Adverse Events After Pars Plana Vitrectomy Among Medicare BeneficiariesAdverse Events After Pars Plana Vitrectomy FREE

Pars plana vitrectomy (PPV), introduced by Machemer¹ in 1969, is a common retinal procedure for treating such conditions as diabetic retinopathy, vitreous hemorrhage, macular holes, infectious and inflammatory conditions of the retina, trauma, and cataract operation complications.^{2 - 3} Numerous technological advances have recently been introduced, allowing for more efficient vitrectomy and expanded surgical indications. Such advances include introduction of smaller (25- and 23-gauge) instrumentation, higher frequency vitreous cutters, and adjunctive agents, eg, heavy liquids, intraocular dyes, and intravitreal medications, including triamcinolone and anti–vascular endothelial growth factor agents. As more retina specialists incorporate these newer techniques, determining whether these innovations affect the safety profile of vitrectomy becomes important. Recent studies have found increased rates of endophthalmitis with the smaller-gauge sutureless techniques,^{4 - 9} while others have not.^{10 - 11}

Although studies have documented potential post-PPV complications, no large-scale study has assessed the frequency of these complications in a nationally representative sample or studied changes in complication rates over time. In this study, we analyzed rates of post-PPV complications among Medicare beneficiaries and assessed whether complications rates have changed from 1994 to 2006.

The Medicare 5% database contains detailed claims records from a random sample consisting of 5% of the overall Medicare population obtained from the Centers for the Medicare and Medicaid Services. Medicare 5% inpatient, outpatient, Part B, and durable medical equipment claims files were used to identify a nationally representative sample of Medicare beneficiaries aged 68 years or older who underwent PPV from 1994 to 2006. The data contained information on beneficiaries' demographic characteristics and diagnostic (International Classification of Diseases, Ninth Revision, Clinical Modification[ICD-9-CM])¹² and procedural (Current Procedural Terminology [CPT], version 4)¹³ codes submitted with claims. Durable medical equipment claims files, which contained Healthcare Common Procedure Coding System codes, were used to identify beneficiaries who used low-vision aids. Data were linked by a unique identifier, permitting construction of longitudinal, person-specific data from January 1, 1991, to December 31, 2006.

We identified all individuals who underwent PPV based on these CPT codes: 67036, 67108, 67038-40, 65260, 65265, 67299, 66990, and 66850. Using this sample, we created 3 cohorts consisting of those who underwent their first PPV during 1994-1995 (hereafter, the 1994 cohort), 1999-2000 (1999 cohort), and 2004-2005 (2004 cohort). A 3-year look-back period was used to account for eye-related diagnoses and procedures that occurred before the index procedure (PPV). To ensure that the index procedure was the first PPV performed on a given beneficiary, we excluded individuals who underwent a PPV during the look-back period. We excluded individuals whose PPV occurred before age 68 years because their pre-PPV Medicare coverage was insufficiently long for a complete 3-year look-back. Individuals in Medicare risk plans or living outside of the United States for 12 months or longer during the look-back period were excluded.

We followed up all individuals for 365 days after the PPV or until censoring. Patients were censored, meaning they could no longer be followed up, when they joined a Medicare risk plan, moved outside the United States, died, or underwent another PPV or any subsequent intraocular operation; this mitigated the possibility of misattributing complications to the original PPV.

Adverse events were identified using ICD-9-CMand CPT codes. Complications were classified as (1) severe (endophthalmitis, suprachoroidal hemorrhage, rhegmatogenous, or tractional retinal detachment); (2) less severe (choroidal detachment, vitreous hemorrhage, retinal edema, glaucoma, retinal tear, hypotony, corneal edema, or recurrent corneal abrasions/erosions); (3) an additional surgical procedure (another PPV, retinal detachment repair performed using another technique besides PPV, glaucoma surgery, enucleation/evisceration, vitreous tap, intravitreal injection, laser retinopexy, panretinal photocoagulation, or focal laser); or (4) development of blindness or low vision (eTable)).

