Ischemic Optic Neuropathy in Cardiac Surgery: Incidence and Risk Factors in the United States from the National Inpatient Sample 1998 to 2013

PERIOPERATIVE visual loss is a rare but devastating complication of nonocular surgery. Ischemic optic neuropathy (ION) is the most common mechanism, with reported cardiac surgery incidence ranging from as high as 1.3 to 0.06%, and most recently estimated at 0.086%.^1–3  There is painless vision loss affecting either the anterior optic nerve or the posterior optic nerve. Anterior optic nerve injury is more prevalent in cardiac surgical procedures, whereas posterior optic nerve injury is more common in posterior spinal fusion surgery.^2,4  ION has been well studied in spinal fusion, with risk factors of age, obesity, male sex, blood loss, increased surgical duration, type of surgical positioning frame, and low colloid/crystalloid ratio for fluid therapy.^5–7  A recent study suggests that ION in spinal fusion has decreased during the past 15 yr.⁸ 

ION in cardiac surgery has been less studied and may result from a different mechanism. A 1982 case series suggested hypotension, hypothermia, and complement activation as possible risk factors.⁹  Three retrospective single-institution cohort studies did not replicate these findings but rather identified other risk factors including severe peripheral vascular disease, cardiopulmonary bypass (CPB) time, packed erythrocyte units transfused, lowest hemoglobin, nonpacked erythrocyte transfusion, and preoperative angiogram less than 48 h before the procedure.^1,2,4  However, these studies are limited by small size, with a combined total of only 34 cases and thus may lack generalizability. Moreover, they are now greater than 15 yr. Since then cardiac surgery has changed considerably, with 25% fewer coronary artery bypass graft (CABG) procedures, and a significant increase in left ventricular assist device insertion.^10,11  Using the National Inpatient Sample (NIS), Shen et al.³  performed the largest analysis to date of ION in cardiac surgery and identified male sex and an increased Charlson index score as risk factors. However, the study did not address factors specific to cardiac surgery such as comorbidities indicative of severe vascular disease, type of cardiac surgery, and use of CPB.

Chronic ophthalmic disease is a possible risk factor due to local hemodynamic impairment known to be associated with glaucoma, age-related macular degeneration, and diabetic retinopathy.^12,13  A recent study supported this notion by demonstrating increased prevalence of diabetic retinopathy in diabetics with incident microvascular ocular motor palsies.¹⁴  On a broader level, a large body of literature supports associations between vascular diseases of the eye with both acute and chronic cerebrovascular diseases that persist after accounting for potentially common causes such as systemic vascular disease.¹⁵ 

Based upon previous studies, we hypothesized that specific patient medical conditions are associated with ION after cardiac surgery and that these are distinct from ION after posterior spine fusion surgery. Moreover, we also hypothesized that specific surgical elements such as CABG and use of CPB would also be associated with ION. To test our hypotheses, we analyzed the NIS to identify risk factors associated with ION after cardiac surgery. Additionally, because of changes in procedure volume in cardiac surgery during the study period, we evaluated the incidence and trends in ION after cardiac surgery involving CABG, heart valve repair or replacement surgery, and left ventricular assist device procedure.

The NIS of the Healthcare Cost and Utilization Project is directed by the Agency for Healthcare Research and Quality (AHRQ, Rockville, Maryland).¹⁶  It is an approximate 20% stratified survey sample of nonfederal U.S. hospital discharges derived from a typical hospital discharge abstract. The NIS includes age, race, total charges, hospital characteristics including teaching status and location, discharge disposition, and 25 diagnostic and 15 procedural codes defined in the International Classification of Diseases, Ninth Revision–Clinical Modification (ICD-9-CM).¹⁷  Beginning in 2012, the NIS was redesigned to improve national estimates, by sampling all participating hospitals rather than a subset.¹⁶  To ensure accurate weighting of the sample across multiple years, AHRQ has provided updated discharge weights for the years 1998 to 2011. We used these updated trend weights, combined with the survey function for all patient-level analysis and regressions, as previously described.⁸  The University of Chicago and University of Illinois (Chicago, Illinois) Institutional Review Boards deemed the study exempt from Institutional Review Board review since there are no patient identifiers.

