Lateral Epicondylitis and Tobacco Use: A Case-Control Study

Abstract

Background

Although lateral epicondylitis (LE) is a very common tendinopathy, we understand little about the etiology of the disease. Tobacco use has been associated with other tendinopathies, and the purpose of this study is to determine if there is an association between the incidence of lateral epicondylitis and tobacco use.

Methods

We performed a retrospective cohort study of adult patients diagnosed with lateral epicondylitis. Patients from a single orthopaedic surgeon's practice with LE were matched to control patients with other common upper extremity conditions based on age, gender, and occupation. A total of 65 case patients and 217 control patients were included in the study. The incidence of smoking in patients with lateral epicondylitis was compared to the incidence of smoking in the control group.

Results

Of the LE patients, 30/65 (46.2%) were non-smokers, 23/65 (35.4%) were former smokers, and 12/65 (18.5%) were current smokers. Of the control patients, 121/217 (55.8%) were non-smokers, 45/217 (20.7%) were former smokers, and 51/217 (23.5%) were current smokers. The odds of LE patients being former or current smokers compared to control patients were 1.45 times higher, but this was not statistically significant. Among people who did not smoke at the time of presentation, the odds of being a former smoker were 2.28 times higher in LE patients than in controls, which was statistically significant

Conclusions

The odds of being a former smoker were significantly higher in patients with lateral epicondylitis compared to patients with other upper extremity conditions. Although it did not reach statistical significance, the odds of being former or current smokers were also higher in the LE group. These results suggest a relationship between smoking history and incidence of lateral epicondylitis, though more research is needed to determine the exact nature of the relationship.

Level of Evidence

Prognostic, Level III

Introduction

Lateral epicondylitis (LE), or “tennis elbow,” is a common cause of elbow pain, effecting nearly 1% of working-age adults^1,2. Despite the prevalence of lateral epicondylitis, the etiology of the disease is not well understood. Several studies have evaluated the relationship between LE and patient factors such as occupational activities, sex, and age^1,2,3. These studies have shown a link between manual labor and LE as well as a higher incidence of LE in patients ages 45–54^1,2,4,5.

Biopsies of tendon tissue in patients undergoing surgery for lateral epicondylitis reveal a pattern of poorly organized collagen and invasion with abnormal vascular structures known as tendinosis⁶. These changes are similar in appearance to tendon changes in other tendinopathies such as rotator cuff tears^7,8. Though there is not a wealth of data on the relationship between smoking status and LE, a link between tobacco use and rotator cuff tears has been shown in several recent studies. Kane et al. found an increased incidence and increased severity of rotator cuff tears in cadavers with a history of smoking⁹. Likewise, Baumgarten et al. found an increased incidence of rotator cuff tears was associated with any history of smoking, a history of smoking within 10 years of onset of shoulder pain, and increased pack-years of tobacco use¹⁰.

Because of the similarity in microscopic pathology between rotator cuff tears and LE, we questioned whether tobacco use was related to LE. We identified two published studies addressing this question. The first study, a cross-sectional study from Finland, found that patients with LE were more likely to be regular tobacco users (odds ratio [OR] 3.4)¹. The second study, a case-control study from England that utilized diagnostic and demographic information from a database, found that previous smoking history was a risk factor for lateral epicondylitis (OR 1.20) but that current smoking status was not¹¹. However, this study was limited in that there were no specific diagnostic criteria for LE.

Understanding the relationship between tobacco use and LE could have an important impact on treatment. Mallon et al. found that smokers had worse outcomes than nonsmokers following open rotator cuff repairs¹². If tobacco use is related to tendinosis, evaluating the efficacy of various treatments in smokers vs. non-smokers could yield important direction for treatment of patients with LE. The goal of this study was to evaluate the relationship between tobacco use and incidence of lateral epicondylitis. We hypothesized that patients with lateral epicondylitis are more likely to regularly use tobacco than the general population. In this study we compared the smoking rates in a group of LE patients to the smoking rates in a control group of patients with other upper extremity conditions. We also compared the smoking rates of the LE patients and control patients to the state average of the study population.

