Study of Lateral Epicondylitis (Tennis Elbow) Using the Health Improvement Network Database

Patients

Data were extracted using Read codes; an National Health Service patient data coding system used in UK general practices. These map other coding systems including The International Classification of Diseases, ninth revision, and the Classification of Surgical Operations & Procedures, fourth revision.

Read codes specific for lateral epicondylitis were identified and used to extract the records of all patients in the database with a first recorded date of diagnosis between 1 January 1987 and 31 December 2006.

Patients who move their registration between general practices have their past medical histories coded at their new practice; this is standard practice. This leads to an acquisition bias in the THIN database; the incidence rate of a condition may be artificially inflated because the diagnosis of that condition may have been made many years previously, although the coded date of diagnosis may be the date of registration at that practice. To remove this effect, we excluded all diagnoses made within the first 6 months of a patient's registration period with their general practice. The ‘crude’ rates in the data are after this exclusion.

For each case of lateral epicondylitis, we extracted data on age, gender, deprivation score, UK health authority and year of first diagnosis. The deprivation score was based on the Townsend score [13]. This represents an index of socioeconomic deprivation in the range 1 to 5, with high scores relating to higher deprivation. Patients were grouped by age in intervals of 5 years for ease of analysis.

The age group incidence rates were adjusted with reference to the 1999 WHO European standard population [14] to obtain age-adjusted incidence rates.

Statistical analysis

The STATA software package, version 10 (StataCorp, College Station, TX, USA) was used for the statistical analysis. Incidence rates per 1000 person-years and 95% confidence intervals were calculated using Poisson regression analysis and compared using a chi-squared test for proportions. Age-adjusted incidence rates for lateral epicondylitis were calculated based on the European standard population using the direct standardization method.

The Poisson regression model was utilized to analyze the effect of age groups, gender, health authority and socioeconomic score on the incidence of lateral epicondylitis. The nested models were compared using the log-likelihood criterion. p < 0.05 was considered statistically significant.

Overall incidence

The overall crude incidence rate of lateral epicondylitis in our cohort between 1987 and 2006 was 2.45 per 1000 person-years [95% confidence interval (CI) = 2.43 to 2.46]. The incidence rate in males during this period was 2.63 (95% CI = 2.60 to 2.65) and was significantly higher than in females whose incidence rate was 2.55 (95% CI = 2.53 to 2.58) per 1000 person-years; p < 0.001).

The crude age incidence rates were adjusted with reference to the 1999 WHO European standard population [14] to obtain age-adjusted incidence rates. The adjusted incidence rate for males was 2.38 (95% CI = 2.36 to 2.40), although the incidence rate for females was significantly higher (incidence rate = 2.42; 95% CI = 2.40 to 2.45 per 1000 person-years; p < 0.001).

Age- and gender-based analysis

The incidence rate of lateral epicondylitis was related to age (Fig. 1) and higher rates were found in the age group aged 40 years to 54 years. The peak incidence rate was highest in the 45-year to 49-year-olds (incidence rate = 7.35; 95% CI = 7.23 to 7.45)

OPEN IN VIEWER

Fig. 1

Lateral epicondylitis incidence by age group.

We further stratified the incidence rate according to age group and gender. Females in the age group 40 years to 54 years had a significantly higher incidence rate than males in the same age category; however, throughout all higher and lower age groups, the male population showed a higher incidence rate than females (Fig. 2).

OPEN IN VIEWER

Fig. 2

Lateral epicondylitis incidence by age group subdivided by gender.

Incidence by deprivation

We used the Townsend deprivation index [13] to further analyse the incidence rate of lateral epicondylitis. The least deprived group (Townsend score: quintile 1) had the highest incidence rate of 2.86 (95% CI = 2.82 to 2.89). The incidence rate decreased such that the most deprived group (Townsend score: quintile 5) had an incidence rate of 2.00 (95% CI = 1.95 to 2.04).

Adjusted incidence rates were then calculated by deprivation score (Fig. 3), with a similar trend of decreasing incidence rate between Townsend quintile one (incidence rate = 2.54, 95% CI = 2.51 to 2.58) and Townsend quintile five (incidence rate = 2.26, 95% CI = 2.22 to 2.31).

OPEN IN VIEWER

Fig. 3

Lateral epicondylitis adjusted incidence rate by deprivation quintile.

The deprivation score quintiles were further stratified according to gender (Fig. 4). Within the Townsend quintile one population, the incidence rate was higher in males (incidence rate = 2.97; 95% CI = 2.92 to 3.01) than females (incidence rate = 2.76; 95% CI = 2.71 to 2.81; p < 0.001). This relationship was reversed in Townsend quintile five, where the female incidence rate (incidence rate = 2.11; 95% CI = 2.06 to 2.18) exceeded the male incidence rate (incidence rate = 1.88; 95% CI = 1.83 to 1.94; p < 0.001).

OPEN IN VIEWER

Fig. 4

Lateral epicondylitis incidence rate by deprivation quintile subdivided by gender.