Our retrospective analysis included inpatient discharges of adults aged 18 yr or older with ICD-9-CM procedure codes for CABG (36.10 to 36.16, 36.17, 36.19), valve repair or replacement (35.1 to 35.14, 35.2 to 35.28), and left ventricular assist device (37.66) procedures from 1998 to 2013. Insertion of an intraaortic counter pulsation balloon pump (37.61) and CPB (39.61, 39.62) were included as secondary procedures. There were no cases of ION in heart transplant, atrial and ventricular septal defect repair, and congenital heart surgery in those aged 18 yr or older; therefore, these procedures were excluded.

To compare the incidence of ION in a nonsurgical inpatient population, we analyzed patients with significant coronary artery disease requiring a percutaneous intervention (PCI). Coronary artery disease was chosen since the majority of ION cases were associated with the CABG procedure. This group consisted of adults aged 18 yr or older with an ICD-9-CM procedure code for percutaneous coronary intervention and placement of either a bare-metal stent (36.06) or drug-eluting stent (36.07). Patients with a concurrent procedure code for CABG, valve repair or replacement, or left ventricular assist device procedure on the same admission were excluded from the PCI cohort. Surgical and PCI patients discharged with a principal or secondary diagnostic ICD-9-CM code of ischemic optic neuropathy (377.41) were considered to have developed ION during the hospitalization.

Patient characteristics analyzed included age (years, continuous variable), sex, length of hospital stay (days), yearly inflation-adjusted total hospital charges, type of admission (elective and emergent), discharge status (routine, short-term hospital, home health care, died, and other), and race.¹⁸  Potential risk factors were identified before analysis based upon previous case series, large database reviews, and case reports as recommended in the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.¹⁹  See Supplemental Digital Content 1, Table, https://links.lww.com/ALN/B374 for ICD-9-CM patient comorbidity codes. We studied obesity, obstructive sleep apnea, smoking, hypertension, diabetes type 1 or 2 uncomplicated, diabetes with renal manifestations, diabetes with neurologic manifestations, diabetic retinopathy, diabetes with peripheral circulatory disorders, hypertensive retinopathy, glaucoma, age-related macular degeneration, cataract, peripheral vascular disease, coronary artery disease, carotid artery stenosis, stroke (postoperative, acute, embolic, and thrombotic), blood transfusion, hyperlipidemia, atrial fibrillation, congestive heart failure (systolic and diastolic), pulmonary hypertension, and anemia. Degenerative eye conditions were included because they may predispose to ION.^20,21  To investigate concern that chronic ophthalmic conditions could appear falsely associated with ION due to increased frequency of eye exams in affected patients, cataract was also studied as a patient comorbidity as it should not act as a risk factor for ION.

The primary outcome was ION during an admission for one of the three cardiac surgical procedures, CABG, valve repair or replacement, or left ventricular assist device or for PCI. AHRQ trend weights and the survey functions of STATA (StataCorp LLC, USA) were used for all national estimates and regressions.⁸  Dividing individual survey estimates from the total count can lead to rounding variation in the NIS data set, as seen in the total number of ION cases when compared to cases of ION for the different cardiac procedures. Data were complete except for race (n = 1,261,433 [23%]), type of admission (n = 1,692,552 [30%]), age (n = 433 [0.01%]), length of stay (n = 185 [less than 0.01%]), and discharge status (n = 7,316 [0.1%]). Due to high rates of missing data for race and type of admission, they were not included in the univariable and multivariable analyses.

Patient characteristics for all cardiac procedures were compared for 1998 to 2013 using the national estimates and reported with 95% CI. We used chi-square with a second-order Rao–Scott correction to detect differences in the incidence of ION in the different cardiac surgery procedures and differences in cardiac surgery procedure rates during the study period. The mean incidence of ION in cardiac surgery and PCI was compared using an adjusted Wald test.²²  Rates of cardiac surgical procedures were calculated as procedures per million adults (greater than 18 yr old) in the U.S. population using U.S. Census data.²³  A univariate logistic regression was performed for patient characteristics and for primary and secondary cardiac procedures to identify covariates associated with ION. Characteristics and procedures with P ≤ 0.2 were then included in a multivariable logistic regression. Those with P ≤ 0.05 in the multivariable logistic regression were considered significant, and the odds ratio (OR), 95% CIs, and P value were reported. Changes in the temporal trend of ION incidence were assessed using multivariable logistic regression, modeling year as a continuous variable and including primary cardiac procedures to account for changes in procedure volume over time.

