ESCRS study of prophylaxis of postoperative endophthalmitis after cataract surgery

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

ARTICLES

ESCRS study of prophylaxis of postoperative endophthalmitis after cataract surgery Case for a European multicenter study David V. Seal, MD, Peter Barry, FRCS, George Gettinby, DPhil, Fiona Lees, MSc, Magnus Peterson, MSc, Crawford W. Revie, MSc, Kirk R. Wilhelmus, MD, PhD, for the ESCRS Endophthalmitis Study Group

PURPOSE: To present the development and design of the European Society of Cataract & Refractive Surgeons multicenter study of the prevention of postsurgical infective endophthalmitis after phacoemulsification and to describe the process for its successful implementation and conduct. SETTING: Twenty-four ophthalmology units and eye clinics in Austria, Belgium, Germany, Italy, Poland, Portugal, Spain, Turkey, and the United Kingdom, with an administrative office in Ireland, coordinating center in England, and data management and statistical unit in Scotland. METHODS: This partially masked randomized placebo-controlled multinational clinical study was designed to evaluate prospectively the prophylactic effect of intracameral cefuroxime and/or perioperative topical levofloxacin on postoperative endophthalmitis after cataract surgery. Random allocation was based on a 2  2 factorial design that included participating centers as a class variable. Real-time electronic data collection monitored study progress and provided weekly outcome tables, monthly recruitment summaries, and quarterly analytical reports for the study’s Data Monitoring Committee, which evaluated the safety and efficacy by Internet-based conferences. RESULTS: A 2-year lead time was required to meet harmonized standards of clinical research in the European Union, obtain ministerial authorization in 3 countries, gain institutional approvals at 24 hospitals, and procure indemnity insurance for surgical centers. Informed consent instruments, designed to comply with national health policies, were translated into 8 languages. The use of information technology to collect study data enabled the organizers to evaluate individual eligibility at enrollment, adherence with study medications during and after surgery, and postoperative status during follow-up. CONCLUSION: This international cooperative study provided the opportunity to estimate the current incidence of endophthalmitis after cataract surgery in Europe and determine whether 1 or both of 2 antimicrobial regimens reduces the risk for postsurgical intraocular infection. J Cataract Refract Surg 2006; 32:396–406 Q 2006 ASCRS and ESCRS

Cataract extraction with intraocular lens (IOL) implantation is the most commonly performed surgical procedure in the elderly population in Europe, with almost 2.5 million operations performed in France, Germany, Italy, Spain, and the United Kingdom in 2003. The frequency varies within European Union (EU) countries, involving 4% to 7% of the population older than 65 years. With an increasing elderly population, the number of those requiring such surgery is Q 2006 ASCRS and ESCRS Published by Elsevier Inc.

396

expected to rise considerably in the near future as a result of increasing longevity and the population’s greater expectation of surgically corrected vision. Although cataract surgery is usually successful in restoring failing eyesight and technical advances have enhanced the efficacy of the procedure, it is also responsible for permanent and significant loss of vision resulting from severe postoperative infective endophthalmitis in up 0886-3350/06/$-see front matter doi:10.1016/j.jcrs.2006.02.014

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

to 0.1% of patients.1–4 As a consequence, an unacceptably high risk for serious postoperative infection with loss of vision remains the most important unsolved problem of the procedure. There is a need to develop safer cataract surgery that will reduce the incidence rate of this complication to much lower levels. In an evidence-based update on bacterial endophthalmitis prophylaxis recently published,5 the authors state that the only method of effective prophylaxis with reliable scientific evidence is the use of povidone–iodine 5% in the conjunctival sac just before surgery.6 Although there is good evidence that povidone–iodine reduces the bacterial numbers in the conjunctival sac by 10- or 100-fold,7,8 the study was retrospective and such studies are notorious for their liability to introduce bias, mainly as a result of selection and poor follow-up of many patients, but also because of confounding factors. Ciulla et al.5 did not find any evidence in randomized or placebo-controlled studies of the benefit of antibiotic use for prophylaxis against postoperative infection, although they did find numerous retrospective, comparative, or single-use studies. All, however, could be suspected of bias or were of insufficient power to provide convincing evidence of benefit. These included reports from high-volume cataract surgeons, 1 of whom claimed only 1 case of

Accepted for publication November 28, 2005. From the European Society of Cataract & Refractive Surgeons (Seal, Barry), Dublin, Ireland; the Department of Statistics and Modelling Science (Gettinby, Lees, Peterson, Revie), University of Strathclyde, Glasgow, United Kingdom; and the Cullen Eye Institute (Wilhelmus), Baylor College of Medicine, Houston, Texas, USA. Funded by the European Society of Cataract & Refractive Surgeons, Dublin, Ireland, with support from Santen GmbH, Germering, Germany. No author has a financial or proprietary interest in any material or method mentioned. The Board of the ESCRS, through its current and recent officers and members, acknowledges several individuals: Peter Barry for conceiving the study and ensuring it became a reality; David Seal for his academic input, tenacity, and voluntary service; and Fiona Lees for meticulous collection and recording of data. The Board acknowledges the contribution of the partners and the doctors, nurses, pharmacists, microbiologists, secretaries, and assistants who have coped magnificently with the additional workload imposed on them by participation in the multicenter study. The Board also acknowledges the promptness and diligence with which the Data Management Committee consistently responded to the electronic information presented to them. In addition, the Board acknowledges the financial contribution of Santen. Finally, the Board acknowledges Mary D’Ardis, CEO, and the ESCRS staff for their unremitting enthusiasm and support. Reprint requests to Peter Barry, ESCRS, Temple House, Temple Road, Blackrock, County Dublin, Ireland. E-mail: [email protected].

