It has been reported that patients with diabetes have a higher prevalence of frozen shoulder and a diminished response to treatment compared with the general population1. There is controversy regarding the influence of glycemic control on the prevalence of frozen shoulder in diabetic patients. It has been suggested that surgical outcomes are poorer for diabetic patients with frozen shoulder than they are for nondiabetic patients with idiopathic frozen shoulder1. If it is possible to decrease the risk of the development of frozen shoulder through the identification of contributing factors, such knowledge would facilitate preventative care counseling and improve medical care. Some investigators have suggested that inadequate glycemic control as measured by the glycosylated hemoglobin A1c (HbA1c) level can identify diabetic patients who are at a higher risk of the development of shoulder pain, shoulder stiffness, and shoulder disability2.
The purpose of this study was to determine the relationship between glycemic control and the prevalence of diabetic frozen shoulder in a large diabetic population. A secondary study objective was to analyze the relationships of diabetic treatment duration, insulin usage, and associated diabetic end-stage sequelae on the prevalence of diabetic frozen shoulder.
A cross-sectional analysis was performed to determine the prevalence of frozen shoulder in diabetic patients stratified by HbA1c level in a large health maintenance organization (HMO) in southern California in 2007. The diabetic database of our regional healthcare system provides a robust population of more than 210,000 diabetics among 3.3 million members. Implementation of a unified electronic medical record system has facilitated the process of capturing large sets of patient data. Patients were identified through a search of the HMO’s electronic health record system for specific International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes. Patients were identified as diabetic if they had received an inpatient or outpatient ICD-9 diagnosis code of 250.xx, 357.2, 362.0, 366.41, or 648.0; an HbA1c result of >6.5%; or a dispensing order for insulin or an oral hypoglycemic medication. Adhesive capsulitis of the shoulder was identified by ICD-9 diagnosis code 726.0. Previous validation work in the database for an unrelated study by independent reviewers to assess a similar algorithm for identifying patients with diabetes mellitus revealed a sensitivity of 93% and a positive predictive value of 95%3.
The following data were obtained: patient age, sex, ICD-9 principal diagnoses codes, date of initial diabetes treatment and frozen shoulder diagnosis, HbA1c laboratory values (with dates of testing), history of frozen shoulder treatment (with dates of treatment), and associated diabetic sequalae on the basis of ICD-9 codes, including trigger finger (727.03), Dupuytren contracture (728.6), autonomic neuropathy (337.1), peripheral neuropathy (357.2), retinopathy (362.0), and nephropathy (250.4).
Hemoglobin A1c levels were identified on the basis of an average of test results obtained within six months prior to the initial diagnosis of frozen shoulder. Hemoglobin A1c levels were used to categorize patients into three different groupings: those with HbA1c levels of <7% (indicating tightly controlled diabetes), levels of ≥7% but ≤9% (indicating mildly controlled diabetes), and >9% (indicating poorly controlled diabetes). Our reasoning for stratification was based on the fact that the American Diabetes Association recommends that the HbA1c level should be below 7% for most diabetic patients4. Other studies have shown that there is a higher postoperative risk of adverse events in patients who have uncontrolled diabetes and elevated HbA1c levels (i.e., above 9%)5,6. For patients without frozen shoulder, HbA1c laboratory results were the average of readings from the first six months of the year 2007.
Medical treatment was separated into diet-controlled, insulin-only, insulin with oral hypoglycemic medication, and oral hypoglycemic medication-only groups. Duration of diabetes treatment was based on the earliest date of identification of the patient as being diabetic. Patients were separated into duration groups of less than five years, five to ten years, and more than ten years of known diagnosis.
