Mortality in patients treated with intravitreal bevacizumab for age-related macular degeneration

This retrospective, population-based analysis accessed data from the electronic medical records of all individuals affiliated with Clalit Health Services who received intravitreal injections of bevacizumab for treatment of AMD between September 2008 and October 2014.

Clalit Health Services maintains a chronic disease registry database that includes information collected from a variety of sources: primary care physician reports, medication-use files, hospitalization records, and out-patient clinic records. The methods of registry acquisition and maintenance were described by Rennert and Peterburg [16].

For all individuals included in the analysis, we extracted information from the registry regarding the following conditions, which have been reported to be more prevalent in AMD patients and to be associated with increased mortality [17‐19]: smoking, alcohol abuse, ischemic heart disease, cerebrovascular disease, congestive heart failure, liver cancer, obesity, and (unilateral/bilateral) pseudophakia.

The definitions in the Clalit database of alcohol abuse are based on the Diagnostic and Statistical Manual of Mental Disorders, version IV. Cerebrovascular disease was diagnosed following the criteria of the National Institute of Neurological Disorders [20]. The clinical data standards of the American College of Cardiology/American Heart Association Task Force were used to define congestive heart failure and ischemic heart disease [21]. A body mass index of 30 kg/m² or higher defined obesity.

Additional information extracted from patients’ files included age, gender, marital status, and socioeconomic status.

The date of death was automatically communicated from the Israeli Interior Ministry via the unique national identity number. The cause of death was not recorded.

Ethics approval was obtained from the Ethics Committee of the Clalit Health Services.

In the nationwide Clalit Health Services records, we identified patients treated by anti-VEGF for wet AMD, and excluded those for whom there was doubt regarding the indication of the treatment. Forty-seven patients were excluded because it was not possible to eliminate diabetic macular oedema as the indication for injections; 29 since high myopia could not be ruled out as the cause of choroidal neovascularisation; 18 as the reason for treatment may have been a concomitant diagnosis of retinal vein occlusion; in 4 patients, inflammatory conditions were identified as the possible etiology of choroidal neovascularisation. Patients who received other intraocular anti-VEGF agents (pegaptanib, ranibizumab, aflibercept) or systemic anti-VEGF therapy at any time were excluded from the analysis.

For each wet AMD patient treated with bevacizumab in the study group, two individuals were matched in age and gender from the members of Clalit Health Services. A matched control had the same age as the case on the date of first bevacizumab injection. Criteria for this reference group were no recorded exposure to anti-VEGF and continual membership in Clalit Health Services from September 2008 until October 2014, excepting death.

For all ratio variables, means and standard deviations were calculated and baseline differences between the groups evaluated using a t-test. For all nominal variables, absolute frequencies and percentages were calculated and baseline differences between the groups were assessed using a Chi-square test. The socioeconomic ordinal variable baseline differences between the groups were evaluated using the Mann-Whitney test. To compare mortality over time between the groups, survival analysis was performed using adjusted Cox regression. The dependent variable was survival. The time-dependent covariate for the treatment group (bevacizumab) was the interval between the first injection to survival or death; and for the control group, the interval between the start of monitoring (date of first injection in the corresponding bevacizumab treated patient) to survival or death, all truncated at 7 years. Adjusted variables were age, smoking, alcohol abuse, hypertension, diabetes, obesity, congested heart failure, liver cancer, ischemic heart disease, and cerebrovascular accident.

Statistical analyses were conducted using the SPSS statistical software (Version 20). The criterion for accepting the research hypothesis was: Alpha (α) = .05 (one-sided). The criterion for negating the preliminary differences between the treatment and the control group was: Alpha (α) = .05 (two-sided).