Secular trend of adhesive capsulitis




Adhesive capsulitis (AC) is a painful shoulder disorder resulting in restrictions of daily activities. Given the recent increase in risk factors for AC, such as diabetes mellitus, research is needed to examine if the incidence of AC has also increased. Therefore, the purpose of this study was to describe the secular trend of AC.


Subjects were from The Health Improvement Network, an electronic medical record database of general practices across the UK from 1995–2008. General practitioners diagnosed AC. We included subjects ages 40–79 years and defined an incident case as the first diagnosis of AC following at least 1 year of enrollment in the database. We examined the potential independent effects of age, calendar time, and birth cohort using an age–period–cohort model. To approximate effect estimates, we applied proportional hazards models and separately adjusted for diabetes mellitus.


Of 2,188,958 subjects (50.0% women), the incidence of AC was 2.36 and 3.38 per 1,000 person-years for men and women, respectively. Age–period–cohort model graphs suggested a secular trend due to a birth cohort, but not calendar effect. For each 10-year increment in birth cohort, the incidence rate increased by 8% in women (hazard ratio 1.08, 95% confidence interval 1.05–1.13). There was no increase in AC incidence for men. Adjusting for diabetes mellitus did not change these associations materially.


We found women born more recently are at a higher risk of developing AC compared with women born earlier. Future research is needed to identify risk factors that may be responsible for this increase.


Adhesive capsulitis (AC) is a shoulder disorder frequently encountered by rheumatologists, rehabilitation professionals, and orthopedic surgeons. AC is characterized by a loss in range of motion and pain, resulting in restrictions with activities of daily living, such as dressing and grooming (1). The duration of symptoms is reported to be between 4 and 24 months on average (2, 3). The disease course of AC typically follows 3 stages: pain, stiffening, and thawing or resolution. The annual health care and non–health care costs of AC have been estimated to be between $7,000 and $8,000 per episode (4, 5). Therefore, AC results in a substantial health burden on patients, on clinicians, and from a public health stand point.

The incidence of AC may be on the rise in the UK. The most recognized clinical risk factor for AC is diabetes mellitus (4–8). People with diabetes mellitus have 5 times the risk of AC compared with those without diabetes mellitus (7). As such, the prevalence of diabetes mellitus in the UK has been reported to increase from 2.8% in 1996 to 4.3% in 2005, and the incidence from 2.71 per 1,000 person-years to 4.42 per 1,000 person-years over the same time periods (9). Therefore, the incidence of AC may have subsequently increased.

To the best of our knowledge, there has yet to be a study that describes the secular trend of the incidence of AC. A large epidemiologic study describing the incidence of AC according to age, sex, calendar time, and birth cohort would be important for health providers who may expect to diagnose and treat more persons with AC, and to public policymakers for the added health care burden related to an increase in AC. Furthermore, investigating secular trends would facilitate our understanding of the etiology of this disabling condition.

To address these issues, we examined the secular trend of AC incidence from 1995 to 2008 among the general population in the UK and whether diabetes mellitus was related to the secular trend of AC.

Significance & Innovations

  • The incidence of adhesive capsulitis (AC) increases with age and is more common in women than men.

  • The incidence of AC is higher among women born more recently compared with those born earlier.


Study participants.

We used data from The Health Improvement Network (THIN), an electronic data set of 425 general practices within the UK. THIN includes 5.7 million patients, one-half of whom are being prospectively followed. Collected data include demographic information along with medical and prescribed drug histories. Subjects in THIN have been found to represent all sections of the UK general population (10). A more detailed description of THIN is provided elsewhere (11). We included subjects from THIN between 1995 and 2008 who were between the ages of 40 and 79 years, had at least 1 year of followup, were born after 1920, and had no prior history of AC. Our age inclusion criteria reflect the range when AC predominantly occurs and is diagnosed by health care professionals.

Defining AC.

