Association Between the Use of β-Adrenergic Receptor Agents and the Development of Venous Leg Ulcers FREE

Although it has been more than 25 years since β-adrenergic receptors were first noted to be expressed in the skin, their functional significance in this tissue has remained unclear. These receptors are expressed in keratinocytes, fibroblasts, and melanocytes of the skin, yet few diseases have been ascribed to their malfunction.¹ The few examples include atopic dermatitis and vitiligo, which are both reported to be associated with a local increase in synthesis and release of the β-adrenergic receptor catecholamine agonists.¹ In addition, in psoriasis, decreased expression of the β₂-adrenergic receptor in keratinocytes is thought to lead to a decrease in intracellular cAMP (cyclic adenosine monophosphate) and a concomitant increase in cell proliferation.² Likewise, few cutaneous clinical effects have been reported as a consequence of the systemic administration of pharmacologic agents that either activate or block these receptors. Two well-recognized associations are the β-adrenergic antagonist exacerbation of psoriasis and the historically important oculomucocutaneous syndrome.^{3 - 4} The current understanding of the mechanism by which β-adrenergic antagonism is associated with exacerbation of psoriasis is that the systemic antagonist blocks endogenous β-adrenergic receptor agonists in the skin, thus resulting in an increase in keratinocyte proliferation.⁵

Recent laboratory work has provided evidence to support a role for the β-adrenergic receptor in cutaneous wound repair. Specifically, β-adrenergic receptor agonists decrease the rate of keratinocyte migration in vitro, impair the ability of cultured keratinocytes to repair a scratch wound in a confluent monolayer, delay wound epithelialization in organ-cultured human skin wounds, and inhibit wound healing in vivo in a murine wound model.^{6 - 8} β-Adrenergic receptor antagonists, however, increase cultured keratinocytes' migratory speed and their ability to heal a scratch wound in vitro, speed up wound healing in organ-cultured human skin, and improve wound epithelialization in vivo in murine wounds.^{9 - 10}

Thus, one might anticipate that in patients who use β-adrenergic receptor agonists, there might be a deleterious effect on the propensity of those patients to heal wounds and, therefore, an increased incidence of chronic, nonhealing wounds. However, complicating the assumed association is the additional finding that although β-adrenergic receptor agonists impair many aspects that contribute to wound epithelialization, they can improve fibroblast function in the healing wound by increasing both fibroblast proliferation and migratory rates. Essentially, agonists have opposite effects on keratinocytes and dermal fibroblasts.¹¹ Thus, one could also reasonably envision a clinical scenario, like lipodermatosclerosis, which is a skin sign associated with VLUs and is likely caused by chronic inflammation, destruction, wounding, and scarring of the dermis and fat layer, where wound repair is mostly dependent on fibroblast function. In this scenario, by virtue of their fibroblast stimulatory effects, administration of β-adrenergic receptor agonists might improve wound repair or prevent lipodermatosclerosis, thereby preventing the onset of a VLU. And, one would thus reasonably propose that systemic administration of either β-adrenergic receptor antagonists or agonists as commonly administered for the treatment of asthma, hypertension, or after a myocardial infarction might have an effect on the onset of nonhealing wounds, such as VLUs. In fact, in an earlier study¹² on the incidence and prevalence of VLUs, we did notice unexplainable associations between patients with VLUs and those with asthma or congestive heart failure, angina, or myocardial infarction (diseases classically treated with β-adrenergic receptor agonists or antagonists) and the onset of a VLU.

Venous leg ulcers occur in about 1% of the population older than 45 years.^{12 - 13} Most are preceded by varicosities, leg edema, and lipodermatosclerosis.¹⁴ For those with lipodermatosclerosis, it is therefore likely that a VLU is really the outcome of repetitive wounding of the dermis that ultimately penetrates the epidermis. We have previously used this logic to demonstrate that estrogens, which have been shown in the laboratory to enhance wound repair, likely clinically promote healing too because those who use them are less likely to develop a VLU.¹⁵

Our current study was, therefore, undertaken to determine whether an epidemiologic association could be established in patients using β-adrenergic receptor agonists or antagonists and the incidence of chronic wounds.