postoperative infection in 12 000 patients when vancomycin was added to the irrigating solution.9 A review of the literature1,5,8 and anecdotal evidence from surgeons show that although many techniques to reduce the incidence of postoperative endophthalmitis are being used experimentally, there is still great uncertainty about the benefits of any of them. A major difficulty in gauging the benefit of a proposed method of prophylaxis is that the incidence of endophthalmitis, with good general hygiene precautions in place, is already so low that a very large study is needed to provide the power to detect a significant reduction. This review also considered various survey studies (S. Masket, MD, ‘‘1997 ASCRS Endophthalmitis Study,’’ presented at the ASCRS Symposium on Cataract, IOL and Refractive Surgery, San Diego, California, USA, April 1998).10,11 In general, such studies achieve response rates of approximately 65%, which is too low to eliminate the risk for significant nonresponse bias. Furthermore, nonresponders are typically not polled again to assess why they failed to respond and what their contribution might have been. Thus, apart from the ‘‘standard’’ use of povidone– iodine 5% for preoperative antiseptic treatment of the eye, there is no single best practice recommended for antibiotic prophylaxis against postoperative infection after phacoemulsification surgery. Most, but not all, surgeons use an antibiotic prophylactically to reduce the perceived risk, but none of these supplementary treatments has proved scientifically to be of benefit. Some have used vancomycin in the irrigation fluid or by intracameral injection, but the use of this antibiotic is not recommended (Prophylactic Use of Vancomycin for Intraocular Surgery; a joint statement of the American Academy of Ophthalmology and the Centers for Disease Control and Prevention [online]. Available at http://www.aao.org/education/library/statements/vancomycin. cfm. Accessed January 31, 2006.) as it is reserved for the treatment of methicillin-resistant Staphylococcus aureus (MRSA) endophthalmitis and other MRSA infections.12 Clearly, the time is appropriate for a systematic investigation of the most promising supplementary means of prophylaxis against postoperative endophthalmitis. The gold standard for an investigation of the incidence of such a rare event is a randomized masked prospective placebo-controlled multicenter study. It was with this in view that the European Society of Cataract & Refractive Surgeons (ESCRS) initiated plans for such a study to be conducted on a large scale and on a Europe-wide basis (see Appendix for full list of contributors). MATERIALS AND METHODS Study Design The ESCRS first envisioned the need for this study in 1995, with 3 grant applications made to the EU (1996, 1998, and 2000). However, none of the grant applications was funded, so

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

397

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

industrial support was investigated. In 2001, the ESCRS approached Santen GmbH. Santen agreed to supply 17500 drop bottles of levofloxacin 0.5% and 17 500 of its antibiotic-free vehicle solution (placebo) as well as the antibiotic Oftaquix (levofloxacin 0.5%) for postoperative prophylaxis. The ESCRS Board and Santen agreed to fund the balance of costs on a 70–30 basis. The protocol was written in 2002, and indemnity insurance was purchased in January 2003 as a 3-year policy for 300 000 euros as this was required to make hospital ethical committee applications. Further insurance of 65 000 euros was required for German and Polish centers. The first hospital started to recruit patients for the study in September 2003; however, most partners were not fully active in the study until early 2004. Study Objectives It was decided that 2 main response variables would be analyzed as the primary objectives of the study: (1) the overall number of patients with presumed infectious postoperative endophthalmitis; (2) the number of patients with infectious endophthalmitis as proven by at least 1 of Gram stain, culture, or polymerase chain reaction (PCR). The principal aim was to establish whether a perioperative antibiotic (either or both of intracameral cefuroxime or topical levofloxacin) helps prevent later development of postoperative endophthalmitis. It was realized that the data from the study would be a valuable resource to assess recently suggested risk factors for increased susceptibility. These include the clear corneal incision,2,4,13 increased risk in summer months,14 and decreased risk when foldable IOLs are inserted using a sterile injector.15 Other risk factors considered are given in the ESCRS Guidelines.1 Thus, data collection was organized to make it possible to investigate all these factors at the close of the study. Inclusion and Exclusion Criteria All patients, including diabetics, having routine cataract surgery in each unit were invited to take part except those who were ineligible. The full list of exclusion criteria is given in the study protocol and includes patients who were allergic to penicillins and cephalosporins; those in long-term nursing homes, pregnant, or younger than 18 years; and all groups severely ‘‘at risk’’ for infection such as those with severe atopic keratoconjunctivitis or active blepharitis. Treatments Investigated The initial decision of paramount importance was the choice of antibiotics to test. In this study, it was decided to investigate the effects of 2 drugs: intracameral cefuroxime16,17 and topical levofloxacin.8,18–23 Cefuroxime was chosen for intracameral use as a consequence of the Swedish study of 32 000 patients, in which efficacy and safety were established.16,17 Levofloxacin was chosen for topical use as a third-generation fluoroquinolone because it is well absorbed into the anterior chamber8 and has enhanced antibacterial activity compared to ciprofloxacin and ofloxacin. In patients receiving it, cefuroxime was injected into the anterior chamber at the end of surgery as 1 mg in 0.1 mL saline (0.9%). Those receiving perioperative topical levofloxacin were given 1 drop of levofloxacin 0.5% eyedrops 1 hour before surgery, a second drop half an hour before surgery, and 3 further drops at 5-minute intervals immediately after surgery. The same dosing