Exclusion criteria were set to eliminate external and overlapping causes of shoulder stiffness. These criteria included an age of less than eighteen years, prior shoulder surgery before the diagnosis of frozen shoulder, or a diagnosis of frozen shoulder with a concurrent diagnosis of shoulder osteoarthritis. These patients were identified with use of demographic data and ICD-9/Current Procedural Terminology (CPT) codes (715.11 [shoulder osteoarthritis], 716.91 [arthropathy], v45.89 [postprocedural status], v43.61 [shoulder joint replacement], 80.21 [shoulder arthroscopy], 81.80 [total shoulder replacement], 81.83 [other repair of shoulder], 83.63 [rotator cuff repair], 23472 [total shoulder arthroplasty], 23470 [arthroplasty, glenohumeral joint; hemiarthroplasty], 29805 [shoulder arthroscopy diagnostic], and 29827 [arthroscopic rotator cuff repair]) and data were verified with use of internal chart review. To further confirm the accuracy of frozen shoulder diagnosis with use of this protocol, prior to our analysis, charts from a random sample of one hundred patients who had been identified as having frozen shoulder in our study were reviewed. Documentation of shoulder pain with loss of passive shoulder motion in elevation and external rotation was identified in 94% of these patient charts.
Institutional review board approval was obtained through our institution before initiating the study.
Source of Funding
No external funding was received for this research.
Statistical Methods
The study was powered to detect an odds ratio of 1.3, based on the relative proportions of diabetics who had an HbA1c level >7.0%, between the diabetic patients with frozen shoulder and those without frozen shoulder. In our population of patients with diabetes, 50% have an HbA1c level >7.0%. If there is a 6.5 percentage point increase in the proportion of diabetic patients who have a HbA1c level >7.0 among our diabetic patients with frozen shoulder (56.5% vs. 50%) and 33.3 times as many diabetic patients without frozen shoulder, it would only take 492 diabetic patients with frozen shoulder and 16,391 diabetic patients without a frozen shoulder to establish the 1.3 odds ratio as significant at the 0.05 level with 80% power. As a post hoc calculation, with 1000 patients in the group of diabetic patients with frozen shoulder and 144,000 in the general group of patients with diabetes, and a proportion difference of 4.5% (54.5% vs. 50%), an odds ratio of 1.2 could be declared significant at the 0.05 level with 80% power.
Continuous variable data were reported as the mean plus the standard deviation. Categorical variables were reported as proportions with 95% confidence intervals (CI). Comparisons of categorical data, including demographics and diabetic sequelae, were analyzed with chi-square tests. P values for association were deemed significant if <0.05.
For our primary study objective, the independent relationship of HbA1c on frozen shoulder was analyzed with logistic regression models, adjusting for insulin use and sex.
For our secondary study objective, we compared the prevalence of frozen shoulder among diabetics by duration using logistic regression models. Frozen shoulder prevalence was compared in insulin-requiring versus non-insulin-requiring diabetic groups with use of chi-square testing. The prevalence of diabetic-associated sequelae in those patients with frozen shoulder versus those patients without frozen shoulder was also analyzed using chi-square testing.
Demographics
There were 201,513 diabetic patients identified in the diabetes patient database; 57,343 did not have an HbA1c lab test result drawn during the time-frame parameters of the study. This left 144,170 patients who fit the criteria for inclusion in the study. The mean patient age was fifty-nine years and 48.1% of diabetics were female. There was a concurrent diagnosis of diabetic nephropathy in 41.2% of patients, and 20.7% of patients received a concurrent diagnosis of diabetic retinopathy (Table I).
There was a 0.65% prevalence of frozen shoulder in all diabetics. Women were 1.27 times more likely to have frozen shoulder compared with men (95% CI, 1.14 to 1.43). In insulin users, women were 1.47 times more likely to have frozen shoulder compared with men (95% CI, 1.18 to 1.84).
HbA1c Analysis
Of 1150 diabetic patients with frozen shoulder, there were 934 patients who had the level of HbA1c measured within six months before the initial diagnosis of frozen shoulder. Although there was a lower percentage of patients with frozen shoulder in the group of 397 diabetic patients (42.5% of the 934 patients with HbA1c data available) who had a mean HbA1c level of <7.0% than there was in the group of 537 diabetic patients (57.5% of the 934 patients) who had a mean level of ≥7.0%, the difference was not significant (p = 0.44) (Fig. 1). The crude odds ratio for mean HbA1c was 0.95 (Table II). After adjusting for insulin use, the adjusted odds ratio for HbA1c increased to 1.06. A second multivariate logistic regression model including insulin use and sex did not make a difference in the odds ratio for mean HbA1c.
