To determine time trends in the epidemiology of rheumatoid arthritis (RA) in a population-based cohort.
To determine time trends in the epidemiology of rheumatoid arthritis (RA) in a population-based cohort.
An inception cohort of residents of Rochester, Minnesota ≥18 years of age who first fulfilled the American College of Rheumatology 1987 criteria between January 1, 1955 and December 31, 1994 (applied retrospectively, as appropriate) was assembled and followed up until January 1, 2000. Incidence rates were estimated and were age- and sex-adjusted to the 1990 white population of the US. A birth cohort analysis was performed, and survival rates over time were examined.
The incidence cohort comprised 609 patients, 445 (73.1%) of whom were female and 164 (26.9%) were male, with a mean age at incidence of 58.0 years. The overall age- and sex-adjusted annual incidence of RA among Rochester, Minnesota, residents ≥18 years of age was 44.6/100,000 population (95% confidence interval 41.0–48.2). While the incidence rate fell progressively over the 4 decades of study, from 61.2/100,000 in 1955–1964, to 32.7/100,000 in 1985–1994, there were indications of cyclical trends over time. Birth cohort analysis showed diminishing incidence rates through successive cohorts following a peak in the 1880–1890 cohorts. Incidence rates increased with age until age 85, but peaked earlier in women than in men. The survival rate in RA patients was significantly lower than the expected rate in the general population (P < 0.001), and no improvement was noted over time.
The secular trends demonstrated in this study population, including the progressive decline in the incidence of RA over the last 40 years, suggest that an environmental factor may play a role in the etiology of RA.
Rheumatoid arthritis (RA) is a chronic disease with prevalence of ∽0.5–1% of the adult population in most Western countries (1). The etiology of this disease is only partly explained by genetic factors, and it has been suggested that environmental factors may play a significant role (2). Changes over time in the incidence and in the age and sex distributions of RA might implicate specific environmental factors and suggest new approaches to etiologic enquiry. Examining trends in survival over time can assist in monitoring the effectiveness of current treatment strategies.
We have previously reported secular trends in RA incidence over a 30-year period, which supported the concept that the epidemiology of RA is a dynamic process. The present study extends these observations for another 10 years, resulting in a 40-year population-based history of RA. Using this unique longitudinal resource, we explored time trends in the incidence and mortality of RA.
The population of Rochester, Minnesota, is well suited for an investigation of the epidemiology of RA because comprehensive medical records for all residents seeking medical care for over half a century are available. A record linkage system allows ready access to the medical records from all health care providers for the local population, including the Mayo Clinic and its affiliated hospitals, the Olmsted Medical Group, the Olmsted Community Hospital, local nursing homes, and the few private practitioners. The potential of this data system for use in population-based studies has previously been described (3, 4). This system ensures virtually complete ascertainment of all clinically recognized cases of RA among the residents of Rochester, Minnesota.
Using this data resource, an inception cohort of all cases of RA first diagnosed between January 1, 1955 and December 31, 1984 among Rochester, Minnesota, residents ≥35 years of age was identified, as previously described (5). We used the same method of case ascertainment to expand this cohort to include all Rochester residents ages ≥18 years who fulfilled the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 criteria for RA (6) between January 1, 1955 and December 31, 1994. The incidence date was defined as the first date of fulfillment of the ACR criteria (4 of the 7 criteria). All cases were followed up longitudinally until January 1, 2000 to determine their vital status.
Age- and sex-specific incidence rates were calculated using the number of incident cases as the numerator and population estimates based on decennial census counts as the denominator, with linear interpolation used to estimate the population size for intercensal years (7). Overall rates were age- and sex-adjusted to the 1990 white population of the US. Ninety-five percent confidence intervals (95% CIs) for the incidence rates were constructed using the assumption that the number of incident cases per year follows a Poisson distribution. Incidence trends were illustrated by plotting age- and sex-adjusted incident rates at the middle of overlapping 3-year time periods. A regression line was calculated for the logarithm of these values and plotted on the untransformed scale.
A birth cohort analysis was performed by grouping all of the incident cases over the 40-year span of the study into 10-year age and 10-year calendar year groups. The RA incidence rate for each birth cohort was plotted as a function of age at diagnosis and birth cohort.
Generalized linear models with the log-link function were used to evaluate the relationship between incidence rates, age, sex, and chronological time. Two-way interactions and higher-order polynomial terms for age were also examined.
Survival curves from the RA incidence date were estimated using the Kaplan-Meier method. Observed and expected survival was compared using the log-rank test, and the standardized mortality ratio was calculated for patients with RA. The expected survival was based on the sex and age of the study population and the death rates from the Minnesota (white population) life tables. Cox proportional hazards models were used to estimate the influence of age, sex, and calendar year on survival from the RA incidence date.
