Smoking is associated with a higher risk of surgical intervention for thyroid eye disease in the IRIS^® Registry

Abstract

Purpose:

To describe the association of smoking status with surgical intervention for TED at the population-level.

Design:

Retrospective cohort study.

Methods:

Setting:

US clinical data registry

Study population:

All adults (age ≥18 years) with Graves disease in the Intelligent Research in Sight (IRIS^®) Registry (1/1/2013–12/31/2020).

Observation:

Surgical intervention for TED, stratified into orbital decompression, strabismus surgery, and eyelid recession surgery.

Outcome measures:

The Kaplan-Meier estimated five-year cumulative probability for each surgical intervention. Multivariable Cox regression was used to evaluate the association between smoking status and each surgical intervention, adjusting for age, sex, race, ethnicity, and geographic region.

Results:

This study included 87,744 patients. Median age was 59 years (IQR, 48–68 years); 81% were female patients. Current smokers had a greater five-year cumulative probability of orbital decompression (3.7% vs.1.9%; p<0.001), strabismus surgery (4.6% vs.2.2%; p<0.001) and eyelid recession (4.1% vs.2.6%; p<0.001) compared to never smokers. After adjusting for demographic factors, current smokers were at greater risk of orbital decompression (HR, 2.1; 95% CI 1.8–2.4; p<0.001), strabismus surgery (HR, 2.0; 95% CI 1.8–2.3; p<0.001), and eyelid recession (HR, 1.7; 95% CI 1.5–1.9; p<0.001) than never smokers. Former smokers were at higher risk of each type of surgery for TED, though at lower levels than current smokers.

Conclusions:

Smoking was associated with increased risk of surgical intervention for TED in the IRIS Registry. Former smokers were at a lower risk than current smokers, supporting the role of smoking cessation on lowering the burden of surgical disease at the population-level.

Table of Contents Statement:

In this study of over 87,000 patients in the Intelligent Research in Sight Registry, smoking was associated with an increased risk of surgical intervention for thyroid eye disease (orbital decompression, strabismus surgery, and eyelid recession). Former smokers were at a lower risk than current smokers, supporting the role of smoking cessation on lowering the burden of surgical disease at the population-level.

Graves disease is the most common form of autoimmune hyperthyroidism, affecting approximately 0.5% of the population.^1,2 Circulating autoantibodies may not only cause thyroid hormone dysregulation, but also stimulate orbital tissue to cause extraocular muscle enlargement and orbital fat expansion. Thyroid eye disease (TED) in the setting of Graves disease may present with exophthalmos, ocular misalignment, and optic neuropathy, all of which may cause ocular morbidity and negatively impact the quality of life.^3,4 Medical therapy for Graves disease generally aims to correct the underlying hyperthyroid state, which may have little impact on TED. A variety of medical and surgical interventions are used to treat TED, including orbital decompression, strabismus surgery, and eyelid recession surgery.

Several risk factors for TED have been identified, with smoking described as the most important modifiable risk factor.⁵ Studies have shown a dose-dependent association between tobacco use and disease progression.⁶ Smoking may exacerbate the underlying autoimmune process in Graves disease, and it has been shown to impact gene expression involved in several disease-related pathways, which appears reversible with cessation.^7,8 Although the relationship between smoking and TED disease progression has been well described, the strength of the association between tobacco use and clinical outcomes in large, diverse cohorts is less well described.

The IRIS^® Registry (Intelligent Research in Sight) includes the electronic health records (EHR) of over 73 million patients cared for in ophthalmology practices in the United States and contains information on smoking history, with several previous studies using these data to study associations with ocular disease.^9,10 This study leverages the IRIS Registry to quantify the impact of current and former tobacco smoking on the risk of surgical interventions for TED.

Methods:

This retrospective cohort study was performed using the EHR data of patients followed at practices participating in the IRIS Registry. The version of the database used in this study was frozen on December 20, 2021 and accessed on June 21, 2022. The data collection methodology of the IRIS Registry has been described previously.¹¹ Briefly, the American Academy of Ophthalmology started the IRIS Registry as an initiative to improve quality and patient outcomes in eye care. Clinical data are extracted directly from the EHRs of participating practices, and deidentified records are stored in a secure, cloud-based data warehouse. Over fifteen thousand clinicians have registered to contribute their data to the IRIS Registry, which includes over 440 million visits from over 73 million unique patients as of April 1, 2022. This investigation was approved by the Massachusetts General Brigham Institutional Review Board with the exemption of informed consent. The research adhered to the Declaration of Helsinki and the United States’ Health Insurance Portability and Accountability Act.