This was a retrospective cohort study consisting of individuals registered in a large patient-record database called the General Practice Research Database (GPRD). Established in the United Kingdom in 1987, the GPRD is a medical records database used by general practitioners (GPs) as their primary means of tracking patient clinical information. We were provided access to the GPRD by the Epidemiology and Pharmacology Information Core (London, England). The GPRD GPs receive both financial inducements and penalties to ensure compliance with providing information in this electronic record. Approximately 1500 GPs representing 500 practices across the United Kingdom participated in the GPRD between 1987 and 2002. Patients were eligible to enroll in our study if they had been registered with a GPRD GP during this period. The total GPRD population available exceeds 9 million patients with over 35 million person-years of follow-up.

A GPRD-registered patient was eligible for our study if (1) they received care from a GP who is a member of the GPRD, (2) they had at least 2 consultations with the GP while they were 40 to 95 years of age, and (3) they did not have a diagnosis of a VLU for the first 6 months after the commencement of their database record. The 6-month period was based on several lines of clinical evidence and was previously evaluated by us.^{12 - 13} Participants maintained their eligibility until they died or transferred out of the practice, or if the practice stopped contributing data to the GPRD.

Diseases are classified in the GPRD database using a hierarchical coding system called Read codes. For this study, an eligible patient was considered to have a diagnosis of VLU using an explicit and previously evaluated (by direct query of the GP) algorithm consistent with the clinical diagnosis of VLU.¹² This algorithm also excludes those with lower extremity peripheral arterial disease. For a VLU to be termed associated with the use of a β-adrenergic receptor agent, it must have occurred in an eligible study participant at least 90 days after exposure to the agent, and for a VLU to be an outcome in those who did not use a β-adrenergic receptor agent, once eligible, the participant had to have been followed for at least 120 days (minimum time between receipt of 2 prescriptions and 90-day period of association).¹³

A participant's use of β-adrenergic receptor agonist or antagonist therapy was determined using the British National Formulary coding system. Separate variables were created for exposure to β-blocking antagonists and β-adrenergic receptor agonists. All individuals needed to have received at least 2 prescriptions separated by at least 30 days to be deemed exposed. All medications dispensed by a GPRD GP are coded in this system because GPs need to be compliant with NHS electronic prescription regulations.

β-Adrenergic receptor agents may be used to treat or prevent many medical conditions, and their use may vary by sex and age. These ailments and other illnesses were evaluated as potential confounders. As a result, the following variables were captured as potential confounders: age; sex; and history of angina, asthma, congestive heart failure, diabetes mellitus, emphysema, glaucoma, hypertension, or myocardial infarction. Confounders were selected for use in our adjusted models based on their potential clinical importance or if they changed the point estimates by more than 10%.

To separately assess the magnitude of the effect of the association between β-adrenergic receptor antagonists and β-adrenergic receptor agonists, single-variable and multivariable logistic regression models were used to estimate odds ratios (ORs). Both unadjusted (from a single-variable logistic model) and fully adjusted (from a multivariate logistic model) ORs are reported with 95% confidence intervals (CIs). Fully adjusted ORs were calculated by including confounders that were a priori considered to be clinically important or statistically important variables (based on a P value of < .10 in the single-variable model) in a logistic model. Age was evaluated both as a continuous variable and in deciles. The effects on the OR were identical, and for simplicity they are reported only as a continuous variable. Two-way interaction terms were also evaluated and were considered to be statistically significant at P < .10. Reported P values are for the Wald statistic or z statistic, calculated as the estimated coefficient divided by its standard error. We used Mantel-Haenszel models to estimate relative risks (RRs).

Evaluation of the appropriateness of the fully adjusted model included analyses for outliers, colinearity, tolerance and covariance, goodness-of-fit, and discrimination. In all cases, the appropriateness of the logistic regression model was confirmed.

To further explore our observations and to attempt to control for treatment selection bias, we also used propensity score methods. Basically, because treatment selection in cohort studies is not random but rather determined by the GP, selection bias could arise if the choice of therapy depends on patient factors related to the probability that a wound will occur. For example, if clinically everyone with asthma who has leg edema uses drug A and leg edema is associated with a VLU, then treatment selection for drug A would be biased toward those with leg edema and also a VLU. One of the important reasons for performing a randomized controlled trial is that it can essentially eliminate treatment selection bias because selection is determined by the randomization procedure. The propensity score technique can be used to statistically model treatment selection, thereby minimizing selection bias attributable to observed covariates. Propensity score techniques mimic the random assignment of a randomized controlled trial by balancing important variables involved in the selection of a therapy between those who received and those who did not receive a therapy.^{16 - 17}

For the current study, separate covariates were included if they were hypothesized to affect the selection of a patient to receive either of the β-adrenergic receptor agents. All of the covariates listed in the “Definition of Outcome, Exposures, and Confounders” subsection as potential confounders were included in these models. The ability of the model to discriminate between those who received 1 of the 2 agents and those who did not was estimated by the area under the receiver operating characteristic curve.