398

was applied in those receiving placebo drops. No injection was given to patients not receiving cefuroxime. Povidone–iodine 5% (Betadine) was chosen for antisepsis. All units agreed to change their previous antiseptic procedure so that povidone–iodine was started outside the operating theater and timed to be applied as 1 drop into the conjunctival sac and onto the cornea for a minimum of 3 minutes before surgery. This was the minimum level of antisepsis with many centers performing additional skin cleansing procedures. No preparations used for antisepsis contained detergent. In addition, as a protective measure against postoperative infection, all patients in the study were given levofloxacin 0.5% eyedrops (Oftaquix) starting the morning after surgery and continuing 4 times daily (ideally 6 hourly) for 6 days. There was an approximate 18-hour gap between surgery and the first postoperative dose to allow assessment of the prophylactic effect of the perioperative antibiotics. Statistical Design to Encompass the Objectives It was decided to conduct the study using a 2  2 factorial design to evaluate optimally the 2 interventions described. Each patient recruited into the study was randomly allocated to 1 of 4 treatment groups: A, B, C, or D (Figure 1). The hypothesis under consideration was that neither the intracameral injection of cefuroxime 1 mg at the end of surgery, nor the administration of perioperative levofloxacin 0.5%, drops, nor the combination, has a significant effect on the incidence rate of postoperative endophthalmitis after phacoemulsification in the presence of preoperative antiseptic preparation with povidone–iodine and postoperative levofloxacin 0.5% drops 4 times daily for 6 days, starting at the first postoperative day. This design provided 2 half-factorial comparisons for each treatment separately and 6 pairwise comparisons for the different combinations. To ensure the 4 treatment groups remained balanced throughout the study, and at each center, the randomization code for allocation of patients to the treatment groups was organized in blocks of 12. Power Calculations and Required Study Size A review of current literature suggested that the background incidence rate of endophthalmitis after cataract surgery is of the order of 0.2%.1,2 Accordingly, the statistical design of the study was carried out with a view to having sufficient power to detect a 4-fold reduction of this risk with a significance level of 5%. If such a reduction could be achieved using these treatment regimens, this knowledge may contribute toward a more widespread adoption of simultaneous bilateral phacoemulsification surgery, which would have important medical and economic benefits. Figure 2 shows how the required group sizes would vary if the rate in any ‘‘treatment’’ group dropped below the assumed background incidence rate of 0.2%. In the light of these and similar calculations, it was decided to aim for 8750 patients in each of the 4 treatment groups, leading to a total study size goal of 35 000 patients. One strength of this statistical design is that if there were no evidence of interaction between cefuroxime and perioperative levofloxacin, higher power would be achieved by combining pairs of groups rather than by investigating the effect of each treatment group separately because the group size would be effectively doubled. Thus, even with less recruitment than planned, there was the prospect of detecting any treatment effect that would be of clinical importance for either of the 2 prophylactic treatments separately.

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

Group A

Group B

Placebo vehicle drops x 5*

Placebo vehicle drops x 5*

Povidone–iodine 5%

Povidone–iodine 5%

No intracameral injection

Intracameral cefuroxime 1 mg

Postoperative levofloxacin 0.5%t

Postoperative levofloxacin 0.5%t Figure 1. Treatment groups to which the patients were randomly assigned using a 2  2 factorial design.

Group C

Group D

Levofloxacin drops 0.5% x 5*

Levofloxacin drops 0.5% x 5*

Povidone–iodine 5%

Povidone–iodine 5%

No intracameral injection

Intracameral cefuroxime 1 mg

Postoperative levofloxacin 0.5%t

Postoperative levofloxacin 0.5%t

*Two drops given preop and 3 drops postop tFrom day 1 to 6

Case Recognition and Laboratory Analysis

to capture and report late-presenting cases. Similarly, each clinical partner was required to actively monitor and report cases presenting more than 6 weeks after surgery. The few patients diagnosed with presumed infective endophthalmitis were immediately investigated by anterior chamber and vitreous taps for (1) Gram stain, (2) microbiology culture, and (3) PCR. Patients were also immediately given what the clinician judged to be the best treatment for the condition. The Gram stains and microbiology cultures were performed locally according to each center’s usual arrangements. The PCR tests were performed centrally and replicated independently at 2 centers: the Institute of Medical Microbiology and Hygiene, University of Regensburg, Germany, and the Instituto Oftalmologico, VISSUM, Alicante, Spain. Any case in which 1 or more of these test procedures yielded a positive result would be reclassified as proven infective endophthalmitis. If all test results were negative, the case would be reclassified as nonproven infective endophthalmitis.

Minimum number of patients per group

It is essential in a multicenter study to have a well-defined procedure for the diagnosis of the condition under investigation. Fortunately for this study, endophthalmitis after cataract surgery is such a severe condition that any patient suffering from it will usually seek medical help immediately. Clinicians participating in the study, however, were instructed to be alert for the symptoms and signs associated with endophthalmitis such as pain, swelling, loss of vision, hypopyon, chemotic conjunctiva, and edema of the lids, conjunctiva, or cornea together with vitritis.1 To provide a positive or negative response to the diagnosis of presumed infective endophthalmitis, all clinicians were required to use clinical judgment regarding the symptoms and signs noted above at each patient’s follow-up visit. At least 1 follow-up visit was required for each patient participating in the study, and it typically took place between 3 weeks and 6 weeks after surgery. Earlier follow-up was permitted if the clinician had a mechanism in place 22,000 20,000 18,000

Figure 2. Size of sample required (number per group) to detect a variety of infection proportions in a treatment (running from 1 to 10 cases per 10 000) against a background rate of 0.002 (ie, 20 cases per 10 000) and assuming a 2-tailed test with 80% power at the 5% significance level.24

16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

6

6.5

7

7.5

8

8.5

9

9.5

10

Reduced incidence rate per 10,000 (given an assumed background rate of 20 per 10,000)