The records of all adults in the IRIS Registry with a clinical encounter in an ophthalmology practice for Graves disease at age ≥ 18 years old between January 1, 2013 and December 31, 2020 were included in this study. Subjects were identified using the International Classification of Diseases (ICD) codes for Graves disease (ICD-9, 242.0; ICD-10, E05.0). Surgical interventions for TED were identified using Current Procedural Terminology (CPT) codes and categorized into orbital decompression surgery, strabismus surgery, and eyelid recession surgery (Table 1). Orbital decompression surgery was categorized into endoscopic vs. non-endoscopic approach. Strabismus surgery was categorized into horizontal, vertical, or combined horizontal and vertical muscle surgery. Eyelid recession surgery included upper or lower eyelids. Patients with a history of surgical intervention for TED prior to the first encounter for Graves disease.

Table 1:

The diagnosis codes used to identify adults with Graves disease and procedure codes used to identify surgical interventions for thyroid eye disease.

Category	Type	Code	Description
Graves disease	ICD-9	242.0*	Toxic diffuse goiter
	ICD-10	E05.0*	Thyrotoxicosis with diffuse goiter
Orbital decompression	CPT	67414	Orbitotomy without bone flap (frontal or transconjunctival approach)
		67445	Orbitotomy with bone flap or window, lateral approach
		31292	Nasal/sinus endoscopy, surgical, with orbital decompression; medial or inferior wall
		31293	Nasal/sinus endoscopy, surgical, with orbital decompression; medial and inferior wall
		31294	Nasal/sinus endoscopy, surgical; with optic nerve decompression
Strabismus surgery	CPT	67311, 67312	Horizontal muscle surgery
		67314, 67316	Vertical muscle surgery
		67318	Superior oblique muscle surgery
		67320	Extraocular muscle transposition
		67334	Posterior fixation suture
Eyelid recession	CPT	67911	Eyelid retraction repair

ICD=International Classification of Diseases; CPT=Current Procedural Terminology;

^*=includes all sub-codes

The main risk factor investigated in this study was smoking status. Patients were categorized based on self-reported tobacco use into current, former, or never smokers. In situations where there was a discrepancy in self-reported smoking status across patient visits, a heuristic algorithm was used to classify smoking status. First, any patient ever recorded as a current smoker was assigned to the “current smoker” group. Next, any patient ever recorded as a former smoker was assigned to the “former smoker” group. Finally, any remaining patient ever recorded as a never smoker was assigned to the “never smoker” group. Patients without a documented smoking status were excluded from this analysis (Figure 1A).

Figure 1:

Flow charts describing A) the heuristic model used to assign smoking status and B) the study inclusion and exclusion criteria.

The time between first clinical encounter with an eye specialist for Graves disease and surgical intervention for TED (orbital decompression, strabismus, or eyelid recession surgery) was collected for each patient. The primary outcome was the cumulative probability of surgical intervention within five years of first encounter for Graves disease, obtained using the Kaplan-Meier estimator of survival. Multivariable Cox Proportional Hazards (PH) regression models were used to quantify the association between smoking status and surgical intervention. The covariates included in the model were age, sex, race and ethnicity and geographic location. Patient age at first encounter was stratified as 18 to 40 years old, 40 to <60 years old, 60 to <80 years old and > 80 years old. Race and ethnicity were also treated as a categorical covariate with four groups: White non-Hispanic, Black non-Hispanic, Hispanic, and other. Geographic location was classified using the four US census regions (Northeast, Midwest, South, and West) corresponding to the state of residence for each individual.