We employed Read codes assigned by general practitioners to their patients to identify cases of AC given the absence of uniform diagnostic criteria for AC. Specifically, we employed 2 diagnostic Read codes, i.e., N210.00 and N210.12, which represent AC and frozen shoulder, respectively. We defined an incident case of AC as a subject who developed AC at least 1 year after entry into the THIN database and did not have a prior diagnosis of AC. We employed a 1-year cushion period to help exclude prevalent cases of AC at study entry.

We randomly sampled 50 cases and reviewed available electronic records for optional notes made within 1 month of the diagnosis of AC. We found notes specifically mentioning range of motion loss in 26% of subjects, shoulder pain in 30%, and a specialist confirmation of AC diagnosis in 4%.

Diabetes mellitus definition.

To identify subjects with diabetes mellitus, we employed a list of Read codes that included the keyword “diabetes” in the description (9). Persons were classified as having diabetes mellitus after the first diabetic Read code was reported in their record and we noted if subjects had diabetes mellitus at the time of AC diagnosis.

The study research protocol was approved by the Boston University Institutional Review Board and Multicenter Research Ethics Committee.

Statistical analysis.

We calculated sex-specific incidence rates of AC across all of the subjects. Specifically, person-years of followup for each subject were computed as the amount of time 1 year from the date of entry into the THIN database to the date of the first of the following events: 1) AC diagnosis, 2) date of transfer out of the practice, 3) date of death, 4) study closing date of December 31, 2008, or 5) date when the subject turned age 80 years. The incidence rate of AC was calculated by dividing the number of incident AC by the person-years of followup accumulated over the calendar years 1995–2008 (study period), and within each birth cohort category (i.e., 1920–1929, 1930–1939, 1940–1949, 1950–1959, and 1960–1969).

We compared the age-adjusted incidence of AC in men with that in women using hazard ratios (HRs) and 95% confidence intervals (95% CIs) calculated from Cox proportional hazards models. Age was entered into the regression model as a continuous variable, defined as the subject's age in years when first eligible for the study, i.e., 1 year after entry into the THIN database and at least 40 years of age.

We plotted the secular trend of AC according to age, calendar time, and birth cohort using an age–period– cohort (APC) model based on Poisson spline regression (12). This model provides a visual depiction of the independent effects of age, calendar time, and birth cohort by minimizing the identifiability problem, i.e., calendar time is the sum of birth date and age. We used the 2002 calendar year and those who were born between 1930 and 1939 as referent groups, since these groups were close to the mean values of our sample.

Since findings from the APC models indicated that the secular trends of AC incidence were strongly associated with birth cohort, we performed additional analyses to estimate the effect of birth cohort. We grouped subjects into 10-year birth cohorts and calculated the incidence rate of AC for each cohort. We estimated the sex-specific effects of birth cohort with proportional hazards models. To test for the linear trend of the birth cohort with the incidence of AC, we entered year of birth as a continuous variable into proportional hazards models. Finally, we examined whether adding the presence of diabetes mellitus (yes/no) prior to the development of AC into the proportional hazards model changed the effect of birth cohort on AC incidence.

We performed a secondary analysis restricting our sample to only those ages 60–69 years to maximize overlap between birth cohorts. All statistical calculations were performed using SAS software, version 9.2.


Of 2,188,958 subjects included in the analysis, 50.0% were women, 12% were born between 1920 and 1929, 16% between 1930 and 1939, 23% between 1940 and 1949, 26% between 1950 and 1959, and the rest were born between 1960 and 1969 (Table 1).

Table 1. Adhesive capsulitis (AC) cases and number of subjects within each birth cohort
Birth years1920–19291930–19391940–19491950–19591960–1969
Men, % of AC cases (no. of all subjects)10.8 (118,260)16.0 (175,432)23.1 (253,088)26.2 (287,430)23.9 (261,353)
Women, % of AC cases (no. of all subjects)13.5 (147,679)16.6 (181,252)22.5 (246,165)24.8 (271,577)22.6 (246,722)

Men and women had 7,146,249 and 7,258,975 person-years of followup time, respectively, during which 16,886 men and 24,505 women developed AC. The overall incidence rate of AC was higher in women (3.38 per 1,000 person-years) than in men (2.36 per 1,000 person-years). After adjusting for age, women had a 40% higher risk of incident AC compared with men (HR 1.40, 95% CI 1.38–1.43).