The preexisting incidence cohort comprised 425 individuals, of whom 1 was excluded from this report because of subsequent refusal to authorize the use of medical records for followup. In order to extend the cohort, we identified from the medical indices a further 2,152 individuals who had one or more diagnoses of inflammatory arthritis. Eighty-five of these medical records (4%) were not available for screening because of refusal to authorize research and 4 (0.2%) were excluded because part of the history was missing. The remaining 2,063 records were screened. Of these, 147 Rochester residents ≥18 years of age first fulfilled the ACR 1987 criteria for RA between 1985 and 1995, and a further 38 individuals 18–35 years of age fulfilled the criteria between 1955 and 1985 (total new cases 185).
Thus, the final incidence cohort comprised 609 cases of RA (according to the ACR 1987 criteria) first diagnosed between January 1, 1955 and December 31, 1994 in subjects who were residents of Rochester, Minnesota and ≥18 years of age at the incidence date. Of these, 164 (26.9%) were men and 445 (73.1%) were women (Table 1). The mean age at incidence of RA in this population was 58.0 years, and the mean followup was 14.2 years. The mean age at incidence remained stable over the 4 10-year time periods we examined (Table 1). Overall, 64% of those tested were positive for rheumatoid factor (RF) at some time, but no trend in this proportion was apparent over time. Of the patients with radiographic data, 51.7% had erosions or destructive changes typical of RA on hand or wrist radiographs, but no clear trend was apparent over the time periods examined (Table 1).
|Total no. of patients||149||148||165||147||609|
|No. (%) female||114 (76.5)||110 (74.3)||120 (72.7)||101 (68.7)||445 (73.1)|
|No. (%) male||35 (23.5)||38 (25.7)||45 (27.3)||46 (31.3)||164 (26.9)|
|Followup, mean ± SD years||18.7 ± 12.1||16.7 ± 9.8||13.4 ± 6.3||8.0 ± 3.6||14.2 ± 9.4|
|Age at incidence, years|
|Mean, median||57.5, 58.1||57.6, 57.7||58.9, 59.1||57.9, 58.7||58.0, 58.2|
|Minimum, maximum||22.9, 85.5||18.5, 88.6||19.9, 92.8||23.4, 89.2||18.5, 92.8|
|No. (%) positive of those tested||70 (65.4)||75 (57.3)||97 (65.5)||99 (67.3)||341 (64.0)|
|No. (%) with definite changes of those radiographed||63 (44.4)||77 (55.0)||79 (51.6)||74 (56.1)||293 (51.7)|
The overall age- and sex-adjusted annual incidence of RA among Rochester, Minnesota, residents ≥18 years of age (for the years 1955–1995) was 44.6/100,000 population (95% CI 41.0–48.2) (Table 2). The incidence of RA in men was extremely low in the 18–34 age group, after which it progressively increased with age until the oldest age group (≥85 years), when it decreased dramatically. In contrast, the incidence of RA in women rose until age 55–64, after which it steadily declined. Age-specific incidence rates varied considerably according to sex, with a 4:1 ratio of women to men in the 35–44 age group compared with a ratio of 1.1:1 in the 75–84 age category.
|No. of patients||Rate†||No. of patients||Rate†||No. of patients||Rate‡|
|Total or mean (95% CI)§||164||30.4 (25.6–35.1)||445||57.8 (52.4–63.2)||609||44.6 (41.0–48.2)|
Table 3 compares the age-adjusted annual incidence rates per 100,000 population ≥18 years of age for the 4 10-year time periods. A progressive decline in overall incidence rates over time was noted, from 61.2/100,000 (95% CI 51.2–71.3) in 1955–1964 to 32.7/100,000 (95% CI 27.3–38.0) in 1985–1994. This trend was apparent in both sexes, but was more marked in women. In fact, the ratio of female to male cases of RA diminished over time, from 2.2:1 in the 1955–1964 cohort to 1.6:1 in the 1985–1994 cohort.