Medians and interquartile ranges (IQR) were reported for continuous variables and frequencies for categorical variables. The Kaplan Meier estimator of survival was used to report the five-year cumulative probability of surgical intervention for TED. Log-rank tests were used to compare the survival curves stratified by smoking status. Chi-square and Wilcoxon rank-sum tests were used to compare the characteristics of individuals with and without surgical intervention for TED in the unadjusted complete case analysis. Multivariate Imputation by Chained Equations (MICE)¹² was used to impute any covariates with >5% missing data in the regression analysis (race and ethnicity, geography). Hazard Ratios (HR) with 95% confidence intervals (CI) are reported from the multivariable Cox PH regression models. The potential relationship between smoking status and patient sex was explored using interaction terms in the regression models. All statistical tests were two-sided, with significance defined as p < 0.05. The Bonferroni-Holm method was used to account for multiple comparisons in pairwise testing, where adjusted p-values (p_adj) are reported. All analyses were performed using R, version 4.2.0 (R Core Team, 2021) with the MICE package, version 3.14.0.¹³

Results:

A total of 94,154 adults with Graves disease were identified in the IRIS Registry during the study interval, and 87,774 were included in this analysis (Figure 1B). The median age at the first clinical encounter was 59 years (IQR, 48–68 years), and most patients (81%) were women (Figure 2). The cohort was subdivided into current smokers (19%), former smokers (24%), and never smokers (57%) (Table 2).

Figure 2:

Histograms demonstrating the distribution of age at first encounter with eye specialist in the IRIS Registry with an ICD-9/10 code for Graves disease stratified by sex.

Table 2:

Baseline patient demographics of patients with Graves disease in the IRIS Registry stratified by smoking status.

		Smoking status
	Overall, N = 87,7441	Current, N = 17,070	Former, N = 20,946	Never, N = 49,728	p-value2
Age group, n (%)					<0.001
18 to <40	12,324 (14)	2,152 (13)	1,634 (7.8)	8,538 (17)
40 to <60	33,248 (38)	7,612 (45)	6,372 (30)	19,264 (39)
60 to <80	37,064 (42)	6,718 (39)	11,223 (54)	19,123 (38)
≥80	5,108 (5.8)	588 (3.4)	1,717 (8.2)	2,803 (5.6)
Female sex, n (%)	70,835 (81)	13,210 (78)	16,075 (77)	41,550 (84)	<0.001
Unknown	324	74	74	176
Race and ethnicity, n (%)					<0.001
White Non-Hispanic	56,237 (77)	11,610 (81)	15,002 (83)	29,625 (73)
Black Non-Hispanic	8,984 (12)	1,731 (12)	1,749 (9.7)	5,504 (14)
Hispanic	4,820 (6.6)	666 (4.7)	913 (5.1)	3,241 (8.0)
Other	2,906 (4.0)	297 (2.1)	394 (2.2)	2,215 (5.5)
Unknown	14,797	2,766	2,888	9,143
US census region, n (%)					<0.001
Midwest	19,041 (25)	4,656 (31)	4,718 (26)	9,667 (22)
South	28,506 (37)	5,198 (35)	6,452 (36)	16,856 (39)
West	13,740 (18)	2,425 (16)	2,999 (17)	8,316 (19)
Northeast	14,981 (20)	2,759 (18)	3,773 (21)	8,449 (20)
Unknown	11,476	2,032	3,004	6,440

¹n (%)

²Pearson’s Chi-squared test

Orbital decompression procedures (either endoscopic or transorbital) were performed on 1212 patients with Graves disease in this cohort. The five-year cumulative probability of orbital decompression was 2.3% (95% CI, 2.2–2.5%). Orbital decompression was more likely to be performed in current smokers compared to never smokers (3.7% vs. 1.9%; p_adj < 0.001). Former smokers had a similar five-year probability of orbital decompression compared to non-smokers (2.1% vs. 1.9%; p_adj = 0.14) (Figure 3a).

Figure 3:

Kaplan-Meier curves comparing the cumulative probability of orbital decompression, strabismus surgery, and eyelid recession in patients with Graves disease stratified by smoking status.

Strabismus surgery was performed on 1392 of the patients included in this study. The type of initial strabismus surgery included horizontal muscles (29%, n = 398), vertical muscles (43%, n = 582), and combined horizontal and vertical muscles (28%, n= 390). The five-year cumulative probability of strabismus surgery for the entire cohort was 2.9% (95% CI, 2.7–3.1%). Strabismus surgery was more likely to be performed in current smokers (4.6% vs. 2.2%; p_adj < 0.001) and former smokers (3.1% vs. 2.2%; p_adj < 0.001) compared to never smokers (Figure 3b).

