Supplements Managed Care Issues in Glaucoma: Emerging Trends and Treatment
The Prevalence of Glaucomatous Risk Factors in Patients From a Managed Care Setting: A Pilot Evaluation
Objective: To report percentage of patients who present with glaucomatous risk factors (RFs) and determine average cumulative 5-year glaucoma progression risk in a subset of ocular hypertensive patients using the Ocular Hypertension Study (OHTS) predictive risk scoring system.
Study Design: A retrospective chart review of patients treated at a large ophthalmology clinic for International Classification of Diseases, Ninth Revision glaucoma-related diagnoses.
Methods: Medical records were screened for demographic, clinical, and ocular RFs. Data were collected on ocular attributes, and descriptive, cross-tabulation, and regression statistics were applied to depict prevalence and potential influence of RFs. For the ocular hypertension subset, the OHTS risk scoring system calculated average cumulative 5-year risk of glaucoma conversion.
Results: With 1189 eligible medical records available, univariate analyses demonstrated significant associations between older age and mean deviation (MD), vertical cup-to-disc ratio (CDR), intraocular pressure, and central corneal thickness (CCT). The presence of diabetes mellitus had a protective effect. Regression analyses identified age, mean CCT, and CDR as significant predictors of MD, whereas CCT was the strongest predictor of vertical CDR. Mean composite OHTS score was 9.7, signifying a 15% cumulative 5-year risk for developing glaucoma.
Conclusions: Known RFs were present in approximately one third of patients. Although analyses confirmed the predictive value of these RFs, existing models may not account for several important RFs or may rely on ocular metrics that might not be routinely or practically assessed in clinical practice. Additional studies are required to incorporate these considerations into future predictive models to enhance their clinical application and the interpretation of risk estimates.
(Am J Manag Care. 2008;14:S28-S36)The importance of the risk factor (RF) awareness in ophthalmology has been underscored in recent years by the heightened availability of validated predictive models estimating the risk of conversion from ocular hypertension (OHT) to glaucoma.
The goals of RF analysis are to identify patients who will most likely benefit from early treatment, make judicious decisions regarding when to initiate treatment, and determine how aggressively treatment goals should be set.1 The Ocular Hypertension Treatment Study (OHTS), a pivotal prospective trial comparing the outcomes of treatment versus observation in subjects with OHT, demonstrated that reducing intraocular pressure (IOP) by at least 20% decreased the cumulative probability of developing glaucoma after 5 years from 9.5% to 4.4%.2,3 The results of this study appeared to indicate that a relatively smaller percentage of untreated patients with OHT will develop glaucoma compared with those not progressing. Yet, multiple population studies have approximated that 3 million to 6 million individuals in the United States alone (including 4%-10% of adults older than 40 years) harbor IOPs of at least 21 mm Hg without overt clinical signs of glaucomatous damage.4-7 Extrapolating the likelihood of glaucoma progression to these figures, it is estimated that without intervention, nearly 0.3 million to 0.6 million Americans are potentially at risk for glaucomatous injury within a 5-year period.
In addition to ascertaining the link between reduction in IOP and the prevention or delay of glaucoma onset, a landmark contribution of the OHTS was the identification of significant RFs for progressing from OHT to glaucoma. RF recognition by the clinician is paramount to instituting timely treatment as well as avoiding the potential risks, inconvenience, and expense of unwarranted intervention. The manner in which ophthalmologists apply their knowledge of RFs in clinical practice, however, continues to evolve. Managing patients with glaucoma has traditionally relied on the following sequence of events: (1) detecting structural or functional damage; (2) establishing a target IOP to reduce or deter damage; (3) initiating treatment; and (4) monitoring for disease progression.1 Because glaucomatous damage is irreversible, however, treatment decisions based reactively on indicators of disease progression rather than RFs may result in substantial vision loss for high-risk patients who might have otherwise avoided glaucomatous injury through timely intervention.
We conducted a retrospective review and analysis of medical records for 1189 patients attending an ophthalmology clinic of a large, urban, managed care organization in Los Angeles, California. The aim of this study was to determine the percentage and characteristics of patients presenting with glaucomatous RFs in this sample and to calculate the average predicted 5-year probability of glaucoma conversion for patients with OHT. For the former analysis, we explored and described the potential associations between demographic, clinical, and ocular RFs for glaucoma progression using cross-tabulation statistics and regression methods. For the latter, we employed he predictive model developed by the OHTS group, which was derived from the findings of the OHTS. This model was developed based on attributes of the observation group in the OHTS and validated recently in the placebo group of the European Glaucoma Prevention Study (EGPS).2,4,8 The pooled predictive model is comprised of a simple risk scoring system, which may be feasibly adopted by the clinician, and has been found to be reasonably discriminatory for estimating the 5-year risk of conversion from OHT to glaucoma.
Methods and Materials
The electronic medical records database was searched, and all medical records for patients treated under the International Classification of Diseases, Ninth Revision (ICD-9) global code index 365.x for glaucoma and glaucoma-related diagnoses (eg, OHT and glaucoma suspect) were eligible for inclusion. All patients had been examined on at least 1 occasion with the initial visit occurring between June 2000 and May 2005.