The variance inflation factor (VIF) examined collinearity. VIF greater than 5 identified possible collinearity between predictors. None of the identified predictors had VIF greater than 5, which suggests no collinearity of predictors. Pearson goodness-of-fit was used to assess the multivariable model fit and was not significant at the 5% level (P = 0.127), thus the multivariable model could not be rejected. STATA v14.0-MP (StataCorp LLC) was used for all data analysis.

There were an estimated 5,559,395 discharges between 1998 and 2013, with procedure codes for a CABG, valve repair or replacement, or left ventricular assist device (fig. 1). A diagnosis of ION was documented in 794 of those procedures, corresponding to an average yearly incidence of 1.43 per 10,000 cardiac surgeries (95% CI, 1.23 to 1.67; table 1). After adjusting for changes in procedural volume, there was no significant change in the yearly incidence of ION in cardiac surgery (OR, 1.01; P = 0.574; fig. 2). There were an estimated 9,520,478 discharges with procedure codes for PCI, and 206 had a diagnosis for ION (see Supplemental Digital Content 2, Figure, https://links.lww.com/ALN/B375, which illustrates the incidence of ION in PCI). The incidence of ION in PCI was 0.22 per 10,000 (95% CI, 0.16 to 0.30), significantly less than that for patients undergoing isolated CABG (P = 0.0001).

Isolated CABG surgery accounted for the highest volume of diagnoses of ION (621 of 794; 78%), followed by a combined CABG and valve repair or replacement (86 of 794; 11%), valve repair or replacement alone (78 of 794; 10%), and left ventricular assist device (10 of 794; 1.3%). The incidence of ION for each procedure category is illustrated in figure 1. The incidence of ION was not different between the different cardiac surgery procedures (P = 0.0596).

Figure 3 shows changes in cardiac surgery procedures from 1998 to 2013. There was a 60% decrease in CABG as the sole procedure from 1998 to 2013 (1,725 to 686 per million adults), 1,092% increase in left ventricular assist device (1.3 to 14.2 per million), 53% increase in valve repair or replacement alone (206 to 315 per million), and a 22% decrease in combined CABG and valve repair or replacement (176 to 137 per million; P = 0.0001).

Characteristics of patients with a cardiac surgical procedure are presented in table 2, and characteristics of patients with a PCI are presented in the Table, Supplemental Digital Content 3, https://links.lww.com/ALN/B376.

A univariable analysis was performed for all cardiac surgery patient characteristics and for primary and secondary cardiac procedures (table 2). Significant characteristics and procedures from the univariable analysis (P < 0.20) were combined to create a multivariable logistic regression (table 3). Patient characteristics, from the multivariable model, associated with increased odds of ION include carotid artery stenosis (OR, 2.70; 95% CI, 1.52 to 4.80; P = 0.001), stroke (OR, 3.43; 95% CI, 1.73 to 6.80; P = 0.0004), diabetic retinopathy (OR, 3.83; 95% CI, 1.84 to 7.95; P = 0.0003), hypertensive retinopathy (OR, 30.09; 95% CI, 6.21 to 145.64; P = 0.0001), glaucoma (OR, 2.68; 95% CI, 1.04 to 6.93; P = 0.042), age-related macular degeneration (OR, 4.50; 95% CI, 1.13 to 17.87; P = 0.032), and cataract (OR, 5.62; 95% CI, 1.71 to 18.45; P = 0.004; fig. 4A). Female sex (OR, 0.59; 95% CI, 0.38 to 0.92; P = 0.019) and uncomplicated diabetes mellitus type 2 (OR, 0.51; 95% CI, 0.32 to 0.83; P = 0.006) were associated with lower ORs (fig. 4B). The primary procedures associated with increased odds of ION were left ventricular assist device (OR, 13.82; 95% CI, 1.75 to 109.01; P = 0.013) and two-vessel CABG (OR, 1.78; 95% CI, 1.04 to 3.06; P = 0.035; fig. 4A). CPB was not associated with increased odds of ION (OR, 0.78; P = 0.219).