We calculated the incidence of the previously listed conditions during the look-back period before PPV. tTests assessed possible differences in adverse event rates between the 1994 and 2004 cohorts and the 1999 and 2004 cohorts. Individuals identified as having had an adverse event listed in the eTable before the PPV were excluded from subsequent incidence calculations and regression analysis for that particular event. For example, for patients who received an endophthalmitis diagnosis before the PPV, no post-PPV endophthalmitis diagnoses were coded as complications of the procedure. For these same patients, however, a new diagnosis of glaucoma after PPV would be considered a surgical complication. Incidence was calculated by dividing the number of individuals who received a first diagnosis of the complication after PPV by the total number of individuals without prior diagnosis of the complication, accounting for censoring. Therefore, if an individual received an endophthalmitis diagnosis following PPV but was censored first, that complication was coded as 0. Alternatively, if the patient received the diagnosis within the 365-day follow-up period before censoring or any subsequent intraocular surgery, the observation was coded as 1.

To calculate whether individuals in the different cohorts were more likely to develop study complications, we performed unadjusted and adjusted time-to-event analyses using Cox proportional hazards model regression. Dependent variables were severe complications, less severe complications, additional operations, blindness or low vision, and endophthalmitis. Individuals were coded as 1 if they developed these complications after vitrectomy and 0 otherwise. Regressions included cohort-identifying variables, with the 2004 cohort as the omitted referent.

In the multivariate analysis, we included a continuous variable for age and binary variables for sex; black race; other race (white, the omitted referent); prior diagnosis of severe complications, less severe complications, diabetes mellitus, proliferative diabetic retinopathy, age-related macular degeneration, eye trauma (open globe or intraocular foreign body), panuveitis, cataract, macular hole, or epiretinal membrane; prior retina surgery; prior intraocular procedure; or same-day procedure (retinal or other intraocular surgery with the PPV).

In a sensitivity analysis, we extended the look-back period from 3 to 5 years for the 1999 and 2004 cohorts to determine whether this would affect the results. Statistical analyses were performed using SAS, version 9.0 (SAS Inc, Cary, North Carolina). P < .05 was considered statistically significant. The Duke University and University of Michigan institutional review boards approved the study.

The total number of beneficiaries who underwent PPV in the 1994, 1999, and 2004 cohorts were 3263, 5064, and 5263, respectively (0.24%, 0.40%, and 0.39%, respectively, of the overall samples). The mean age in the cohorts ranged from 76.4 to 77.3 years. Compared with the other 2 cohorts, the 2004 cohort had a higher proportion of men and individuals of nonwhite and nonblack races (P < .001) (Table 1).

Relatively common pre-PPV retinal diagnoses included age-related macular degeneration, epiretinal membrane, rhegmatogenous retinal detachment, and retinal edema (Table 1). These conditions were not necessarily the indications for PPV (eg, macular degeneration) but were included in the analyses to assess whether they could have influenced post-PPV complication rates. The 2004 cohort had significantly higher proportions of patients with histories of retinal tears, retinal edema, epiretinal membranes, and diabetes mellitus than the 1994 and 1999 cohorts.

Percentages of beneficiaries in the 1994, 1999, and 2004 cohorts with at least 1 severe adverse event diagnosed before the PPV were 33.5%, 26.1%, and 29.6%, respectively. Less severe adverse events occurred in 53.0%, 51.8%, and 60.1% of patients, respectively. Furthermore, during the look-back period, nearly 25% of each cohort had undergone an intraocular operation besides a cataract operation before the PPV.

Post-PPV incidences of severe adverse events were similar among the cohorts (Table 2), ranging from 4.8% to 5.5% (P > .2). Post-PPV rates of endophthalmitis were 0.3% in the 1994 cohort, 0.4% in the 1999 cohort, and 0.4% in the 2004 cohort (P > .2). The post-PPV incidence of blindness or use of low-vision aids ranged from 2.5% to 3.3% (P > .2 for both comparisons).