Based on the evidence from the OHTS and other published clinical trials,2,9-17 we identified 15 RFs that have been reputedly associated with glaucoma progression (Table 1). The medical records were screened for the following RF information: age, racial descent, family history of glaucoma (ie, parents and/or siblings), and documented diagnoses of coexistent diabetes mellitus (DM), systemic hypertension, cardiovascular disease, migraine, or vasospasm. Cardiovascular disease, systemic hypertension, DM, and migraine were considered present if a clear documentation of diagnosis or related treatment was recorded on the medical chart. Additionally, we collected data on reported ocular measurements of IOP, vertical cup-to-disc ratio (CDR), central corneal thickness (CCT), and visual field indices, including pattern standard deviation (PSD), and mean deviation (MD). Documentation of the presence or absence of myopia greater than –3 diopters, pseudoexfoliation, and optic disc hemorrhage was also recorded.
All statistical analyses were performed with the aid of Statistical Package for the Social Sciences (SPSS, Inc, Chicago, IL) for Windows (version 14.0, Microsoft Corporation, Redmond, WA). Missing data were handled using listwise deletion when the pattern of missing data was determined to be random and the number of missing cases did not exceed 5%. For data not meeting these criteria, an expectation maximization (EM) algorithm (SPSS, 1999) was employed. The EM algorithm consists of a 2-step process. First, the expected values missing observations are computed using regression equations based on the observed data. Then, the missing values are replaced by the conditional means derived from the regression equations.18 Statistical significance was defined as P <.05.
We applied descriptive statistics to indicate the frequency of demographic and clinical RFs. Measurements of age and ocular indices, including IOP, MD, vertical CDR, CCT, and PSD, were reported as the mean ± 1 standard deviation (SD) and analyzed as continuous variables. For cases documenting ocular metrics for the right and left eyes obtained on 1 or more visits, the calculated mean ocular measurements represented the average of the mean respective values for the right and left eyes. Demographic and clinical parameters, which were characterized as categorical or nominal data, were expressed as number counts and the percentages of the total number of patients.
In exploring the data, we addressed the following research questions for the study sample:
1. Are there any identifiable patterns of association among demographic, clinical, and ocular RFs?
2. What is the relative influence of nonocular and ocular RFs in predicting visual field loss and optic nerve damage as quantified by MD and CDR, respectively?
3. What is the 5-year risk of progression to glaucoma for patients with OHT?
In addressing the first question, we determined the distribution of RFs among patients by dividing the study sample into subgroups using the presence or absence of demographic, clinical, and ocular RFs as stratification variables. Threshold values for defining the presence of risk versus the absence of risk for ocular RFs were based on standards established in the published literature2,9-17 or by expert opinion provided by practicing ophthalmologists19 (Table 1). A preliminary series of univariate analyses was performed, and nonocular and ocular RFs demonstrating statistically significant associations were cross-tabulated in separate two-by-two contingency tables. The strength and significance of associations between the presence of these demographic and/or clinical RFs and the presence of ocular RFs were assessed using χ2 statistics and expressed as a phi coefficient for nominal variables.
For the second investigation, 2 separate multiple regression models, Model 1 and Model 2, were analyzed using a stepwise selection method. The explanatory variables consisted of the RFs from the prior analysis. The mean bilateral MD and vertical CDR were modeled as dependent variables in Model 1 and Model 2, respectively, and were controlled for in Model 2 and Model 1, respectively. These ocular measures were elected as dependent variables because of their dual association as indicators of disease severity, and given the fact that rather than true RFs, these parameters herald signs of early glaucomatous damage.4 Nominal data for demographic and clinical RFs were coded as binary variables, with “0” and “1” signifying the absence and presence of the RF, respectively, whereas age and ocular RFs were analyzed as continuous variables. Goodness of fit and the assumption of independent errors were assessed via R2 and the Durbin-Watson statistic, respectively.
For the final inquiry, we calculated the 5-year risk for progression to glaucoma for the subset of patients with OHT by applying the validated risk scoring system developed by Medeiros and colleagues.4,8 This predictive model has been recently updated and validated by the OHTS group and the EGPS group, and details of the scoring method have been previously published.4 For each patient with a documented diagnosis of OHT, we assigned points ranging from 0 (denoting lowest risk) to 4 (denoting highest risk) to each of 5 predictors: baseline age, IOP, CCT, vertical CDR, and Humphrey visual field PSD. Ocular measurements represented the average of the mean values for the right and left eyes. All points for the 5 RFs were summed, and the 5-year risk for conversion to glaucoma was determined based on the composite risk score (Table 2). Additionally, we calculated the average composite risk score and the average 5-year risk for conversion from OHT to glaucoma for the total OHT sample.
Risk Factor Prevalence and Patterns of Association. Medical records were available for 1189 patients who were evaluated between June 2000 and May 2005 for conditions classified under the ICD-9 global code index 365.x for glaucoma. One patient, aged 4 years, was deemed an outlier and excluded from the analyses. Tables 3 and 4 list the demographic, clinical, and ocular characteristics of the sample. The largest percentages of missing data were evident for the age (28.7%) and race (31.6%) parameters and measurements of CCT (81.0%).