|No. of patients||Rate(95% CI)†||No. of patients||Rate(95% CI)†||No. of patients||Rate(95% CI)‡|
|1955–1964||35||36.9 (24.2–49.6)||114||83.0 (67.6–98.3)||149||61.2 (51.2–71.3)|
|1965–1974||38||31.0 (20.9–41.0)||110||61.8 (50.1–73.4)||148||47.3 (39.5–55.0)|
|1975–1984||45||32.0 (22.4–41.6)||120||59.1 (48.4–69.7)||165||46.0 (38.9–53.2)|
|1985–1994||46||25.6 (18.1–33.2)||101||39.9 (32.0–47.8)||147||32.7 (27.3–38.0)|
In Figure 1, the 3-year moving average annual incidence rates per 100,000 population are plotted to illustrate trends over the entire study period from 1955 to 1995. The overall progressive decline over time is clearly shown by the regression line fitted to the data. There appears to be a cyclical pattern in the annual incidence rates, with peak incidence occurring in the late 1950s to early 1960s, the mid 1970s, and the early 1980s, with troughs occurring later in these decades. When men and women were evaluated separately (Figure 2), these peaks and troughs were found to occur at different times. In the early 1960s and in the period around 1970, the RA incidence in men was low, while that in women peaked, causing separation of the curves at these times. A contrasting pattern was present in the late 1980s, when relatively low rates in women and high rates in men led to a crossing of the incidence curves.
Birth cohort analyses of the incidence rates by 10-year intervals for women and men are illustrated in Figures 3 and 4, respectively. The graphs show that the 1880 and 1890 birth cohorts had unusually high incidence rates in their eighth decade of life, and in women of the 1890 cohort, there was a high incidence in the seventh decade. Examining successive birth cohorts, the incidence rates for RA fell progressively in both men and women in all age groups. This is consistent with the previously observed decline in incidence rates over time.
The generalized linear models analysis examining the incidence of RA showed that calendar year, age, and sex were significantly related to the incidence of RA (Table 4). There was a significant (P< 0.001) linear decrease in incidence with calendar year. Both linear and quadratic effects of age were significant in this model (P< 0.001 for both), which is consistent with our earlier observation that the incidence of RA increases with age until age 85, after which it falls again. In addition, the age–sex interaction noted in Table 2 was found to be significant, since the age at which the RA incidence peaks in women is significantly lower than that in men (P< 0.0001). The apparent change in the male:female ratio over time was not significant, however, in this model (P = 0.35 for interaction term).
|Variable||Coefficient||Standard error of the coefficient||P|
|Age and sex||0.024||0.006||<0.001|
Survival in this cohort was significantly lower than that in the general population, with a standardized mortality ratio of 1.27 (95% CI 1.13–1.41). Death certificate information was available on 329 of the 334 patients who had died at the end of the followup period (98.5%). Of these, cardiovascular disease was listed as the primary cause of death in 37.4%, cerebrovascular disease in 9.4%, pulmonary disease in 10.0%, malignancy in 10.3%, infection in 15.2%, and in 17.6%, death was attributed to other causes.
Kaplan-Meier curves comparing survival among patients in each of the 4 decades of RA incidence were nearly identical, and the log-rank test showed no statistically significant difference among them (P = 0.97). Cox proportional hazards models showed that calendar year was not a significant predictor of survival (P= 0.56). The hazards ratio for death per 10-year increment in calendar year was 1.018 (95% CI 0.91–1.14). Together, these analyses demonstrated no evidence of change in survival over time.
This study is the first to show 40-year trends in the incidence and mortality of RA in a geographically defined population. Our findings show a definite and statistically significant progressive decline in the incidence of RA over the study period, 1955–1995. We also found a difference in the age distribution between men and women and an apparent cyclical variation in disease incidence over time.
The overall age- and sex-adjusted incidence rate in this study was lower than that previously reported from this population: 44.6/100,000 compared with 75.3/100,000 in the previous study (5). This is, in large part, due to the inclusion of the 18–35 age group in the present study. The incidence in this age group, which was excluded from the previous study, is considerably lower than that in older groups.
The overall incidence rate reported here (44.6/100,000 population) is similar to the reported rate in Finland between 1980 and 1985, where the overall RA incidence using the same criteria was 46/100,000 population (8). The rate we report for women during the most recent time period (39.9/100,000) was also similar to the rate reported from the Norfolk Arthritis Register in the UK for the period 1990–1991 (34.3/100,000) (9). However, the rate we report for men (25.6/100,000) is notably higher than the 12.5/100,000 population reported in that study. The reason for this difference in incidence rates in men is unclear and may simply reflect random error associated with the relatively small number of men with RA in both studies.
Our finding of a progressive decline in RA incidence over time, affecting both men and women, has been documented in a number of populations over the last few decades. This finding was first reported in a study performed in this center more than 20 years ago (10). Since then, a similar decline has been noted in other US populations (11), as well as in the UK (12, 13) and Finland (8). Furthermore, a marked drop in both the incidence and prevalence of RA in Pima Indians over a 25-year period provides convincing evidence that rates are decreasing in this population, which previously had a relatively high incidence of RA (14).