The proportion of patients with new diagnoses of less severe adverse events was lower in the 1994 (17.3%; P < .05) and 1999 (14.8%; P < .001) cohorts than in the 2004 cohort (20.0%) (Table 2). Rates of additional operations were also lower in the 1994 (22.3%; P < .001) and 1999 (30.2%; P < .001) cohorts than the 2004 cohort (32.3%). The proportion of patients who underwent a second PPV was higher in the 2004 cohort (22.8%) than the 1994 cohort (12.5%) (P < .001). The rate of post-PPV intravitreal injections in the 2004 cohort was 11 times as high as the 1994 cohort (6.6% vs 0.6%; P < .001), and the rate of post-PPV vitreous taps was twice as high in the 2004 cohort (1.4% vs 0.6%; P < .001).

In the multivariate analysis, the likelihood of a severe adverse event did not differ between the 2004 and the 1994 (adjusted¹ hazards ratio [HR], 0.99; 95% confidence interval [CI], 0.78-1.25) or 1999 (adjusted HR, 0.95; 95% CI, 0.77-1.17) cohorts (Table 3). Compared with the 2004 cohort, the risk of a less severe adverse event was lower in the 1994 (adjusted HR, 0.78; 95% CI, 0.66-0.91) and 1999 (adjusted HR, 0.75; 95% CI, 0.65-0.86) cohorts. The likelihood of an additional operation was 40% lower in the 1994 cohort (adjusted HR, 0.60; 95% CI, 0.55-0.66) and 11% lower in the 1999 cohort (adjusted HR, 0.89; 95% CI, 0.83-0.96) than in the 2004 cohort.

Covariates that were associated with an increased risk of a severe adverse event were male sex (adjusted HR, 1.31; 95% CI, 1.10-1.57), trauma (adjusted HR, 1.91; 95% CI, 1.01-3.60), and panuveitis (adjusted HR, 3.64; 95% CI, 1.49-8.85). In contrast, same-day surgery (PPV and another intraocular procedure simultaneously) was associated with a reduced risk of a severe adverse outcome (adjusted HR, 0.69; 95% CI, 0.51-0.92). Patients with a history of a severe adverse event had an increased risk of a less severe event after the PPV (adjusted HR, 1.21; 95% CI, 1.05-1.38). Patients with previously diagnosed macular degeneration (adjusted HR, 0.68; 95% CI, 0.59-0.79) or macular holes (adjusted HR, 0.72; 95% CI, 0.60-0.86) had a reduced risk of a less severe adverse event (Table 3).

Covariates associated with undergoing an additional operation after PPV included having a history of a severe adverse event before the PPV (adjusted HR, 1.23; 95% CI, 1.15-1.32) and a retinal operation (adjusted HR, 1.53; 95% CI, 1.41-1.66) or proliferative diabetic retinopathy (adjusted HR, 1.14; 95% CI, 1.02-1.27). Patients who had another intraocular operation and the PPV performed simultaneously had a 25% reduced risk of an additional operation (adjusted HR, 0.75; 95% CI, 0.67-0.84) (Table 3).

Racial disparities in the likelihood of developing endophthalmitis were observed. The risk of endophthalmitis was nonsignificantly higher in black patients (adjusted HR, 2.32; 95% CI, 0.94-5.68) and significantly higher in patients of other races (adjusted HR, 3.06; 95% CI, 1.18-7.95) compared with white patients. In a sensitivity analysis, increasing the look-back period from 3 to 5 years did not materially affect our results (results not shown).

To our knowledge, this is the first study of adverse events after PPV using a nationally representative sample of Medicare beneficiaries. Rates of severe, sight-threatening complications were relatively low and were unchanged from 1994 through 2006. After adjustment for covariates in the multivariate analysis, rates of severe complications did not differ between cohorts. Rates of less severe adverse events and additional operations were significantly higher in the 2004 cohort than in the 1994 or 1999 cohort.