Eyelid recession surgery was performed on 1324 patients in this study. The five-year cumulative probability of eyelid recession surgery for the entire cohort was 2.9% (95% CI, 2.7–3.1%). Eyelid recession surgery was more likely to be performed in current smokers (4.1% vs. 2.6%; p_adj < 0.001) compared to never smokers. Former smokers and never smokers had a similar probability of undergoing eyelid recession (2.6% vs. 2.6%; p_adj = 0.45) (Figure 3c).

After adjusting for age, sex, race and ethnicity, and geographic region, current smokers were at greater risk for undergoing orbital decompression surgery (HR, 2.1; 95% CI 1.8–2.4; p < 0.001), strabismus surgery (HR, 2.0; 95% CI 1.8–2.3; p < 0.001), and eyelid recession (HR, 1.7; 95% CI 1.5–1.9; p < 0.001) than never smokers. Former smokers were also at higher risk for undergoing orbital decompression (HR, 1.3; 95% CI 1.1–1.5; p = 0.001), strabismus surgery (HR, 1.3; 95% CI 1.1–1.5; p < 0.001), and eyelid recession (HR, 1.2; 95% CI 1.0–1.4; p = 0.016) than current smokers (Table 3).

Table 3:

Hazard Ratios (HR) from the multivariable Cox regression models for the association between smoking status and surgical intervention for TED adjusting for all covariates including age, sex, race, ethnicity, and geographic location.

	Orbital decompression		Strabismus surgery		Eyelid recession
	HR (95% CI)1	p-value	HR (95% CI)1	p-value	HR (95% CI)1	p-value
Smoking status
Never	1.00		1.00		1.00
Former	1.27 (1.09 to 1.47)	0.002	1.28 (1.12 to 1.46)	<0.001	1.19 (1.03 to 1.36)	0.016
Current	2.08 (1.82 to 2.38)	<0.001	1.99 (1.75 to 2.26)	<0.001	1.67 (1.46 to 1.90)	<0.001
Age group
18 to <40	1.00		1.00		1.00
40 to <60	0.94 (0.79 to 1.11)	0.46	2.13 (1.63 to 2.79)	<0.001	0.87 (0.73 to 1.02)	0.089
60 to <80	0.55 (0.46 to 0.66)	<0.001	2.46 (1.89 to 3.20)	<0.001	0.54 (0.46 to 0.65)	<0.001
≥80	0.32 (0.22 to 0.47)	<0.001	1.89 (1.35 to 2.66)	<0.001	0.30 (0.21 to 0.44)	<0.001
Female	0.65 (0.57 to 0.74)	<0.001	0.50 (0.45 to 0.56)	<0.001	1.09 (0.95 to 1.27)	0.23
Race and ethnicity
White Non-Hispanic	1.00		1.00		1.00
Black Non-Hispanic	1.06 (0.89 to 1.27)	0.52	0.64 (0.51 to 0.79)	<0.001	0.76 (0.62 to 0.92)	0.005
Hispanic	2.02 (1.67 to 2.45)	<0.001	1.24 (1.00 to 1.55)	0.051	1.40 (1.14 to 1.73)	0.001
Other	1.19 (0.88 to 1.61)	0.26	0.99 (0.72 to 1.36)	0.94	0.98 (0.71 to 1.36)	0.89
US census region
Midwest	1.00		1.00		1.00
South	1.67 (1.42 to 1.96)	<0.001	1.55 (1.34 to 1.80)	<0.001	1.80 (1.54 to 2.11)	<0.001
West	1.23 (1.00 to 1.51)	0.049	1.38 (1.15 to 1.64)	<0.001	1.08 (0.88 to 1.32)	0.48
Northeast	0.66 (0.52 to 0.83)	<0.001	0.91 (0.76 to 1.10)	0.34	1.12 (0.93 to 1.35)	0.24

¹HR = Hazard Ratio, CI = Confidence Interval

Compared to current smokers, former smokers had a lower five-year probability of orbital decompression (3.7 vs. 2.1; p_adj < 0.001), strabismus surgery (4.6 vs. 3.1; p_adj < 0.001), and eyelid recession (4.1 vs. 2.6; p_adj < 0.001). Similarly in the adjusted regression model, former smokers had a lower risk of undergoing orbital decompression surgery (HR, 0.6; 95% CI 0.5–0.7; p < 0.001), strabismus surgery (HR, 0.65; 95% CI 0.56–0.75; p < 0.001), and eyelid recession (HR, 0.77; 95% CI 0.67–0.89; p < 0.001) compared to current smokers.