We performed a birth cohort analysis to examine whether the reduction in RA incidence over time might be partly explained by year of birth. Such analyses have been shown to be useful for elucidating possible etiologic mechanisms in other chronic diseases (15–17). We observed a relatively high incidence of RA in individuals born in the latter years of the nineteenth century that progressively declined over subsequent birth cohorts. The only other study in recent years in which a birth cohort analysis was performed found a contrasting pattern, with peak incidence in women occurring in later birth cohorts (from 1915 to 1939) (18). Both analyses do, however, show a progressive decline in RA incidence through more recent birth cohorts.
These trends in incidence over time and through successive birth cohorts are likely to be explained by changing exposure to some environmental factor or factors. Such an environmental factor may take the form of a decline in exposure to some factor that promotes the development of RA or, alternatively, an increase in exposure to a protective factor.
The declining incidence rate in women, along with the frequent use of exogenous estrogens for contraceptive and hormone replacement purposes since the 1960s, led to the suggestion that exogenous estrogens might be a protective factor (10, 19). A number of cohort and case–control studies examining this question have been performed, but the results have been inconsistent (20).
A possible hormonal influence is also suggested by the differing age distributions in men and women, which has also been noted in studies from the UK and Finland (9, 18). The peak RA incidence in women occurs ∽10 years before that in men and coincides with the early years after menopause, when endogenous estrogen exposure declines dramatically.
The temporal and cyclical pattern we observed in the incidence of RA is consistent with an infectious etiology, since this pattern could reflect the cyclical change in the epidemicity of an infectious disease. Although a recent study found no evidence of time or seasonal clustering of incident cases as would be expected with a viral etiology, an endemic agent with a latent period between exposure and disease could be responsible for the development of RA in genetically susceptible individuals (21). It is interesting to note that the incidence of (and mortality from) certain infections in the general population, for example, tuberculosis, dramatically declined in the earlier half of the last century. Thus, the later birth cohorts in this study are considerably less likely to have been exposed to such infectious diseases. However, it is unclear how an infection would cause RA incidence rates to peak in men and women at different times. An infectious etiology for RA has been extensively sought, but no causative microorganism has, as yet, been identified (22).
We did not find evidence that the proportion of patients with RF positivity or erosive changes on radiographs had diminished over time. This is in contrast to the findings of 2 previous studies (8, 23). One of these studies reported that the severity of RA, as measured in terms of RF, rheumatoid nodules, and radiographic erosions, is declining (23). It should be noted, however, that the 3 factors used as severity markers in that study are included in the ACR 1987 criteria for RA. Thus, if the severity of RA were diminishing, and these features were becoming less common in RA patients, patients with inflammatory arthritis would be less likely to fulfill the criteria for RA. This could, in turn, contribute to a decline in RA incidence rates.
Finally, we were unable to demonstrate a statistically significant decline in mortality among RA patients over the 40-year time period. This finding concurs with previous findings that survival in patients with RA has not improved in recent years, despite significant advances in treatment (24, 25).
The strengths of this study include its population-based sampling and use of a systematic and standardized approach to case identification. It has been suggested that the apparent decline in incidence may reflect changes in diagnostic practice, with the application of more rigid criteria over time. The application of the same standardized criteria for case ascertainment over the 40-year study period ensured that this potential bias did not play a role in our study.
While selection bias is a possible limitation, it is expected to be minimal in this study because all incident cases of RA diagnosed in the study population during the period of interest have been ascertained through a medical record linkage system. Although use of such a system for case ascertainment has the limitation that RA cases that do not come to medical attention may have been missed, this pattern is unlikely to have changed over time and would not explain the variation in incidence rates over time. Moreover, we believe that all cases of RA are likely to come to medical attention.
Another potential limitation relates to the intensity of testing for RF and bony erosions. It is possible that patients with RA incidence in earlier years had blood tests and radiographs performed less often and that this may have masked a true decline in the prevalence of RF and bony erosions.
As a result of the introduction of a privacy law in Minnesota in 1997, one of the original cohort of 425 cases was excluded from the present study and 85 histories from among 2,152 potential subjects were not available for screening because these individuals had refused authorization for research. This is unlikely to have significantly affected the incidence rate, since the number of RA cases resulting from screening these extra histories would not have been expected to exceed a total of 8.
Finally, some racial and ethnic groups are underrepresented in Rochester, Minnesota, where the population in 1990 was 96% white, according to the US census data. The results of our population-based study are therefore only generalizable to the white population of the US.
In conclusion, we have demonstrated sharp and progressive declines in the incidence of RA in a US population over the last 4 decades. These findings point to the need for additional research into the role of environmental, infectious, and hormonal factors in the etiology of this disease.