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

399

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

All nonproven cases were subsequently reviewed to determine whether some could be identified as toxic inflammatory noninfective uveitis (C/ÿ hypopyon), such as cases caused by toxic anterior segment syndrome,25 when 3 or more cases usually occur together, or phaco-toxic uveitis (often resulting from hypersensitivity to lens protein or a residue of ethylene oxide in the IOL). Such noninfective cases often present acutely within 24 to 48 hours of surgery but usually respond well to treatment with steroids. Any such cases were removed from the total endophthalmitis count. A more difficult problem is posed by late-presenting cases of chronic granulomatous or saccular (bag) endophthalmitis, which is often caused by infection with Propionibacterium acnes, coagulase-negative staphylococci, or diphtheroids. Partners were asked to report all such cases that presented within 6 months of surgery. Interim and Final Analyses for Safety, Efficacy, and Futility Summary statistics regarding cases of endophthalmitis were provided to the chairman, study coordinator, Data Monitoring Committee (DMC), and all other partners and parties on a monthly basis. Monitoring took place at 3-month intervals, and statistical analyses were applied by the information technology (IT) and statistics teams on a confidential basis to provide advice to the DMC only as to whether intervention by the committee may be warranted. In particular, the incidence of presumed and proven endophthalmitis in a treatment group was tested for evidence of a tripling of the expected background incidence rate of 20 infections per 10 000 operations. For every sample size, the significance level (probability of a false alarm) was kept below 1%. At the start of the study, when the available sample was small, this provided low power to detect a tripling of the incidence rate and the power to detect this was reported along with the significance level. As recruitment increased, the available samples became larger and setting the significance level as close below 1% as possible provided a power of detecting a tripled incidence rate of endophthalmitis of much greater than 80%. In this situation, the threshold value at which the alarm was to be triggered was chosen so that it just achieved a power over 80% to detect a tripled incidence rate, which was reported with the lower significance level. Further account was taken of the group-sequential approach and the small number of repeated significance tests (up to 9 repeated analyses) that could take place during the 30-month study period. The scheme was adopted so that the significance level for each interim test was always below 1% and when the sample size became large, it was considerably lower as only 80% power was used to detect a tripling of the incidence rate to 60 per 10 000. Individual centers were similarly monitored using the same criteria and all available cases, or the most recent 2000 cases were used when the number of cases exceeded 2000. Monitoring was also undertaken for early evidence of beneficial treatment effect to prevent delay in acting on early proof of the benefit of any treatment. Such evidence was required to show that not only did the treatment intervention result in a significantly lower incidence of endophthalmitis than the presumed background rate of 20 cases per 10 000, it also resulted in an incidence rate that was significantly lower than that observed in the group or groups of the other interventions because it is possible the assumed background rate was higher than the overall rate achieved in the study. As a consequence, if the test incidence rate was the background rate, a standard lower 1-tailed test of significance of a proportion for the observed incidence rate in that group was carried out using a 1% significance level. If the test incidence rate was

400

the pooled rate in the other 3 groups, a standard 2-tailed test for a difference of proportions between the test group and the other 3 groups was done at the 1% significance level. If in any 1 of the groups the test result was significant, a report was issued to the DMC. Finally, monitoring is undertaken specifically for the effect of the 2 prevalent surgical techniques (clear cornea and scleral tunnel) in the same way as the entire study is assessed for possible increased rates of the incidence of endophthalmitis among treatment groups. At the conclusion of the study, efficacies of treatments will be examined by comparing the incidence of proven and presumed endophthalmitis cases among treatment groups. The analysis technique will be the use of chi-square tests to determine whether the incidences of endophthalmitis in the 4 treatment groups are significantly different at the 5% significance level and to provide estimates and their associated confidence intervals for the incidence rates of endophthalmitis for each of the treatments. A binary logistic regression model will be constructed to estimate odds ratios of risk associated with each treatment group, taking account of any center effects that might be present.26 Administration of a Large Multicenter Multinational Study The preliminary work of preparing a protocol for a study such as this was extensive and time consuming. The protocol took 1 year to write and contained 200 pages. It was required to comply with the EU Clinical Trial Directives 2004 (EU Clinical Trial Directives. The Medicines for Human Use [Clinical Trials] Regulations 2004. Statutory Instruments (SI). No. 1031. 2004 [online]. Available at http://www.opsi.gov.uk/si/si2004/20041031. htm. Accessed January 31, 2006); the World Health Organization Declaration of Helsinki, with particular reference to paragraph 29 for the use of a placebo control; Good Clinical Practice (GCP) regulations; and Investigational Medicinal Products legislation. It also had to conform to the national guidelines of each EU country and Turkey, including the provision of a patient information sheet. A DMC had to be created to monitor the results of the study on an ongoing basis. Negotiations were undertaken by the study coordinator with the medical administrations of numerous EU countries to ensure that all ethical and safety guidelines could be duly observed. This was done in collaboration with each hospital unit and its ethical committee and often, but not always, with the national ministry of health. Appropriate indemnity insurance cover had to be negotiated, with conditions to meet the requirements of all the different legislatures of the 8 EU countries (Austria, Belgium, Germany, Italy, Poland, Portugal, Spain, and United Kingdom) and Turkey. Arrangements for the supply of the medications for the study by Santen Oy (Finland) and for the labeling of the bottles of perioperative eyedrops incorporating the randomization code by Brecon Pharmaceuticals (Wales, United Kingdom) were undertaken by the industrial consultant/pharmaceutical qualified person (QP) in collaboration with the study coordinator. A large-scale study such as this, operating across the continent of Europe, requires a well-designed administrative structure during its conduct if it is to run successfully. The bodies and their respective responsibilities and interactions are shown in Figure 3. The main responsibility for the oversight of the study and the maintenance of satisfactory standards for its conduct rested with the study chairman, who was a board member of ESCRS. This was carried out in close cooperation with the study coordinator,

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

Industry Sponsor

Industrial Consultant / QP

Production of Supply

Medication Supply and Logistics

Clinical Partners

Infectious Disease Partner

Patient Recruitment

Study Co-ordinator

Figure 3. Schematic diagram of the study’s administration.

Study Management

University of Strathclyde

Team

Data Management and Analysis

ESCRS Board Member Study Chairman

Data Monitoring Committee = Interaction

Data Monitoring and Patient Safety

who had the difficult role of maintaining close contact with the clinical partners and ensuring they were able to understand and observe the details of conduct in the study protocol. Monitoring and quality assurance visits were required at least once every 6 months to each unit to ensure that all procedures were being followed correctly. With 24 hospitals in 9 countries taking part in the study, these visits were costly and time consuming. To facilitate communication and exchange of knowledge, partners’ meetings were organized twice a year during the ESCRS Annual Summer and Winter Meetings. Without the resources of modern IT, such a study would be unfeasible. The oversight of that side of the study, combined with the task of organizing the statistical analyses required at various stages, was delegated to small dedicated statistics and IT teams at the University of Strathclyde in Glasgow. Finally, it was important to ensure adequate independent checking of the data as they arrived from the clinical partners to ensure that the study did not continue if it became ethically indefensible.

