye patching for minor corneal trauma. Ann Ophthalmol. 2000; 32:316 9. 24. Patrone G, Sacca SC, Macri A, Rolando M. Evaluation of the analgesic effect of 0.1% indomethacin solution on corneal abrasions. Ophthalmologica. 1999; 213:350 4. 25. Bijur PE, Latimer CT, Gallagher EJ. Validation of a verbally administered numerical rating scale of acute pain for use in the emergency department. Acad Emerg Med. 2003; 10:390 2. 26. Scherer RW, Crawley B. Reporting of randomized clinical trial descriptors and use of structured abstracts. JAMA. 1998; 280: 269 72. 27. Sanchez-Thorin JC, Cortes MC, Montenegro M, Villate N. The quality of reporting of randomized clinical trials published in ophthalmology. Ophthalmology. 2001; 108:410 5. 28. Hawkins BS. The CONSORT Statement: will it lead to improved reporting of clinical trials in ophthalmology? Arch Ophthalmol. 1999; 117:677 80. 29. Begg C, Eastwood S, Horton R. Improving the quality of reporting of randomized controlled trials. JAMA. 1996; 276: 637 9. 30. Brown MD, Cordell WH, Gee AS. Do ophthalmic nonsteroidal anti-inflammatory drugs reduce the pain associated with simple corneal abrasion without delaying healing? Ann Emerg Med. 1999; 34:526 34. 31. Weaver CS, Terrell KM. Update: do ophthalmic nonsteroidal anti-inflammatory drugs reduce the pain associated with simple corneal abrasion without delaying healing? Ann Emerg Med. 2003; 41:134 40. 32. Guyatt GH, Sackett DL, Cook DJ. Users guide to the medical literature II. How to use an article about therapy or prevention. A. Are the results of the study valid? Evidence-Based Medicine Working Group. JAMA. 1993; 270:2598 601.