Patients were stratified into quintiles based on the distribution of propensity scores. Quintile-specific rates for the development of a VLU were calculated for those who used a β-adrenergic receptor agents and those who did not. An overall estimate of the association of the agent with the onset of a VLU was then calculated by summarizing the stratum (quintile) specific RRs using a Mantel-Haenszel technique. Before combining the quintile-specific data into a summary score, we used the Q-statistic for heterogeneity to determine whether the size of the treatment effect varied across quintiles. Finally, to estimate the summary of effectiveness in the setting of confounding and effect modification, multivariable logistic regression models were used.

Statistical analyses were conducted using Stata statistical software for Windows XP (version 9.2; StataCorp, College Station, Texas). This study was reviewed by the institutional review board of the University of Pennsylvania and previously reviewed by the Scientific and Ethical Advisory Group of the United Kingdom.

A total of 414 887 GPRD-registered patients met our study criteria for eligibility (Table 1). Their mean (SD) age was 61.8 (14.1) years (95% CI, 61.8-61.9) and the median age was 61 years (range, 49-73 years). With respect to sex, 240 592 (58%) were female, and 174 295 (42%) were male. Of these individuals, 62 886 were exposed to a β-adrenergic receptor agonist and 54 861 were exposed to a β-adrenergic receptor antagonist (6620 used both β-adrenergic receptor antagonists and agonists). Of those exposed to a β-adrenergic receptor agonist, 15.5% developed a VLU, whereas 18.4% of those not exposed developed a VLU (Table 1). Of those exposed to a β-adrenergic receptor antagonist, 18.2% developed a VLU, whereas 19.9% of those not exposed developed a VLU (Table 1). The OR of association between β-adrenergic receptor antagonist use and VLUs was 1.02 (95% CI, 0.99-1.04). The OR of association between β-adrenergic receptor agonist and VLUs was 0.84 (95% CI, 0.82-0.86). As has been previously shown,¹² those who developed a VLU were more likely to be older (the percentage with a VLU increased from younger to older age groups; P < .001) and female (P < .001).

Several multivariate analyses were conducted (Table 2). The relationship between β-adrenergic receptor antagonist use and VLUs was not notably confounded by (ie, altered point estimate by more than 10%) other covariates (Table 2) (see the “Definition of Outcome, Exposures, and Confounders” subsection in the “Methods” section for the listing of potential confounders). Indeed, the fully adjusted OR (adjusted for sex; age; and history of myocardial infarction, hypertension, angina, or congestive heart failure) was 1.04 (95% CI, 0.98-1.11). The relationship between β-adrenergic receptor agonist use and VLUs was confounded (changed the point estimate of the unadjusted model by more than 10%) by asthma and prior oral or inhaled glucocorticoid use. The adjusted OR (adjusted by sex, age, and history of asthma and glucocorticoid use) was 0.44 (95% CI, 0.42-0.45), meaning that those who used β-adrenergic receptor agonists were about 66% less likely to develop a VLU. This indicates that the use of glucocorticoids and a history of asthma masked the protective association of β-adrenergic receptor agonists use on the onset of a VLU. Finally, the most frequently used β-adrenergic receptor agonist in our data set was salbutamol (also known as albuterol). The fully adjusted OR for comparing users of salbutamol with those who did not use it was 0.40 (95% CI, 0.39-0.41), with respect to the onset of a VLU.