Data Monitoring Committee and Masking There were no formal guidelines for a DMC in the EU at the time of the study; therefore, the Draft United States Guidelines (www.fda.gov/cber/guidelines.htm) were adopted. Membership of the study’s DMC included statisticians, clinicians, pharmacists, microbiologists, and a lay member to oversee the interests of patients. Data monitoring committees have an important role in clinical studies involving placebos and must be structured so they can monitor progress and be sure that decisions are based on statistical evidence. The study chairman, coordinator, and all clinical partners remained masked27 while the study continued. Reasons for early termination of the study were considered in setting up the remit of the DMC. These were on safety grounds, for strong early evidence of a significant effect, and for evidence of futility of continuing. These were monitored on a monthly basis and tested on a quarterly basis. According to the protocol, while the study was running, the DMC would remain masked. The treatment groups would be identified to the DMC only by code. The DMC would know the number of patients followed up in each group or the number of cases of endophthalmitis that had occurred within each group and at each hospital.

If the DMC decides that the situation requires the masking be broken for them, its request for unmasking must be met. It is their decision thereafter as to whether to remove the mask also for the study chairman, who would assume responsibility for further action.

Informatics Provision for the Study From the outset, the study team realized that the most effective way to capture informative surgical and follow-up data for the planned 35 000 patients was to design a study-specific data entry tool. A database permitting direct data entry in both the operating theaters and the outpatient clinics was created to perform this task. This method of data collection offered several benefits. Patient randomization was performed by the database, eradicating the need for sealed envelopes or other inflexible methods containing the treatment details of each patient. Data were checked automatically for completion and accuracy at the time of entry. Surgical notes could be printed for the patient’s file, eliminating handwritten records. The database allowed for the rapid transmission of study data to the University of Strathclyde for monitoring and interim analysis. Following a pilot study using a prototype database in 1998, the IT team at the University of Strathclyde developed a standalone database application using the widely available database package Microsoft Access 2000. The benefit of selecting this configuration was that the database could run on any Windows-based desktop computer and could use each hospital’s licensed local copy of Microsoft Access. The software was designed for multilingual use and was prepared in Dutch, English, French, German, Italian, Polish, Portuguese, Spanish, and Turkish. The flexible design of the database allowed each hospital to configure the system to meet local requirements. Of the 24 hospitals participating in the study, 21 opted to use the specially designed database. The 3 units that chose not to use the database had in-house data entry systems available to them. Staff in these hospitals were familiar with their own systems and were keen to avoid the burden of double data entry. By working with the study’s IT team, each of these units adapted their in-house systems to meet the study’s data requirements. Capturing the Data The data captured in the operating theater were critical to the success of the study, with the most important piece of data being

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

401

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

Figure 4. Operation data-entry screen.

the antibiotic/placebo regimen given to the patient. Before surgery, each patient was allocated a drop bottle containing levofloxacin 0.5% or its antibiotic-free vehicle solution. A unique sequential subject ID identified each bottle. On entering this ID into the operating theater database, the surgeon was instructed whether to administer an intracameral injection of cefuroxime at the end of surgery or not to do so. The subject ID associated with each bottle was recorded in the operating theater database along with a range of additional information relating to the patient, surgeon, surgery, IOL, operating theater, and phacoemulsification machine. Complications that occurred at the time of the surgery were also documented. Once the operation record was complete (Figure 4), a copy of the surgical note was printed for the patient’s medical file. A follow-up sheet was also produced for the patient’s attendance in the outpatient clinic 1 to 6 weeks later. This sheet contained questions to be completed by medical staff at the follow-up visit and included the following: ‘‘EndophthalmitisdYes or No.’’ The completed follow-up sheet was returned to the unit secretary for entry into the outpatient clinic database. Follow-up was required for every patient. At the end of each week, all new study data were exported to the University of Strathclyde. A simple button click prompted the database to generate an export file that was attached to an e-mail. The database automatically made all data anonymous before they were exported in an encrypted format to ensure data remained confidential. On receiving the data, it was double-checked for

402

completeness and validity. Once any issues were resolved, the data were safely stored in the study’s master database. If it was necessary to make amendments to the study data, the change was agreed with the hospital involved and documented in the master database. Follow-up The surgical data recorded for each patient only became useful once the outcome of the surgery had been documented at a follow-up visit. Typically, study patients attended an outpatient clinic at the hospital at which the surgery was performed. However some large centers found that patients did not always return to the same unit but may have attended an outpatient clinic closer to home for follow-up. In these cases, it was necessary for staff at the operating hospital to telephone the patient and/or the patient’s physician to obtain the necessary follow-up data. Ongoing statistical analyses were based on the number of patients who had been followed up rather than the number of patients who had been recruited. Because of their importance, data on endophthalmitis cases were reported promptly by telephone or facsimile. If nonendophthalmitis follow-ups were delayed, the distortion in the data could show the rates of infection to be beyond safety thresholds when in fact they were not. To overcome the above problem, overdue follow-up reports were regularly sent to each clinical partner for action. Centers were also asked to make sure that patients who were followed

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

up for less than 1 week after surgery were rechecked at a later date to ensure that late-presenting cases of endophthalmitis were not missed. Supply and Distribution of the Drugs A modern multicenter study requires drugs to be labeled professionally with a high standard of quality control and quality assurance. This is set against a background of ever-increasing legislation requiring release by a QP, and since the introduction of the EU Clinical Trial Directives 2004 (EU Clinical Trial Directives. The Medicines for Human Use [Clinical Trials] Regulations 2004. Statutory Instruments [SI]. No. 1031. 2004 [online]. Available at http://www.opsi.gov.uk/si/si2004/20041031.htm. Accessed January 31, 2006), the approval of each EU National Ministry of Health. The drop bottles were manufactured according to full pharmaceutical GMP requirements. The task of labeling was outsourced to a pharmaceutical company specializing in this area. Labels for boxes included drop-bottle subject IDs and their expiration date. Instructions for use were given in the local language. Labels for drop bottles were given 4 codes: date of labeling, expiration date of drug/placebo, code number of batch manufacture, and the patient’s unique subject ID. The crucial subject ID was linked to the contents of the bottle (levofloxacin or vehicle) and to the centrally held randomization file, which dictated whether an injection of cefuroxime should be given to the patient at the end of surgery. The address and telephone number of the sponsor were also given. Telescopic labels were used to accommodate the required 9 languages. DISCUSSION