Several clinical trials and limited-site observational studies have reported post-PPV complication rates (Table 4).^{6 ,8 ,10 - 11 ,15 ,17 - 41} At 0.3% to 0.4%, our rate of endophthalmitis is 10 times as high as the 0.039% reported in an analysis of 15 326 patients from 1984 to 2003.²¹ Studies of endophthalmitis after 20-gauge PPV include 1 that reported a rate of just 0.018%⁶ and another that reported a rate of 0.046%.¹⁹ Two other such studies include a review of 20-gauge PPVs performed at Moorfields Eye Hospital from 1986 to 1990, with an endophthalmitis rate of 0.15%¹⁶ and an analysis of 1500 post-PPV eyes with a rate of 0.2%.¹⁴ Methodological differences may account for the considerably lower rates of endophthalmitis in some studies compared with ours. In most studies, the follow-up time was between 6 and 25 weeks, compared with 1 year in our study. Furthermore, 1 study excluded all late-onset diagnoses of endophthalmitis.¹⁹ Limiting our follow-up to 6 weeks would reduce incident cases of endophthalmitis from 49 to 34, which still yields a higher rate (0.26%) than those of most studies. No study reported mean patient age at the time of PPV, making direct comparisons with our study difficult.

Reported rates in the literature of endophthalmitis after 25-gauge PPV have ranged from 0.03% to 0.84%, into which our observed rates, 0.3% to 0.4%, fall.^{6 ,8 ,11 ,15 ,17 ,19 ,21 ,24 ,32 ,37 ,39} Because smaller-gauge PPV is often a sutureless procedure, some investigators have expressed concern that inadequate closure of the sclerotomies may increase patients' risk of adverse outcomes, eg, postoperative hypotony and endophthalmitis.⁴ While some studies have demonstrated elevated incidence rates of endophthalmitis with smaller-gauge vitrectomy,^{6 ,8} Shimada and colleagues¹¹ found no difference in endophthalmitis rates between conventional 20-gauge and smaller-gauge PPV. Unfortunately, the claims data we used did not specify whether cases used smaller- or larger-gauge instrumentation. Therefore, we cannot know with certainty what percentage of cases performed in 2004-2005 used the newer, smaller-gauge technology. While no study has reported trends in the use of smaller-gauge PPV, a survey by the American Society of Retina Specialists estimated that in 2004-2005 approximately 30% of retina surgeons performed mostly smaller-gauge (25-gauge) vitrectomy.⁴² If we assume that a similar percentage of cases in our 2004 cohort underwent smaller-gauge vitrectomy and that the risk of endophthalmitis is roughly 10-fold higher with smaller-gauge instruments, as some researchers have suggested, our sample size would have had ample power to detect differences in endophthalmitis rates between the 2004 cohort and the 1994 and 1999 cohorts.

After adjustment for other potential factors, the rates of postoperative endophthalmitis did not differ among the cohorts. Although these rates did not differ on the basis of ICD-9-CMcodes, the incidence of post-PPV vitreous tap (CPT code 67015)—a procedure done almost exclusively to confirm an endophthalmitis diagnosis—was higher in the 2004 cohort than in the 1994 cohort. However, the proportion of patients from all cohorts who underwent postoperative vitreous taps was more than double that receiving a diagnosis of endophthalmitis. Therefore, some providers may have used the code for the vitreous tap to bill for diagnoses other than endophthalmitis.

There are several possible explanations for the increase in rates of less severe complications and additional operations from the earlier cohorts to the 2004 cohort. First, the ability to detect certain retinal conditions increased over time. Owing to improvements in imaging technology and quality coupled with a reduction in the price of imaging modalities, the use of diagnostic imaging devices (eg, optical coherence tomography) to identify retinal diseases has risen dramatically.⁴³ The elevated incidence of less severe complications (eg, epiretinal membranes and retinal edema) in the 2004 cohort may partly reflect more widespread use of optical coherence tomography and other imaging devices.⁴⁴ Moreover, it is unsurprising that compared with the 2 other groups, the 2004 group had a 10-fold higher incidence of postoperative intravitreal injections (CPT code 67028)—a common treatment for retinal edema, which is more easily identified using these imaging devices. Intravitreal injections of triamcinolone acetonide and anti–vascular endothelial growth factor agents have also become increasingly popular recently for treating many ocular conditions, eg, diabetes mellitus complications and age-related macular degeneration.^{44 - 48}