The incidence of ION in CABG, valve repair or replacement, and left ventricular assist device, from 1998 to 2013, was 1.43 per 10,000 with no significant trend over the study period. The estimated incidence of ION after PCI, in the same database, was 0.22 per 10,000, near the lower end of previous national estimates of ION incidence in the general adult population at 2 to 10 per 100,000.^24,25  The seven-fold higher incidence of ION in cardiac surgery as compared to PCI and previous national estimates supports previous studies identifying cardiac surgery as high risk for developing ION and suggests that our findings are due to the development of ION during the perioperative period, and not due to a preexistent diagnosis. Overall, no differences in the incidence of ION were identified between the different cardiac surgeries; however, in the multivariable analysis, left ventricular assist device and two-vessel CABG were associated with increased odds of ION. Two-vessel CABG accounted for the highest percentage of isolated CABG procedures and diagnoses of ION; although this was accounted for in the multivariable analysis. It should be noted that left ventricular assist device procedural volume was smaller relative to the other cardiac surgeries, which led to wide 95% CIs for the estimate, thus the result should be interpreted cautiously.

Carotid artery stenosis and stroke were associated with increased risk of ION in our study. Carotid artery stenosis has been listed in numerous case reports as a diagnosis in affected patients, and atherosclerosis was previously identified in a single-center retrospective cohort as a risk factor.^2,9,26,27  However, neither stroke nor carotid stenosis has been identified as a risk factor for spontaneously (nonperioperative) occurring ION.^28,29  Taken together, these findings suggest that in cardiac surgery, hypotension, systemic inflammation, or other as yet unknown perioperative mechanisms may interact with already decreased perfusion via the carotid circulation to heighten the risk of ION.

Our study identified chronic eye conditions associated with altered posterior eye circulation such as glaucoma, age-related macular degeneration, diabetic retinopathy, and hypertensive retinopathy as risk factors in perioperative ION. Glaucoma and age-related macular degeneration may predispose the optic nerve to ischemia due to impaired autoregulation of blood flow.^30–32  Similarly, diabetic retinopathy and hypertensive retinopathy are characterized with endothelial damage, a leaky blood–retinal barrier, vascular occlusion, and ischemia, leading to neovascularization.³³  Development of ION may be associated with these conditions due to impaired microcirculation diffusely within the eye including the optic nerve. Thus, these degenerative eye conditions could be a marker of more widespread ocular circulatory abnormalities.³⁴  A 2014 analysis of posterior spine fusion surgery in the NIS found an increased risk of visual loss in patients with diabetes with end-organ damage, but the association with ION is unclear as the study included discharges with cortical blindness, and retinal artery occlusion in addition to ION.³⁵ 

These chronic degenerative eye conditions may be markers of increased risk of ION after cardiac surgery, but the results should be interpreted cautiously. Patients who developed perioperative ION likely underwent a detailed ophthalmologic exam to confirm the diagnosis leading to a heightened diagnosis intensity of degenerative eye conditions.³⁶  Increased diagnosis intensity may have led to the association seen between ION and cataract, as there is no theoretical reason for its association. In contrast, glaucoma requires multiple exams and more sophisticated technology for a diagnosis, and an initial diagnosis was unlikely to occur during the hospital admission for cardiac surgery.³⁷ 

Female sex and uncomplicated diabetes were associated with a decrease in the odds of ION after cardiac surgery. The female visual pathway is not known to anatomically differ versus the male visual pathway; however, hormonal factors such as estrogen may play a role. Estrogen improves vascular function and decreases atherosclerosis; however, female sex had decreased odds of ION in our study even when controlling for vascular disease and carotid artery stenosis in the multivariable regression leaving the mechanism for this decrease unclear.^38,39  The reason why ION risk in this study differed between uncomplicated diabetes mellitus type 2 and diabetic retinopathy is not known and will require further study.

ION was not associated with factors specific to cardiac surgery that may decrease optic nerve blood flow or embolic events, including CPB, cardiogenic shock, congestive heart failure, atrial fibrillation, pulmonary hypertension, and anemia.^1,2,4  Several case series and case reports of ION after cardiac surgery identified anemia, blood loss, and lowest hemoglobin level as possible risk factors for ION; however, we did not find anemia or transfusion as significant risk factors. Kalyani et al.⁴  found a trend toward a perioperative change in hemoglobin levels and risk of ION but not a significant association. Nuttall et al.²  and Shapira et al.¹  identified low postoperative hemoglobin concentrations as an independent risk factor for ION in cardiac surgery after CPB. In the NIS, these data may not accurately reflect perioperative hemoglobin values.