Taban et al.2 performed a metaanalysis of 215 studies that addressed endophthalmitis after cataract surgery and met their selection criteria. A total of 3 140 650 cataract extractions were pooled from extracapsular cataract extraction and phacoemulsification surgery, giving an overall incidence of 0.128% for postoperative endophthalmitis. They found this incidence varied with time, from 0.265% in 2000 to 2003, 0.087% in the 1990s, 0.158% in the 1980s, and 0.327% in the 1970s. However, Taban et al. also reviewed the limitations of this metaanalysis study, which depended mostly on retrospective studies with limited statistical power. They commented on the scarcity of prospective randomized case-controlled studies and on the unreliability of estimates derived from retrospective studies, many of which had differing definitions of crucial variables, such as the criteria for diagnosing endophthalmitis. West et al.28 make a similar point. They note that large-scale administrative databases are typically limited to establishing prevalence rate estimates and determining temporal trends. Furthermore, in a recent UK study to demonstrate the benefit of electronic data collection, it was noted that efficient collection and analysis of data ‘‘require specialty-specific clinical systems.’’29 These observations underline the timeliness of the present study, which has the potential to provide a much needed reliable set of

data on the epidemiology of this condition in Europe at present as well as to give excellent guidance on the best way to combat it. Finding volunteer surgeons for a multicenter study was time consuming and difficult. It involved the recruitment of surgeons who were not using intracameral antibiotics as well as surgeons already using intracameral antibiotics and irrigation fluid containing antibiotics (including in some cases vancomycin). In practice, the ESCRS study has included surgeons in both categories. Recruiting a large number of patients was challenging. Meeting GCP requirements is a prerequisite for a study to be recognized by today’s standards of evidence-based medicine. These ensure that all patients are given a patient information sheet and are asked to take part in the study. Recruitment is most successful in units in which all patients are asked to participate as part of the routine practice of the unit. Even then, not all patients wish to ‘‘volunteer’’ to join the study. The original goal was to enlist 75% of those asked to do so, before any exclusions were investigated. However, this goal was not achieved at many hospitals. For this reason, it was necessary to include more hospital units than originally planned. In practice, some hospitals found it easier to recruit volunteer patients than others. Our experience was similar to that described in recent research30; that is, having a staff member ask the patient to take part in a study was more successful than any other approach (eg, a letter). Flory and Emanuel30 also found that asking the patient to participate in a study was better than using a multimedia approach (eg, CD-ROM disk). Of the many factors influencing recruitment success, the most important are leadership of the clinical team and disposition of the clinical staff to participate in clinical studies. It is recognized as best practice that in all studies with a control treatment, both patients and clinical staff should be kept in ignorance of which of the different treatments is being given to which patient. As far as possible, this policy was followed in the present study. The mask was maintained for all clinical staff and for the study chairman and study coordinator absolutely. The IT staff held the key to the masking and could break the mask at any time if it became ethically necessary. The masking of the DMC was important to its members’ ability to perform their function. A secure online forum was created for the study, and all weekly, monthly, and quarterly reports were posted to this forum. Proving they had the correct permission, the DMC and other members of the study team could access the reports they were entitled to view through this forum and post comments or questions. This was crucial to the ability of the DMC to perform its main function. There have been many lessons learnt in setting up a large multicenter multinational study, and most tasks would be approached differently in the future, but this is

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

403

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

the essence of good quality clinical researchdto explore, improvise, investigate, and finally to analyze data in a reliable, statistically sound way. If evidence-based medicine is to be advanced through the use of prospective randomized masked placebo-controlled multinational studies, it is important to be familiar with current legislation and not to underestimate the challenges. In this study, a constant challenge was securing cost-effective indemnity insurance that appropriately covered the period patients were recruited in each country, yet met stringent regulatory requirements for cover to be in place before regulatory approval could be given. The advantages of working within a medical (or surgical) society became obvious during the inception and implementation of the study. The exhaustive work of study coordination was given voluntarily to the society by the study chairman and coordinator. The society’s officers also provided unstinting administrative support, in particular in securing indemnity insurance cover. The clinical partners at each hospital gave their time to implement, support, and run the study with payments provided only for secretarial expenses. It is the umbrella membership of the society that allowed this study to progress. It also provided through its twice yearly meetings a natural forum at which the essential face-to-face contact between clinical partners and the others running the study could take place without the prohibitive expense of calling staff from across Europe for short 1-off meetings. This is unique at the present time and, to our knowledge, no such large multicenter study has been performed by a medical membership society with substantive support from members’ funds. It is recommended as a model for other European-based medical societies to follow. Additional financial support from government institutions or industry, provided that full independence is retained as in this study, is almost a prerequisite to enabling a costly project of this type to be conducted. The ESCRS study was scheduled to continue recruiting patients until March 31, 2006, with follow-up completed by May 31, 2006. However, the study was terminated early, in January 2006, for reasons described in an article detailing the results.31 As all data were electronically collected and handled and checked on arrival at the data-coordinating center, it is expected that final results will be presented at the ESCRS congress in London in September 2006. This will represent a record time for producing a final analysis of such a large multicenter multinational study. During the time the study was underway, the issue of compulsory registration of clinical trials once again became an area of public focus.32 A statement by the International Committee of Medical Journal Editors in September 2004 promoted the registration of all trials starting recruitment after July 1, 2005, while ongoing trials were also strongly advised to register (International Committee of Medical

404

Journal Editors. Is This Clinical Trial Fully Registered? A statement from the International Committee of Medical Journal Editors, 2005 [online] Available at http://www. icmje.org/clin_trialup.htm. Accessed January 31, 2006). In line with these recommendations, the study was registered with ClinicalTrials.gov and was allocated the unique trial identifier NCT00136344. APPENDIX The ESCRS Endophthalmitis Study Group Study Management Team Peter Barry (study chairman), Royal Victoria Eye & Ear Hospital and St. Vincent’s University Hospital, Dublin, Ireland; David Seal (study coordinator), Applied Vision Research Centre, City University, London, United Kingdom; George Gettinby, Magnus Peterson, Crawford Revie, and Fiona Lees, Department of Computer and Information Sciences/Department of Statistics and Modelling Science, University of Strathclyde, Glasgow, United Kingdom; Mary Dardis and Betsan Bradley ESCRS, Dublin, Ireland.