Patients and Methods

Subject Selection

This was a retrospective case control study approved by the site's institutional review board. We reviewed patients who presented to a single orthopaedic surgeon specializing in upper extremity conditions between 2009 and 2012. Patients were identified by query of billing codes (ICD9 codes) through hospital databases. Data was collected from the electronic medical record.

Eligible subjects for study were 18 years or older at the time of presentation, and had either a diagnosis of LE by ICD-9 code (case patients), or one of the ten common upper extremity injuries listed in Table 1 (control patients). Patients were excluded if their records contained incomplete smoking status or occupation information, or if the cause of LE was determined by the principal investigator to be due to high energy trauma. All patients were examined by a single orthopaedic surgeon. The diagnosis of LE was made based on the presence of tenderness over the lateral epicondyle and pain with resisted wrist extension with the elbow in extension. Because the pool of potential controls was large, a subset was selected using the random number generator (RAND) function in Excel (Microsoft Corporation, Redmond, WA, USA).

Table 1.

ICD9 Diagnosis Codes

ICD9 Code	Description	Cohort
726.32	Lateral epicondylitis	Case
726.32A	Lateral epicondylitis	Case
354.2	Ulnar nerve lesion	Control
816.01	Fx middle/proximal phalanx- hand-CL	Control
727.05	Tenosynovitis hand/wrist NEC	Control
815.00	Fx metacarpal NOS-CLOSED	Control
727.43	Ganglion NOS	Control
727.04	Radial styloid tenosynovitis	Control
715.34	LOC osteoarthritis NOS-HAND	Control
727.03	Trigger finger	Control
813.42	Fx distal radius NEC-CL	Control
354.0	Carpal tunnel syndrome	Control

Case-control match

Occupations were organized into three classes (light, moderate, and heavy) based on the perceived frequency that workers would engage in high risk activities identified by Van Rijn et al., Walker-Bone et al., and Haahr and Anderson^2,5,13. The high risk factors for lateral epicondylitis identified in the literature included forceful work, repetitive movements, working with hands or arms in a non-neutral position, working with the neck twisted, working with neck bent forwards >2h/day, handling tools > 1 kg, handling loads > 20 kg at least 10/day, use of hand/arm vibrating tools >1h/day, carrying weights on one shoulder, and lifting weights >5kg in one hand. Table 2 shows the classification of specific occupations and occupation fields according to the above criteria. A separate investigator independently assessed occupation class in a subset of 20 LE patients and 20 control patients for purposes of demonstrating inter-rater agreement. Each LE patient in the case cohort was matched with up to four controls using a greedy matching algorithm according to age, gender, and occupation class¹⁴. In order to be considered a match, a control had to be within 5 years of age from a case, and have the same occupation class and gender. Any cases or controls that remained unmatched were not included in the analysis.

Table 2.

Classification of Occupations

Occupation Class	Type of work
Light	Health care worker, lab worker, office worker, stay at home mom, teacher, student, retired*, unemployed*
Medium	Law enforcement, food service, army, truck driver, mail handler, physical/message therapy
Heavy	Construction, labor, maintenance, mechanic, factory/assembly line worker

^*Due to variability in the documentation of jobs in the electronic medical records, past work history was not factored into the classification of retired or unemployed patients.

Data Analysis

The sample was described in terms of mean age, proportion female and proportion of subjects reporting moderate or high occupational risks of injury. We used the AC-1 statistic for assessing the extent of agreement by multiple reviewers about a subject's occupation class.