We further defined our associations by conducting a propensity score study (Table 2 and Table 3). Our propensity score for the selection of exposure to a β-adrenergic receptor antagonist had an area under the receiver operating characteristic curve of 0.78. We estimated an RR of 1.11 (95% CI, 1.09-1.14) for the association between use of β-adrenergic receptor antagonist and VLUs. However, there is notable heterogeneity in this estimate in that individuals in propensity quintile subgroups 1 to 4 did seem to have a protective association with respect to drug exposure and VLUs (Table 3) (eg, group 4: RR, 0.39 [95% CI, 0.35-0.43]). In group 5, which comprises those most likely to receive a β-adrenergic receptor antagonist, those who used a β-adrenergic antagonist were at increased risk for developing a VLU. For β-adrenergic receptor agonist use, our propensity score for the selection of exposure to a β-adrenergic receptor agonist had an area under the receiver operating characteristic curve of 0.83. Using the Mantel-Haenszel technique to combine the quintiles, the propensity score RR summary for β-adrenergic receptor agonist use was 0.76 (95% CI, 0.75-0.78). Clinically important heterogeneity was not noted. For both of our models, important covariates were evenly balanced between groups within each quintile.

Laboratory investigations have suggested that β-adrenergic receptors may be involved in cutaneous wound repair. β-Adrenergic receptors are present on keratinocytes and fibroblasts of the skin; however, their effects on keratinocytes and fibroblasts differ.^{9 ,11} In our current study, we have shown that those exposed to β-adrenergic receptor agonists are less likely to develop a VLU. This effect was noted in a propensity score model as well. We were unable to find a consistent association between the use of β-adrenergic receptor antagonist using multivariable logistic regression and the onset of a VLU. However, in our propensity score model we were able to demonstrate that a protective effect was noted in many of the propensity deciles (Table 2). This suggests that in a subset of patients, β-adrenergic receptor antagonists may also provide some protective effect.

Previous work^{9 - 10} in our laboratory has shown that β-adrenergic receptor antagonists increase cultured keratinocyte migratory speed and their ability to heal a scratch wound in vitro, speed up wound healing in organ-cultured human skin, and improve wound epithelialization in vivo in murine skin wounds. Although we were able to show in our propensity score study that some individuals who received β-adrenergic receptor antagonists were less likely to develop a VLU (Table 3; 2, 3, and 4), our adjusted estimate does not show a protective effect. This apparent heterogeneity may have 2 simple clinical explanations. First, we and others^{13 ,18 - 19} have previously shown that subtle peripheral vascular (arterial) disease of the lower extremity (eg, a lower limb ankle-brachial index of 0.7-0.9) is a risk factor for the development and healing of a VLU. As noted, the individuals in group 5 were the ones most likely to use β-adrenergic receptor antagonists, which means that they are also most likely to have hypertension and atherosclerotic vascular disease of the heart. These are also, therefore, individuals at highest risk for peripheral vascular disease, which, when subtle, we know from experience cannot be properly ascertained in this database.¹² As a result, the RRs that we note in group 5 may be caused by residual confounding. Second, genetic polymorphisms commonly for β-adrenergic receptor agents.²⁰ These polymorphisms have been shown to have an effect on the regulation of the receptor itself.²⁰ It is entirely possible that the effects of this agent depend on the genetic polymorphism that is present and that our heterogeneity has a genetic basis.

However, β-adrenergic receptor agonists increase both the migration and proliferative rate of dermal fibroblasts, and it is likely that this cell type is the one that is initially damaged in patients with lipodermatosclerosis.^{11 ,14} Thus, β-adrenergic receptor agonists could be predicted to enhance the repair process of the woundlike environment of early lipodermatosclerotic lesions and thus prevent overlying ulceration. There is also evidence that implicates β-adrenergic receptor activation with enhancement of angiogenesis.^{21 - 22} Thus, in addition to their function on dermal fibroblasts, β-adrenergic receptor agonists could improve chronic wounds by up-regulating their vascularization. Furthermore, inflammation plays an important part in the pathogenesis of lipodermatosclerosis and chronic VLUs.^{23 - 25} There are many examples in the literature (reviewed by Miyamoto et al²⁶ ) demonstrating the anti-inflammatory effects of β-adrenergic receptor agonists, which includes the fact that neutrophils express the β₂-adrenergic receptor and synthesize catecholamines. In addition, β-adrenergic receptor agonists decrease many aspects of neutrophil function, including their chemotaxis and recruitment, their ability to adhere to the endothelium, and their ability to generate reactive oxygen species and cytokine inflammatory mediators.²⁷ Therefore, either by its effects on dermal fibroblasts, angiogenesis, or on the inflammatory process, the systemic administration of β-adrenergic receptor agonists may decrease the individual's propensity to develop a nonhealing venous ulcer.