Graphs generated from the APC model indicated that age and birth cohort were associated with AC incidence, but calendar period was not (Figures 1–3). As shown in Figure 1, incidence of AC increases with age until 60 years, then somewhat plateaus for men and women thereafter. For each 5-year increment in age, AC incidence increased by 14% in men (HR 1.14, 95% CI 1.13–1.15) and by 8% in women (HR 1.08, 95% CI 1.07–1.08).

Figure 1.

Incidence rate and 95% confidence intervals of adhesive capsulitis by age.

Figure 2.

Hazard ratio of the incidence of adhesive capsulitis and 95% confidence intervals by calendar period.

Figure 3.

Hazard ratio of the incidence of adhesive capsulitis and 95% confidence intervals by birth cohort.

Compared with women born between 1930 and 1939, women born between 1920 and 1929 were 8% less likely to develop AC, and women born between 1940 and 1949, 1950 and 1959, and 1960 and 1969 were 1%, 11%, and 14% more likely to develop AC, respectively (Table 2). For each 10-year increment in birth cohort, the incidence rate of AC increased by 8% for women (HR 1.08, 95% CI 1.05–1.13). Adjustment for diabetes mellitus did not change these associations materially (Table 2).

Table 2. Effect of birth cohort with incidence of AC*
Birth cohortAC casesTotal person-yearsIncidence rate per 1,000 person-yearsHR (95% CI)HR (95% CI)
  • *

    AC = adhesive capsulitis; HR = hazard ratio; 95% CI = 95% confidence interval.

  • Age-adjusted HR.

  • Age- and diabetes mellitus–adjusted HR.

 1920–19293,262881,2443.700.92 (0.87–0.97)0.92 (0.87–0.98)
 1930–19395,9671,489,6654.001.00 (reference)1.00 (reference)
 1940–19497,5801,986,3383.821.01 (0.96–1.07)1.02 (0.97–1.07)
 1950–19596,0982,023,4483.011.11 (1.04–1.19)1.12 (1.05–1.20)
 1960–19691,598878,2791.821.14 (1.04–1.26)1.15 (1.04–1.27)
 1920–19291,983694,2502.850.94 (0.88–1.01)0.95 (0.88–1.02)
 1930–19394,4771,398,3583.201.00 (reference)1.00 (reference)
 1940–19495,7882,040,4082.830.97 (0.92–1.03)0.98 (0.93–1.04)
 1950–19593,7312,103,0701.770.99 (0.92–1.08)1.01 (0.93–1.09)
 1960–1969907910,1620.990.95 (0.84–1.07)0.97 (0.86–1.09)

We found no increase in the incidence of AC for men born more recently, and the overall trend for each 10-year increment in birth cohort was not associated with AC incidence (HR 0.97, 95% CI 0.93–1.01) (Table 2).

We found similar results when restricting the analysis to those subjects ages 60–69 years. For each 10-year increment in birth cohort, the incidence rate of AC increased by 10% for women ages 60–69 years at the time of diagnosis (HR 1.11, 95% CI 1.04–1.18). Adjustment for diabetes mellitus only attenuated this effect slightly to 8% (HR 1.08, 95% CI 1.01–1.15). For men there continued to be no statistically significant association of birth cohort with the incidence rate of AC adjusted for age and adjusted for age and diabetes mellitus, with effect estimates being close to the null (HR 1.00, 95% CI 0.99–1.01 and HR 0.99, 95% CI 0.99–1.01, respectively).


Among the general population within the UK, we found that the incidence of AC increases with age and is more common in women than men. Our results also suggest that the incidence of AC is higher among women born more recently compared with those born earlier. We did not find this trend among men, and adjusting our models for diabetes mellitus did not materially change these findings.