Another possible explanation for the increased incidence rates over time is expanded indications for PPV and the emergence of various alternative interventions for residual pathology after PPV. Perhaps cases deemed untreatable after 1 PPV in 1994 were being treated with repeat or other operations in the 2004 cohort.

Despite the apparent improvement in access to PPV during the past decade among older nonwhite patients, rates of post-PPV endophthalmitis among nonwhite patients in our study were higher than those for white patients. After adjustment for covariates, the rate of post-PPV endophthalmitis was more than twice as high in black patients and more than 3 times as high in those of other races than in white patients. Although the difference in endophthalmitis rates between black patients and white patients was statistically insignificant (P = .07), given the devastating vision loss associated with endophthalmitis, this finding may be clinically significant. Similar findings of elevated endophthalmitis rates among nonwhite patients have been observed after cataract and glaucoma operations.^{49 - 50}

The use of large administrative health care databases to study adverse outcomes after intraocular surgery was first described by Javitt and colleagues,⁵¹ who assessed rates of endophthalmitis after cataract operations among Medicare beneficiaries. Others have used Medicare claims databases to study complications after glaucoma operations.⁵² Unlike single-center observational studies, which can be influenced by case mix, surgical technique, and surgeon experience, national health care claims databases capture care provided by many different surgeons and therefore may better represent the surgical experience of patients in the community. Furthermore, unlike observational studies and clinical trials, these databases are often large enough to allow study of rates of infrequent adverse events, such as endophthalmitis. Finally, the longitudinal nature of Medicare claims data enables longer and more uniform follow-up to monitor for postoperative complications.

Use of administrative databases for research purposes limits our study in several ways.⁵³ First, Medicare claims exist to collect payment, not to capture details of clinical encounters. Thus, capturing an adverse event using this database requires the clinician to both identify and bill for the complication. Second, we cannot confirm whether any given event resulted from a particular surgical procedure. Third, because claims data do not specify which eye has undergone surgery or experienced the adverse event, some adverse events in this series could have occurred in the nonsurgical eye. Finally, because only Medicare beneficiaries were studied, our findings may not reflect complication rates of patients who are younger, have another insurance type, receive care through the Veterans Affairs system, or reside outside of the United States.

In summary, among Medicare beneficiaries who underwent PPV, rates of severe complications, including endophthalmitis, were unchanged throughout the past decade, though rates of less severe complications and subsequent intraocular operations increased significantly. Given the aging population, rising prevalence of proliferative diabetic retinopathy⁵⁴ and other indications for PPV, and the increasing influence of imaging techniques for identifying potential PPV candidates, it is important that recently implemented technological advancements designed to improve the efficiency of PPV are not compromising patient safety. Unchanged rates of severe complications suggest that technological advances have not adversely affected patient risk. However, additional studies should further explore the factors that contribute to elevated rates of less severe complications and subsequent intraocular operations, including vitreous taps, and disparities in outcomes among individuals of racial minorities undergoing PPV.

Correspondence: Joshua D. Stein, MD, MS, Department of Ophthalmology and Visual Sciences, University of Michigan, 1000 Wall St, Ann Arbor, MI 48105 (jdstein@med.umich.edu).

Submitted for Publication: March 16, 2009; final revision received May 23, 2009; accepted June 1, 2009.

Financial Disclosure: None reported.

Funding/Support: Partial support for this research came from grant 2R37-AG-17473-05A1 from the National Institute on Aging.

Previous Presentations: This paper was presented at the 2009 Association for Research in Vision and Ophthalmology Annual Meeting; May 7, 2009; Ft Lauderdale, Florida.