Our study has limitations secondary to using an administrative database. The NIS is a stratified probability sample of inpatient discharges and hospitals in the sampling frame of all nonfederal hospitals in the United States. As such, because ION is a rare event, oversampling or overweighting of discharges containing a diagnosis of ION may have increased the incidence of ION. However, the NIS is a robust 20% sample of the entire NIS universe and has been rigorously validated.⁴⁰  Discharge records are susceptible to ICD-9-CM coding errors for a diagnosis of ION. The incidence of ION is low, and small errors in diagnostic coding could have a large impact on our findings. However, our unweighted sample size is greater than 1 million admissions, which mitigates systematic reporting bias, and previous studies have used the NIS to identify the incidence of ION in various surgical populations.⁸ 

Our study is limited to identifying patient conditions and surgical procedures associated with increased odds of ION after cardiac surgery. As such, patient conditions reported in the NIS using ICD-9-CM codes may not accurately reflect the clinical spectrum of severity, such as in cardiogenic shock or anemia. Thus, while our study did not find a diagnosis of anemia to be a significant risk factor for developing ION, it is possible that the lack of granularity may have impacted our results. Furthermore, the NIS does not contain any intraoperative information about the anesthetic or intraoperative care such as duration of CPB. Additionally, ICD-9-CM codes do not exist for certain aspects of the procedure such as redo sternotomy, which may impact the development of ION.

The temporal relationship between a diagnosis of ION and the timing of the cardiac surgery cannot be determined in the NIS, thus the cardiac surgery population may have had a higher baseline prevalence of ION. However, the seven-fold higher incidence of ION after CABG as compared to PCI and previous national estimates strongly suggests that the majority of ION we observed occurred after cardiac surgery. The severity of visual loss cannot be determined from the database as well as longitudinal follow-up for progression or improvement after discharge. The NIS represents a single inpatient admission and does not contain any patient identifiers to allow patients to be identified for follow-up.

In conclusion, cardiac surgery has an incidence of perioperative ION with an average yearly incidence of 1.43 per 10,000 procedures; the majority of the cases were associated with isolated CABG. This incidence was about seven times that of a comparable inpatient group undergoing percutaneous cardiac intervention, as well as the known incidence of spontaneously occurring ION in the general population. The yearly incidence has not changed despite decreases in CABG, increases in left ventricular assist device, and an overall decrease in cardiac surgery volume between 1998 and 2013. ION is most frequently associated with carotid artery stenosis, stroke, male sex, degenerative eye conditions, two-vessel CABG, and left ventricular assist device. Uncomplicated diabetes type 2 and female sex were associated with a lower risk of ION. Further research is needed to identify potential therapeutic interventions to decrease the risk of this rare and devastating complication.

The authors are grateful to Ms. Chuanhong Liao, M.S., Senior Biostatistician, Department of Public Health Sciences, University of Chicago, Chicago, Illinois, for assistance with the statistical analysis.

Supported by the following grants from the National Institutes of Health (Bethesda, Maryland): RO1 EY10343, to Dr. Roth; UL1 RR024999, to the University of Chicago Institute for Translational Medicine (Chicago, Illinois); UL1 TR002003, to the University of Illinois at Chicago Center for Clinical and Translational Science (Chicago, Illinois); and K23 EY024345, to Dr. Moss. Also supported by core grant No. P30 EY001792 from the Department of Ophthalmology and Visual Sciences of the University of Illinois, the University of Chicago Pritzker School of Medicine Summer Research Program (Chicago, Illinois; to Ms. Matsumoto), and an unrestricted grant from Research to Prevent Blindness (New York, New York; to Dr. Moss, Dr. Joslin, and Dr. Roth) and from the Department of Ophthalmology and Visual Sciences, University of Illinois at Chicago (Chicago, Illinois; to Dr. Moss, Dr. Joslin, and Dr. Roth).

Dr. Roth has served as an expert witness in cases of perioperative eye injuries on behalf of patients, physicians, and hospitals. The other authors declare no competing interests.