Diabetic Medication Treatment Analysis
Insulin users (with or without use of oral hypoglycemics) were 1.93 times more likely than non-insulin-dependent patients to have a diagnosis of frozen shoulder (p < 0.0001) (Table II). This increased to 1.96 times more likely when adjusted for HbA1c level. Insulin-only users (without use of oral hypoglycemics) were 2.8 times more likely to have frozen shoulder than those who did not use insulin or oral hypoglycemic medications (p < 0.0001). Those on oral hypoglycemic drugs were 1.5 times more likely to have frozen shoulder than those who did not use insulin or oral hypoglycemic drugs (p < 0.0001). A comparison of mean HbA1c levels between insulin users and non-insulin users showed insulin users had a higher mean HbA1c level (8.08 vs. 7.45, p < 0.0001).
Effect of the Duration of Diabetes Treatment
When compared with patients who had less than five years of diabetes treatment, the patients who had five to ten years of treatment had an odds ratio of 1.40 for the development of frozen shoulder (95% CI, 1.20 to 1.63), while those whose treatment lasted longer than ten years had an odds ratio of 1.85 (95% Wald confidence limit: 1.59, 2.16) to develop frozen shoulder, after controlling for insulin use. All odds ratios were significant. This association was significant (p < 0.0001), indicating that patients living with and being treated for diabetes longer are more likely to develop frozen shoulder.
Diabetic Soft-Tissue and Microvascular Sequelae
When identifying diabetic sequelae among all diabetic patients, those with frozen shoulder were significantly more likely (p < 0.0001) to have a diagnosis of each of the four microvascular conditions than patients without frozen shoulder. For instance, autonomic neuropathy was diagnosed in 8.4% of diabetic patients with frozen shoulder and in 5.3% of diabetic patients without frozen shoulder (p < 0.0001); peripheral neuropathy was diagnosed in 33% of diabetic patients with frozen shoulder and in only 25% of diabetic patients without frozen shoulder (p < 0.0001); and nephropathy was diagnosed in 41% of diabetic patients with frozen shoulder and in only 38% of diabetic patients without frozen shoulder (p = 0.047). Twenty-six percent of diabetic patients with frozen shoulder had diabetic retinopathy compared with only 19% of diabetic patients without frozen shoulder (p < 0.0001). Furthermore, 19.8% of diabetic patients with frozen shoulder also developed trigger finger (stenosing tenosynovitis of the digit) and/or Dupuytren contracture as compared with only 9.4% of diabetic patients without frozen shoulder (p < 0001).
It is unclear why frozen shoulder in diabetic patients has such a prolonged and recalcitrant course as compared with the course of frozen shoulder in the normal population7,8. Increased glycosylation of connective tissue is a commonly proposed theory. The HbA1c blood level is proportional to the average blood glucose concentration over the previous months. Measuring glycosylated hemoglobin levels assesses the effectiveness of glycemic control therapy by monitoring long-term serum glucose regulation9. Accumulating irreversible cross-links between various protein molecules may lead to joint stiffness through glycosylation10. In fact, multiple soft-tissue contracture disorders of the hand have also been associated with diabetic populations11. In our study population, a higher percentage of diabetic patients with frozen shoulder also developed trigger finger (stenosing tenosynovitis of the digit) and/or Dupuytren contracture compared with diabetic patients without frozen shoulder. This fact may support a systemic etiology for the development of joint contracture in diabetics. Prior studies imply that if diabetic patients can effectively manage their disease, they will exhibit fewer symptoms of sequelae (retinopathy, neuropathy, and nephropathy) compared with diabetics who have higher HbA1c levels12. If adhesive capsulitis of the shoulder is a direct consequence of poor glycemic control, its clinical appearance and natural progression should be related to HbA1c levels, duration of diabetes, or severity of associated diabetes complications.
There is controversy regarding the effects of glycemic control on the prevalence of frozen shoulder. Thomas et al. found no significant differences in HbA1c level between their diabetic patients without frozen shoulder and their thirty-four diabetic patients with frozen shoulder13. Pal et al. studied 109 consecutive diabetic patients and found no correlation of HbA1c levels with limited joint mobility14. Others have found a higher prevalence of subjective shoulder pain and disability in patients with elevated HbA1c levels and prior eye surgery2. We did not find a significant relationship between HbA1c level and prevalence of frozen shoulder among diabetic patients in our study. It may be that HbA1c levels do not adequately reflect glycemic instability on a time scale that is necessary to detect an increased prevalence of frozen shoulder. Although the acute onset of symptoms is inconsistent with this hypothesis, it may be that frozen shoulder develops only when a specific glycemic threshold is reached.