All observational studies may be limited by bias and confounding. Treatment selection bias is a critical concern when evaluating the efficacy of a medication in a study that does not randomly allocate treatment. Although the ultimate proof of the effectiveness of β-adrenergic receptor agents in wound repair will require a randomized clinical trial, we feel that treatment selection bias is less likely an issue in our study than in most observational studies. First, during the period of patient observation in this study (1987-2002), no one suspected that β-adrenergic receptor agents were likely to have clinically important effects on wound repair. More important, no one suspected that they had clinical activity with respect to the treatment of VLUs. Treatment selection with respect to these agents is likely based on the illnesses for which they are treatment options (eg, asthma, glaucoma, hypertension, myocardial infarction, and congestive heart failure). Second, with respect to β-adrenergic receptor agonists, the association of these agents and the association of asthma with respect to VLUs are in opposite directions. Although a decreased association with VLUs was noted in our unadjusted model, this effect was unmasked (more protective) once our models were adjusted for history of asthma and systemic glucocorticoid use. This effect was also noted across all strata using a propensity score model, which attempts to adjust for treatment selection bias. In essence, we have 2 different models that show that the use of β-adrenergic receptor agonists protects against the onset of VLUs. Finally, it is possible that our results are biased owing to confounding by indication. However, there are multiple different diseases associated with the use of these medications; our propensity scores, which are based on treatment selection, revealed similar results; and β-adrenergic receptor agents are used for illnesses both with an increased and decreased association with VLUs.¹²

Information bias is also problematic for observational studies. We have previously demonstrated our ability to use the GPRD to determine whether an individual has developed a VLU.¹² Although we do not know if our participants actually used β-adrenergic receptor agents, we do know that they were prescribed more than once. Our definition of exposure to β-adrenergic receptor agents was based on at least 2 prescriptions to the agent, and for an outcome to be associated with 1 of these agents it must have occurred 90 days after receipt of the agent. In any event, if individuals received these agents and did not use them, then the direction of the information bias should have biased our results toward the null (no association). It is very unlikely that anyone would have used β-adrenergic receptor agents without knowledge of the participant's GP, but again, if this did occur then the bias would have been toward the null (fewer unexposed individuals would have developed a VLU), and therefore our true effect estimate would have been even more protective.

In summary, we have shown a protective association between β-adrenergic receptor agonists and VLUs. There is strong laboratory evidence to support β-adrenergic receptor agonist–mediated modulation of dermal fibroblast function as the mechanism underpinning this epidemiologic finding. It is, however, possible that the benefit noted in our study is caused by another mechanism, such as local changes in arterial perfusion, and we encourage others to investigate this possibility. Although we were unable to find consistent evidence of an association between β-adrenergic receptor antagonist and VLUs, there was, however, a subgroup of patients who benefited from the use of these agents too. It may also be possible that individuals with specific illnesses, such as diabetes mellitus or congestive heart failure, might benefit more from these agents. This needs further study. It is important to realize that the evidence in this study should not be used as a rationale for treatment of VLUs with β-adrenergic receptor agents but rather should be compelling data for the consideration of a randomized clinical trial.

Correspondence: David J. Margolis, MD, PhD, 815 Blockley Hall, 423 Guardian Dr, University of Pennsylvania, Philadelphia, PA 19104 (dmargoli@cceb.med.upenn.edu).

Financial Disclosures: Dr Isseroff has a patent application on the use of adrenergic agents in wound repair that both she and the University of California, Davis, are currently preparing.

Accepted for Publication: November 9, 2006.

Author Contributions: Dr Margolis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Margolis, Hoffstad, and Isseroff. Acquisition of data: Margolis and Hoffstad. Analysis and interpretation of data: Margolis, Hoffstad, and Isseroff. Drafting of the manuscript: Margolis, Hoffstad, and Isseroff. Critical revision of the manuscript for important intellectual content: Margolis, Hoffstad, and Isseroff. Statistical analysis: Margolis and Hoffstad. Obtained funding: Margolis and Isseroff. Administrative, technical, or material support: Margolis, Hoffstad, and Isseroff. Study supervision: Margolis, Hoffstad, and Isseroff.

Funding/Support: This study was funded in part by National Institutes of Health grants AR02212 and AR44695 (Dr Margolis) and by AR44518 and Shriners Hospitals for Children grant No. 8550 (Dr Isseroff).

Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; or in the preparation, review, or approval of the manuscript.