Pharmaceutical Associates Klaus Geldsetzer, Santen GmbH, Germering, Germany; David Lloyd, Renaissance Health Care, Surrey, United Kingdom.

Optometric Advisor Alison Finlay, Applied Vision Research Centre, City University, London, United Kingdom.

Clinical Partner Ophthalmologists Austria: Guenther Grabner, Stefan F. Egger, and Josef Ruckhofer, University Eye Clinic, Paracelsus Private Medical University, Salzburg. Belgium: Johan Blanckaert, Jan Yperman Ziekenhuis, Campus Zwarte Zuster, Ieper; Camille Budo, Oogheelkunde, Melveren; Albert Galand, Jessica Crommen, Jean Rakic, Gae¨l Xhauflaire, Centre Hospitalier Universitaire, Liege; Marie-Jose´ Tassignon, UZA Ophthalmology, Edegem; Hugo Verbraeken and Rita de Donker, Universitair Ziekenhuis, Gent. Germany: Stefanie Schmickler, Augenklinik Ahaus, Ahaus. Italy: Roberto Bellucci, Simonetta Morselli, and Sandro Soldati, Ospedale Borgo Trento, Verona; Fausto Vigasio, Marco Bertelli, Andrea Bottoli, Marta Cassamali, Fabrizio Danieli, Samer Khuri, and Luigina Rosa, Azienda Ospedaliera di Desenzano del Garda, Desenzano del Garda. Poland: Jerzy Szaflik and Justyna Izdebska, Kierownik Katedry i Kliniki, Samodzielny Publiczny Kliniczny Szpital Okulistyczny, Warsaw. Portugal: Concieca˜o Lobo, Jose Cunha Vaz, and Joaquin Mira, Department of Ophthalmology, Coimbra. Spain: Augusto Abreu, Jose Aguilar, Victor Arteaga, Luis Cordoves, Valentin T. Dı´az-Aleman, Manuel Gonzalez de la Rosa, and Cristina Mantolan, Servicio de Oftalmologia, Hospital Universitario de Canarias, Tenerife; Pedro Abreu, Jorge Alvarez-Marin, and Maria Antonia Gil, La Candelaria University Hospital, Tenerife; Jorge Alio´, Instituto Oftalmologico VISSUM, Alicante; Miguel Teus, M.T. Alvarez, and J.M. Roma´n, Hospital Oftalmologico Internacional, Madrid. Turkey: Suleyman Kaynak, Retina Ophthalmic Research Centre, Izmir. United Kingdom: David Allen, Peter Phelan, David Steel, and Chris Wood, Sunderland Eye Infirmary, Sunderland; Carol Cunningham, Michael Miller,

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

Ramesh Moorthy, Andleeb Zafar, Moorfields Eye Outreach Unit, Northwick Park Hospital, Harrow; Alex Ionidis, Damian Lake, and Graham Thompson, Moorfields Eye Outreach Unit, St. George’s Hospital, London; John Jacob, Daniel Byles, Casper Gibbon, Andrew Kleinschmidt, Roland Ling, Anthony Quinn, Peter Simcock, and George Sturrock, West of England Eye Unit, Royal Devon & Exeter Hospital (Wonford), Exeter, and Axminster Hospital, Axminster; William Kiel, Ipswich Hospital, Ipswich; Denise Mabey, David Spalton, and Anupma Kumar, Department of Ophthalmology, St. Thomas’ Hospital, London; Paul Rosen, C.K. Patel, and John Salmon, Oxford Eye Hospital, Oxford.

Ophthalmology Nurse Managers and Administrative Assistants Belgium: Eveline Callens, Gent; Danny Mathysen, Antwerp; Franc¸oise Molemans, Ge´rald de Rassenfosse, and Christel Schenkeveld, Melveren. Germany: Andrea Eckelmann and Andreas Haselhoff, Ahaus. Poland: Renata Franczuk, Warsaw. Portugal: Liliana Carvalho and Ana Catarina, Coimbra. Spain: Laurent Bataille and Elena Jime´nez, Alicante; Ana Martı´n de Nicola´s, Madrid. Turkey: Hakan Kuheylan, Izmir. United Kingdom: Sue Bovill and Ian Tate, Sunderland; Suzanne Cabral and Tim Withers, London; Deborah Cox and Trudi Yeates, Exeter and Axminster; Jean Dash, Linda Lindsell, and Rebecca Turner, Oxford.

Molecular and Microbiologists Consuelo Ferrer, Instituto Oftalmologico VISSUM, Alicante Spain; Roland Koerner, Sunderland Royal Hospital, Sunderland, United Kingdom; Udo Reischl and Anke Behr, Institute of Medical Microbiology and Hygiene, University of Regensburg, Regensburg, Germany.

Data Monitoring Committee Steven Barrett (chairman), Chelsea & Westminster Hospital, London, United Kingdom; Susanne Gardner, St. Joseph’s Hospital/Research, Atlanta, Georgia, USA; Susan Kennedy, National Ophthalmic Pathology Laboratory, Royal Victoria Eye and Ear Hospital, Dublin, Ireland; John Ludgate, London, United Kingdom; Per Montan, St. Erik’s Hospital, Stockholm, Sweden; Kirk Wilhelmus, Department of Ophthalmology, Cullen Eye Institute, Baylor College of Medicine, Houston, Texas, USA.