Similar to Bridgman8, we found that insulin use was associated with increased risk for frozen shoulder in diabetic patients. Thomas et al.13 did not find a similar effect; however, they did find that a longer duration of diabetes correlated with an increased prevalence of frozen shoulder. When controlling for insulin use, we found that the odds ratio of developing frozen shoulder was 1.8 when the duration of diabetes was more than ten years as compared with less than five years. Furthermore, diabetic sequelae, such as peripheral neuropathy, nephropathy, autonomic neuropathy, and retinopathy, were all significantly increased (p < 0.0001) in the patients with frozen shoulder as compared with the patients without frozen shoulder. This compares similarly with the results from the study by Balci et al., who demonstrated an odds ratio of 2.2 for retinopathy together with frozen shoulder15. Our analysis did not show a direct cause and effect for these factors with frozen shoulder and suggested the use of insulin to be a covariate. Intuitively, our studied factors (HbA1c, insulin use, diabetic duration, and end-stage disease) influence one another, which makes it difficult to analyze those factors independently. While insulin use and duration may be a sign of disease severity, a normalized level of HbA1c is a sign of effectiveness of diabetes treatment. Diabetic duration, therefore, may not be directly linked to the etiology or pathophysiology of frozen shoulder.
Prior studies have assessed diabetic patients for the presence of frozen shoulder via a questionnaire on musculoskeletal complaints at the time of presentation in a medical clinic8,13,15. Other studies have attempted to identify diabetic patients who had nonspecific pain and stiffness in the shoulder joint (i.e., capsulitis) but have not specifically identified frozen shoulder as a diagnosis16,17. Our study was based on a captured diabetic population within a large HMO, and we made use of specific diagnosis codes for frozen shoulder as determined by the medical provider. This allowed us to analyze one of the largest frozen shoulder populations in the literature.
Limitations of the study include its retrospective nature and reliance on multiple physicians for diagnoses. It is possible that some providers misdiagnosed frozen shoulder and diagnosed it as another condition, such as impingement syndrome or rotator cuff tear, which would cause the prevalence of frozen shoulder to be underestimated in our population. Similarly, patients with other diagnoses might have been misdiagnosed as having frozen shoulder. Although the authors were unable to individually validate all diagnoses of frozen shoulder, a random sampling of patients’ charts prior to this analysis demonstrated a high rate of accuracy (94%) for the diagnosis of frozen shoulder. Not all diabetic patients with frozen shoulder had measurement of HbA1c levels within the predetermined six-month window period after initial diagnosis of frozen shoulder. This was not a controllable variable but theoretically could have affected the final analysis. However, we sought to include only HbA1c levels drawn near the onset of frozen shoulder to best characterize the influence of HbA1c level on the disease process. Examination of these missing patients revealed that they were ethnically similar to the patients who had HbA1c measurement but were, on the average, slightly younger and slightly more likely to be male. Similarly, the HbA1c levels that were used for the diabetic patients without frozen shoulder may not have been a complete reflection of the patients’ history of hyperglycemic control over the entire duration of their diabetic condition. Patients with frozen shoulder in our study may also have inadvertently been followed more closely for end-stage sequelae of the disease, thereby falsely elevating percentages by introducing detection bias. Lastly, the variable of diabetic duration was based on capturing specific identifying events for each patient that led to the diagnosis of diabetes. However, if the patient was enrolled in the healthcare insurance plan after the development of diabetes, his or her prior diabetic history would not have been identified. This possibility could have resulted in an underestimation of the actual time period from the patient’s initial diagnosis.
In summary, the results of the present study demonstrate that increased HbA1c levels do not affect the risk of the development of frozen shoulder in diabetic patients. The duration of diabetes treatment may be associated with an increased risk of the development of frozen shoulder, although a cause-and-effect relationship between the two has not been determined.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. None of the authors, or their institution(s), have had any financial relationship, in the thirty-six months prior to submission of this work, with any entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, no author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.