Our study adds to the literature that the incidence of AC is increasing among women born more recently. This is noteworthy given the burden this disease places on patients, providers, and society in general. While interventions exist to treat AC, such as corticosteroid injections (13), surgery (14), and physical therapy (15), limitations in shoulder function have been reported to last between 2 and 11 years (3). Our study also confirms that the incidence of AC increases with age and is more common in women than men. These findings suggest that clinicians should be aware that they may encounter more new cases of AC, especially among women and older adults.

Diabetes mellitus has been reported as a risk factor for AC (4–8), and has increased in the UK over the past 15 years (9). However, adjusting for the presence of diabetes mellitus prior to developing AC did not change the association between birth cohort and the incidence of AC materially. In general, a birth cohort effect is more pronounced by a factor that tends to be acute and only present within a specific generation (16). However, diabetes mellitus is a chronic factor whose risk has increased across several birth cohorts, albeit more so among those who were born more recently. Other chronic diseases associated with AC may have possibly increased over time as well, including Dupuytren's contractures, thyroid disorders, and bronchitis, although future work is needed to investigate the secular trends of these conditions.

One reason why AC incidence has increased could be the change in employment practices over the past 20 years for women. The number of working women ages 50–64 years has increased to a record high in the UK according to the 2002 Labour Force Survey, with increases in jobs with low levels of physical activity (17). Since AC is reported to be more common in sedentary vocations than in more active jobs (18), these shifts toward more recently born women taking more sedentary jobs could be a possible reason for the increase in AC incidence.

Our study has some limitations. First, the diagnosis of AC in our sample could not be validated beyond being diagnosed by a general practitioner. We were unable to employ a strict diagnostic criteria set, e.g., limited external shoulder rotation (19). Upon review of optional notes from 50 electronic records, general practice physicians most frequently noted shoulder pain and range of motion loss at the time of diagnosis of AC. While specialist-diagnosed AC would likely be more accurate, only a few specialist- referral records were available in this general practitioner–based database, likely reflecting the overall pattern of AC in the UK. However, the few specialist notes that were available for review from our sample records all confirmed the diagnosis of AC. While it is possible that some misclassification of this diagnosis occurred or that an increase of diagnosis occurred following more awareness of AC, we believe our sample is a reasonable reflection of general practitioner–diagnosed AC with our estimates of AC incidence approximating previously published data (20). Second, we did not differentiate between idiopathic versus secondary AC, i.e., painful restriction of shoulder range of motion due to rotator cuff tears, subacromial bursitis, or surgery (21). We performed a sensitivity analysis excluding all cases with a known history of shoulder surgery or pathology, such as rotator cuff tears, tendonitis, or bursitis, and found similar results. Third, all APC models, including the one we employed, still have limitations for separating the effects of age, calendar period, and birth cohort (16). Therefore, our study findings need to be replicated over a longer followup time before firm conclusions can be made.

Despite these limitations, our study has several strengths. First, we examined the incidence of AC within a large population-based study followed for more than a decade, which allowed us to examine the secular trend of the incidence of AC. Our findings not only suggest an overall increase in AC incidence, but also suggest that factors other than diabetes mellitus are responsible for this increase. Second, our findings were similar to previously published research studies. We report an incidence rate of 3.38 and 2.36 per 1,000 person-years for women and men, respectively, and this is consistent with a large Dutch observational study finding of 2.4 per 1,000 person-years for both men and women (20). Also, we found a peak incidence at age 60 years, which is consistent with a retrospective study of 2,370 persons with AC undergoing physical therapy (15) and the previously mentioned Dutch observational study (20). We report 59.2% of cases to be women, which is consistent with previous reports that slightly more women are likely to have AC compared with men (20, 22).

In conclusion, we found AC to be more common in women than in men, and the incidence of AC to increase with age. We found a secular trend of an increasing incidence of AC among women born more recently. The presence of diabetes mellitus did not materially attenuate the association of birth cohort with diabetes mellitus incidence, suggesting other factors are responsible for this increase.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. White 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 conception and design. White, Choi, Zhang.

Acquisition of data. White, Choi, Zhang.

Analysis and interpretation of data. White, Choi, Peloquin, Zhu, Zhang.