REFERENCES 1. ESCRS Guidelines on Prophylaxis, Management and Therapy of Postoperative Endophthalmitis. Dublin, European Society of Cataract & Refractive Surgeons, version 1, 2005 2. Taban M, Behrens A, Newcomb RL, et al. Acute endophthalmitis following cataract surgery; a systematic review of the literature. Arch Ophthalmol 2005; 123:613–620 3. Khan RI, Kennedy S, Barry P. Incidence of presumed postoperative endophthalmitis in Dublin for a 5-year period (1997–2001). J Cataract Refract Surg 2005; 31:1575–1581 4. Nagaki Y, Hayasaka S, Kadoi C, et al. Bacterial endophthalmitis after small-incision cataract surgery; effect of incision placement and intraocular lens type. J Cataract Refract Surg 2003; 29:20–26 5. Ciulla TA, Starr MB, Masket S. Bacterial endophthalmitis prophylaxis for cataract surgery; an evidence-based update. Ophthalmology 2002; 109:13–24 6. Speaker MG, Menikoff JA. Prophylaxis of endophthalmitis with topical povidone-iodine. Ophthalmology 1991; 98:1769–1775

7. Min˜o de Kaspar H, Chang RT, Singh K, et al. Prospective randomized comparison of 2 different methods of 5% povidone-iodine applications for anterior segment intraocular surgery. Arch Ophthalmol 2005; 123:161–165 8. Peyman GA, Lee PJ, Seal DV. Endophthalmitis Diagnosis and Management. London, New York, NY, Taylor & Francis, 2004 9. Gimbel HV, Sun R, de Brof BM. Prophylactic intracameral antibiotics during cataract surgery; the incidence of endophthalmitis and corneal endothelial cell loss. Eur J Implant Refract Surg 1994; 6:280–285 10. Schmitz S, Dick HB, Krummenauer F, Pfeiffer N. Endophthalmitis in cataract surgery; results of a German survey. Ophthalmology 1999; 106:1869–1877 11. Morlet N, Gatus B, Coroneo M. Patterns of peri-operative prophylaxis for cataract surgery: a survey of Australian ophthalmologists. Aust NZ J Ophthalmol 1998; 26:5–12 12. Centers for Disease Control. Recommendations for preventing the spread of vancomycin resistance. MMWR Recomm Rep 1995; 44(RR-12):1–13 13. Wallin T, Parker J, Jin Y, et al. Cohort study of 27 cases of endophthalmitis at a single institution. J Cataract Refract Surg 2005; 31:735–741 14. Rubio EF. Climatic influence on conjunctival bacteria of patients undergoing cataract surgery. Eye 2004; 18:778–784 15. Mayer E, Cadman D, Ewings P, et al. A 10 year retrospective study of cataract surgery and endophthalmitis in a single eye unit: injectable lenses lower the incidence of endophthalmitis. Br J Ophthalmol 2003; 87:867–869 16. Montan PG, Wejde G, Koranyi G, Rylander M. Prophylactic intracameral cefuroxime; efficacy in preventing endophthalmitis after cataract surgery. J Cataract Refract Surg 2002; 28:977–981 17. Montan PG, Wejde G, Setterquist H, et al. Prophylactic intracameral cefuroxime; evaluation of safety and kinetics in cataract surgery. J Cataract Refract Surg 2002; 28:982–987 18. Jensen MK, Fiscella RG, Crandall AS, et al. A retrospective study of endophthalmitis rates comparing quinolone antibiotics. Am J Ophthalmol 2005; 139:141–148 19. Koch H-R, Kulus SC, Roessler M, et al. Corneal penetration of fluoroquinolones: aqueous humor concentrations after topical application of levofloxacin 0.5% and ofloxacin 0.3% eyedrops. J Cataract Refract Surg 2005; 31:1377–1385 20. Bucci FA Jr. An in vivo study comparing the ocular absorption of levofloxacin and ciprofloxacin prior to phacoemulsification. Am J Ophthalmol 2004; 137:308–312 21. Colin J, Simonpoli S, Geldsetzer K, Ropo A. Corneal penetration of levofloxacin into the human aqueous humour: a comparison with ciprofloxacin. Acta Ophthalmol Scand 2003; 81:611–613 22. Raizman MB, Rubin JM, Graves AL, Rinehart M. Tear concentrations of levofloxacin following topical administration of a single dose of 0.5% levofloxacin ophthalmic solution in healthy volunteers. Clin Ther 2002; 24:1439–1450 23. Graves A, Henry M, O’Brien TP, et al. In vitro susceptibilities of bacterial ocular isolates to fluoroquinolones. Cornea 2001; 20:301–305; errata, 546 24. Zar JH. Biostatistical Analysis, 4th ed. Upper Saddle River, NJ, Prentice Hall, 1999 25. Monson MC, Mamalis N, Olson RJ. Toxic anterior segment inflammation following cataract surgery. J Cataract Refract Surg 1992; 18:184–189 26. Hosmer DW Jr, Lemeshow S. Applied Logistic Regression. New York, NY, John Wiley, 1989 27. Ellenburg SS, Fleming TR, DeMets DL. Data Monitoring Committees in Clinical Trials: a Practical Perspective. New York, NY, John Wiley, 2002 28. West ES, Behrens A, McDonnell PJ, et al. The incidence of endophthalmitis after cataract surgery among the U.S. Medicare population increased between 1994 and 2001. Ophthalmology 2005; 112:1388–1394

J CATARACT REFRACT SURG - VOL 32, MARCH 2006

405

ESCRS STUDY ON POSTOPERATIVE ENDOPHTHALMITIS

29. Johnston RL, Sparrow JM, Canning CR, et al. Pilot National Electronic Cataract Surgery Survey: I. Method, descriptive, and process features. Eye 2005; 19:788–794 30. Flory J, Emanuel E. Interventions to improve research participants’ understanding in informed consent for research; a systematic review. JAMA 2004; 292:1593–1601

406

31. Barry P, Seal DV, Gettinby G, et al. ESCRS study of prophylaxis of postoperative endophthalmitis after cataract surgery; preliminary report of principal results from a European multicenter study. J Cataract Refract Surg 2006; 32:407–410 32. Abbasi K. Compulsory registration of clinical trials. BMJ 2004; 329: 637–638

J CATARACT REFRACT SURG - VOL 32, MARCH 2006