Several additional factors associated with surgical interventions for TED were identified in this cohort (eSupplement). Compared to male patients, female patients were less likely to undergo orbital decompression (HR, 0.65; 95% CI 0.57–0.74; p < 0.001) and strabismus surgery (HR, 0.50; 95% CI 0.45–0.56; p < 0.001), but not eyelid recession (HR, 1.09; 95% CI 0.95–1.27; p = 0.23). The relationship between smoking status and surgical intervention was similar for male and female patients (all p > 0.18). The probability of undergoing surgical intervention varied with age (all p < 0.001) with orbital decompression and eyelid recession peaking in the 40 to <60 year-old group and strabismus surgery peaking in the 60 to <80 year-old group. Hispanic patients were more likely to undergo orbital decompression (12% vs. 8%; p_adj < 0.001) and Black patients were less likely to undergo strabismus surgery (7% vs. 12%; p_adj < 0.001) compared to non-Hispanic White patients. The regional differences in TED surgical interventions included a greater number of procedures performed in the South and West census regions and fewer in the Midwest and Northeast census regions.

Discussion:

Smoking was associated with a greater risk of surgical intervention for TED in this large cohort of patients with Graves disease in the IRIS Registry. The risk for undergoing orbital decompression, strabismus surgery, and eyelid recession surgery was elevated in smokers compared with non-smokers, and was slightly lower for former smokers compared to current smokers after adjusting for demographic and regional variations in the rates of surgical intervention. The findings in this large registry-based analysis support the existing literature on the association between smoking and TED^5,14–18 and provide novel insight on the potential impact of smoking cessation on lowering the risk of surgical intervention for TED at the population-level.

Few studies have described the incidence of surgical interventions associated with TED in large cohorts.^19,20 Among 120 incident cases of TED in Olmsted County, Minnesota, the four-year cumulative incidence of strabismus surgery was 5% and orbital decompression was 4%.²⁰ Among 4106 cases in a Danish nationwide study, the four-year cumulative incidence was 8% for strabismus surgery and 5% for orbital decompression.¹⁹ The lower cumulative probabilities in this study are most likely due to the inclusion of all patients with Graves disease followed in the IRIS Registry, which may include non-incident cases of TED as well as patients with Graves disease who are followed in an ophthalmology practice but have not developed TED. A retrospective cohort study of 501 patients with TED at a single center reported increased risk of strabismus surgery among active smokers compared to nonsmokers (age-adjusted HR of 2.2 increasing to 4.9 among individuals who did not require orbital decompression), which is comparable to the findings in our study.²¹

Smoking has been described as an important modifiable risk factor for TED, with the degree of risk proportional to the number of cigarettes smoked per day.^22,23 There is a higher prevalence of TED in patients with Graves disease who smoke.^5,17 Smokers tend to have more severe TED⁵ and are more likely to have worse outcomes after radioiodine^24,25 or immunosuppressive treatment.^18,26,27 Orbital fibroblasts play an important role in the pathogenesis of TED by expressing antigens that respond to the circulating antibodies. Fibroblast activation triggers connective tissue remodeling, contributing edema of the periorbital tissue, while smoking generates reactive oxygen species, which may enhance the effects of oxidative stress and intensify orbital fibroblast production;⁸ however, the exact mechanism is poorly understood. Smoking cessation may reverse some of these processes,^15,28 and our findings supplement the existing literature by showing that former smokers are at lower risk of surgical intervention for TED than current smokers.