The primary hypothesis was tested using conditional logistic regression, which modeled the probability of having the exposure of smoking as a function of case/control status, conditional on the age, gender, and occupation cluster to which the individual patient was assigned by the matching algorithm. A priori, we decided to use two definitions of smoking exposure: (1) current smoking (CS) and (2) past or current smoking (ever smoking; ES). Comparisons of smoking rates with the current state averages were done using exact binomial tests, treating estimates from the 2010 [Blinded] Adult Tobacco Survey¹⁵ as known population values (16.7% current smokers, 27.3% former smokers, and 56.6% never smokers). In all cases, tests were two-sided, with α = 0.05. Thus, 95% confidence intervals (CI) are used in this manuscript to describe variability in our estimates. When describing the uncertainty around a sample proportion, the Clopper-Pearson confidence bounds are used. Analysis was done using SAS 9.2 (SAS Institute Inc., Cary, NC, USA).

Results

A total of 65 patients with LE met our entry criteria. From 934 eligible control patients identified by ICD9 code, we randomly selected 408 of them to classify according to occupation intensity, in order to make them eligible for matching with cases. Out of these 408 subjects, 217 of them were matched and included in the study. Of the 65 LE patients, 33 were matched to four control subjects, 22 were matched to three controls, nine were matched to two controls each, and one LE patient could only be matched with a single control. Thus, the sample consisted of 282 subjects (Figure 1). In the 40 subjects for which occupational intensity was assessed by multiple investigators, the inter-rater reliability between the two investigators was high (AC1 0.971).

Figure 1. Breakdown of total study population.

Lateral epicondylitis patients had a mean age of 49.4 years (CI: 47.0 – 51.0), with the observed range of ages spanning from 25 to 87 years. Of the 65 cases, 23 (35.4%) were male (CI: 23.9 −48.2%). Fifty-two LE subjects were classified as having light occupational activity (80%; CI: 68.2 – 88.9%), while seven were classified as having moderate occupational activity (10.8%; CI: 4.4 – 20.9%) and six were classified as having heavy occupational activity (9.2%, CI: 3.5 – 19.0%).

Thirty of the 65 LE patients (46.2%) were never-smokers, 23/65 (35.4%) were former smokers, and 12/65 (18.5%) were current smokers. Of the control patients, 121/217 (55.8%) were never-smokers, 45/217 (20.7%) were former smokers, and 51/217 (23.5%) were current smokers.

When analyzed within age, gender, and occupation-matched clusters, the odds of LE patients being ever smokers (current or former) were 1.45 (CI: 0.83 – 2.52, p = 0.194). LE patients were not more likely to be current smokers than controls (OR 0.71, CI 0.36 −1.41, p =0.327) (Table 3). However, LE patients were significantly more likely to be former smokers compared to controls (OR 2.28, CI 1.11–4.68, p=0.024).

Table 3.

Smoking Rates in Lateral Epicondylitis Cases and Controls

Smoking Status	Control Rate (%)	LE Case Rate (%)
Never	55.8 (48.9 - 62.5)	46.2 (33.7 - 59.0)
Former	20.7 (15.5-26.7)*	35.4 (23.9 - 48.2)
Current	23.5 (18.0-29.7)*	18.5 (9.9 - 30.0)
Ever Smoking	44.2 (37.5 - 51.1)	53.8 (41.0-66.3)

^*Indicates a p<0.05 difference from the [blinded] state average rate in 2010

The 23.5% rate of current smoking in the control group was significantly higher than the state average of 16.7% (p=0.012)¹⁵. The rate of current smoking seen in LE patients of 18.5% was not different at a statistically significant level from the [Blinded] state average (p = 0.803)¹⁵. Compared to the general state population, the rate of former smoking was significantly lower in controls (p=0.032), but not significantly different in LE cases (p=0.190). The rate of ever smoking was not different from state averages in either LE cases (p=0.141) or controls (p = 0.996).

Discussion

Previous research has shown a link between tobacco use and tendinopathies. The purpose of this study was to determine if there was an association between patient smoking history and lateral epicondylitis. We hypothesized that patients with lateral epicondylitis were more likely to use tobacco than the general population. We did not find a statistically significant difference in current smokers between the LE patients and control cohorts for this study. The data did trend towards our hypothesis, but due to a small sample size, it did not reach statistical significance. However, there was a significantly higher rate of former smokers in the LE patients when compared with the control cohort and when compared with the state averages¹⁵.