The deprivation score was further examined in the age group 40 years to 49 years, where the incidence rate of lateral epicondylitis is the highest (Fig. 5). Within these age groups, the incidence rate of lateral epicondylitis in females is not significantly different between Townsend quintiles one and five. However, the male incidence rate in quintile 5 is significantly less than quintile 1.

OPEN IN VIEWER

Fig. 5

Lateral epicondylitis incidence rate by deprivation quintile subdivided by gender-specific age groups.

Incidence rates by geographical area

We compared the incidence rate of lateral epicondylitis in 13 health authorities across the UK for which data were available (Fig. 6). Incidence rates were comparable for all regions (range 2.05 to 2.68), except London where the incidence rate was 1.75. Adjusted incidence rates also confirm London as an outlier.

OPEN IN VIEWER

Fig. 6

Lateral epicondylitis incidence rate by UK health authority.

Incidence rates over time

Over a period of 20 years, the crude and adjusted incidence rate has fluctuated between 1.5 and 3 (Fig. 7). The incidence rate was lower in the first 4 years (1987 to 1990) and there was a further dip during the years 1995 to 1999.

OPEN IN VIEWER

Fig. 7

Lateral epicondylitis incidence rate per year.

Main findings

We have analyzed the incidence of lateral epicondylitis in the UK general population and further examined the effect of age, gender and social deprivation on these rates. The overall incidence rate of lateral epicondylitis was 2.45 per 1000 person-years. This was more common in males than females. After direct standardization, the age adjusted rates were 2.38 for males and 2.43 for females.

Highest crude incidence rates were found in the group aged 45 years to 55 years. Between the ages of 40 years to 54 years, females had the highest incidence rates. The least deprived areas of the population had the highest incidence rates (2.86 per 1000 person-years). Regional distribution of the incidence rates showed a fairly even spread across 13 UK Health Authorities, with the exception of London where incidence rates were significantly lower.

Strengths and limitations

The data employed in the present study are acquired from computerized general practice records used for routine clinical care. This has the advantage of reflecting real-life experience rather than the research setting, and being a prospective continuous data collection that avoids recall bias. One disadvantage, however, is that the general practitioner is likely to have made the diagnosis of lateral epicondylitis on varying strengths of clinical suspicion and without referral to a musculoskeletal specialist. However Herrett et al. found that estimates of the validity of data within the General Practice Research Database (GPRD) were high [15]. The GPRD is the formative name for THIN database.

Other studies using general practice computerized records to examine musculoskeletal conditions such as carpal tunnel syndrome [16] and Dupuytren's disease [17] found the data to be valid when examining the database for cases that had undergone surgical intervention. This type of analysis was not appropriate for lateral epicondylitis, which is mainly managed in the community, and is surgically treated in a minority of cases.

The presenting features of lateral epicondylitis are quite specific but studies use varying diagnostic criteria. Consensus on diagnostic criteria for upper limb disorders in a work setting was reached in 1998 [18]. Davis suggested that the specific criteria to be used in epidemiological studies of lateral epicondylitis should include; lateral epicondylar pain and epicondylar tenderness and pain on resisted extension of the wrist [19]. It is quite likely, however, that general practitioners have a heterogenous set of diagnostic criteria for an entry of lateral epicondylitis into the patient record. Indeed, it is impossible to confirm on what basis the recorded diagnoses were made in the THIN database. However, data from practices are routinely validated by internal checks and there are specific audits of the data supplied by individual practices. Only data meeting the minimum standards are added to the research database. It is probable, however, that a diagnosis of lateral epicondylitis has a good specificity but low sensitivity.

Comparison with the existing literature

Most previous studies either examined a cohort of patients or screened a population using surveys and questionnaires to identify a specific subsection of the population for further examination. The reliance of such studies on physical examination obviously limits the number of participants that can be included and therefore reduces the power of subsequent analysis. Only one other study that we identified used a population database, although these data were restricted to members of the US military [5].

Walker-Bone et al. used a two-staged cross-sectional study employing a screening questionnaire and physical examination in symptomatic individuals [5]. It is likely that this design permits improvements in sensitivity and specificity, although the total number of subjects (n = 9969) is still very small compared to the present study. They estimated the prevalence of lateral epicondylitis to be 1.3% in males and 1.1% in females. This value is lower than the incidence reported in the present study, although it is important to note that the point prevalence may be lower than the annual incidence because the condition is likely to be relapsing and remitting.

It is interesting to observe that males have a higher crude incidence rate than females, whereas females have a significantly higher adjusted incidence rate.

Social deprivation was assessed using the Townsend score. The findings of the present study are interesting because they suggest that lateral epicondylitis is mainly a disease of the higher socioeconomic groups. This is in conflict with previously held assumptions suggesting that lateral epicondylitis is a disease of manual workers. Unfortunately, a population study of this size was unable to examine occupation, sporting and social activities. Further well designed large-scale population studies are required to assess the impact of these factors.

The reason for the fluctuation in incidence rates over the period of the study data is uncertain. It is possible that progressive computerization means that, as practices join THIN, their data are coded retrospectively. This may result in an over-reporting of incidence in later years.