The secondary findings in our study also support existing literature on risk factors for surgical interventions for TED. Women are at higher risk of Graves disease and indeed represent the majority of Graves disease patients included in the IRIS Registry.²⁹ The role of sex on TED is controversial, with older studies suggesting a slightly increased risk for men and recent studies suggesting no sex-related risk.^23,29 In the Danish cohort, Boulakh et al. found that men were twice as likely to undergo strabismus surgery, but equally likely to undergo orbital decompression compared to women.¹⁹ In this cohort, female sex was associated with lower risk of both orbital decompression and strabismus surgery. It has been reported that smoking increases the risk more in women than men.³⁰ There was no interaction between sex and smoking status in this cohort, suggesting that while female sex increases the risk of developing Graves disease and TED, smoking increases the risk of surgical intervention for TED similarly for men and women.

The prevalence of TED increases with age, peaks in the fifth and sixth decades of life, and may even have a bimodal distribution varying by sex.²⁰ In this cohort, fewer surgical interventions for TED were performed on adults before age 40 and after the age of 80. Patients with Graves disease were generally less likely to undergo orbital decompression and eyelid recession with increasing age. This finding may be a result of older patients choosing to avoid major surgery, and perhaps younger patients being more likely to undergo orbital decompression to improve cosmesis. In contrast, individuals with Graves disease were less likely to undergo strabismus surgery at a younger age, with the rate peaking in the 6^th and 7^th decades of life and decreasing thereafter. This pattern differs from the overall age-related decline in eye muscle surgery that has been reported previously using the IRIS Registry,³¹ and may suggest that the cumulative effects of extraocular muscle inflammatory changes in older patients with TED are more likely to require surgical intervention.

Despite the size and scope of the IRIS Registry, several limitations should be considered when interpreting these findings. First, smoking status used in this study was self-reported and may underestimate the true prevalence of smoking in this population. The EHR data in the IRIS Registry was not collected primarily for research purposes, and there exists potential for miscoded records of smoking status. There is also potential for selection bias as we excluded individuals with undocumented smoking status and patients may underreport their smoking habits. The heuristic algorithm used to assign smoking status attempts to minimize this potential source of error, but it may overestimate the number of current smokers by inadvertently including individuals who recently quit smoking – approximately 20% of patients had more than one code listed and required use of the algorithm for classification. Second, many patients with newly diagnosed Graves disease have no ocular involvement,³² and some who develop TED improve spontaneously.³³ Therefore, not all patients with Graves disease are referred for ophthalmic care, potentially skewing the cohort toward including individuals with more severe ocular disease. This should not influence the goal of the present study, which evaluated the risk for progression to surgery in those already followed in an ophthalmology practice. Unfortunately, granular data does not exist to describe each patient’s specific clinical course and thyroid status over time. This study may underestimate the overall probability of surgical intervention in cases where patients receive subsequent surgical care in practices not participating in the IRIS Registry. Academic medical centers are underrepresented in the IRIS Registry¹¹ and this study may not capture patients with complex pathology referred to these centers for surgical management. Finally, the landscape of surgical intervention for TED has likely changed with the approval of Teprotumumab in the US in 2020 and the impact of this medication could be investigated in the future using the IRIS Registry.

In conclusion, smoking among patients with TED was associated with a higher risk of undergoing orbital decompression, strabismus surgery, and eyelid recession. Former smokers had a lower risk of surgical intervention compared to current smokers, supporting existing literature on the role of smoking as an important modifiable risk factor for TED related morbidity using a large clinical registry of over 80,000 patients. Physicians should be aware of the long-term burden of surgical disease associated with smoking in TED and counsel active smokers with Graves disease on the benefits of smoking cessation.

Appendix:

Members of the IRIS^® Registry Analytic Center Consortium

Suzann Pershing, MD;¹ Leslie Hyman, PhD;² Julia A. Haller, MD;² Aaron Y. Lee, MD MSCI;^3,4

Cecilia S. Lee, MD MS;⁴ Flora Lum, MD;⁵ Joan W. Miller, MD;⁶ Alice C. Lorch, MD, MPH⁶

1. Stanford University, Palo Alto, CA, USA

2. Wills Eye Hospital, Philadelphia, PA, USA

3. eScience Institute, University of Washington, Seattle, WA, USA

4. Department of Ophthalmology, University of Washington, Seattle, WA, USA

5. American Academy of Ophthalmology, San Francisco, CA, USA

6. Department of Ophthalmology, Massachusetts Eye and Ear, Harvard Medical School, Boston, MA