The lifetime smoking rate (ever smoking) was noticeably higher in LE patients compared to controls and the state average, but this was not statistically significant (Table 3). Failure to detect a statistically significant difference could have been due to the small sample size. A priori, we calculated that a study that obtained 60 cases and 180 age, gender, and occupational intensity-matched controls would allow for a conditional logistic regression to have 84% power to detect a true increase in odds ratio of 2.5. If we assume that the true underlying odds of lateral epicondylitis are equal to our best estimate from this data (OR = 1.52), then a sample of this size has approximately 37% power to detect such an effect. Thus, it should be noted that a true increase of LE odds among tobacco users may indeed exist, but the likelihood of obtaining a sample of this size that demonstrates such an effect with the typical level of statistical certainty (p<0.05) is poor. Further studies examining this effect in similar populations should recruit 200 or more cases, in order to have a type II error rate of <20%.

The results from our study are similar to the findings from Titchener et al. which used data from The Health Improvement Network, a large database of electronic medical records collected from general practices throughout the United Kingdom¹¹. They also found a statistically significant increase in risk of LE in former smokers, but not in current smokers¹¹. In our study population, we found a significant age difference in age between levels of smoking exposure. Former smokers were significantly older (two-sided Wilcoxon p value= 0.03) than current or never smokers (Table 4). One explanation for the higher odds of lateral epicondylitis occurring in a former smoker than a current smoker or non-smoker in our study is that the extra risk of lateral epicondylitis with smoking is one that accumulates over time.

Table 4.

Differences in Age by Smoking Status

Smoking Status (Overall Sample)	N	Mean Age	95% Confidence Interval
Never Smoker	151	49.5	47.9 - 51.1
Former Smoker	68	52.4	49.5 - 55.3
Current Smoker	63	48.0	45.9 - 50.0

Although our study includes fewer patients than Titchener, a strength of our study is that all patients were examined and diagnosed based on specific clinical criteria by a single orthopaedic surgeon fellowship trained in hand surgery.

Another limitation of this study is the failure to quantify tobacco exposure in our patients. Several studies support the dose dependent effects of tobacco on rotator cuff pathology, which may also be the case in regards to LE.⁹ In our study, there was an increased incidence of history of any tobacco use in patients with LE compared to the control patients. Although the rates of current smokers were very similar between the two groups, it is possible that the increased incidence of LE in former smokers could be because the damage from tobacco has already occurred prior to smoking cessation. Future research efforts should be made to examine the possible dose dependent relationship between lateral epicondylitis and tobacco use.

A limitation that this study shares with all other case-control studies is that, due to the fact that we do not know the size of the population being sampled (i.e., the denominator is unknown), we cannot make causal inferences about smoking upon lateral epicondylitis incidence. However, the design does allow us to observe and report associations (odds ratios) without assigning causality.

Another limitation of our study was that our control group had a higher rate of current smoking than the general population of the study state¹⁵. We elected to use other clinic patients as our controls so that we could account for age and occupation, but a comparison to a control group from the general population may have yielded a larger difference between the LE patients and the control group with regard to tobacco use. A related limitation is that our study treats the estimated smoking rates reported in the 2010 [Blinded] Adult Tobacco Survey as known population values, rather than as proportions observed in a sample of roughly 7,000 interviewees. This assumption was used because in the format that the [Blinded] Adult Tobacco Survey results were presented, the exact numerators for current, former and never smokers were not available for use in a comparison such as Fisher's exact test¹⁵. We felt that the assumption was justifiable due to the large sample size of the survey, especially relative to our own data.

Conclusion

In conclusion, our study found a higher incidence of LE in patients with any history of tobacco use. Current smokers did not have an increased incidence of lateral epicondylitis in our study. Larger studies could better define the possible dose dependent nature of the relationship between tobacco and tendinopathies such as lateral epicondylitis.