Despite lower socioeconomic status (SES) and higher disease burden, Hispanics in the US paradoxically display equal or lower mortality on average than non-Hispanic whites. Our objective was to determine if the “Hispanic paradox” occurs among patients with rheumatoid arthritis (RA).
In a cohort of 706 RA patients, we compared differences in RA severity and comorbidity between Hispanic and non-Hispanic white ethnic groups at baseline. Cox proportional hazards models were used to estimate and compare mortality risk between Hispanics and non-Hispanic whites.
We studied 706 patients with RA, of whom 434 were Hispanic and 272 were non-Hispanic white. Hispanics had significantly lower SES, greater inflammation, as well as higher tender and swollen joint counts. Patients were observed for 6,639 patient-years, during which time 229 deaths occurred by the censoring date (rate 3.4 per 100 person-years; 95% confidence interval 3.0, 3.9). Age- and sex-adjusted mortality was not significantly different between the 2 ethnic groups (hazard ratio [HR] 0.96). After adjustment for comorbidities, RA severity, and level of acculturation, mortality among Hispanics was lower (HR 0.56, P = 0.004).
Despite greater severity in most clinical manifestations and lower SES among Hispanics, paradoxically, their mortality was not increased. Further research is needed to understand the mechanisms underlying this survival paradox.
Rheumatoid arthritis (RA) is a chronic disease characterized by inflammation within the joints and affects approximately 1.3 million Americans (). Although ultimate causes of death in patients with RA are similar to the general population (), mortality rates in RA patients are 1.5–1.6 times higher, with cardiovascular disease being the most frequent cause of death ([2, 3]). The disease pathology and increased mortality seen in RA are incompletely understood, although there is thought to be a genetic component to disease severity (). Identification of a group of patients protected from the higher mortality of RA may open insights into the mechanisms of the increased mortality.
Hispanics are the fastest-growing ethnic group in the US. According to the US Census Bureau, more than half of the total population growth between 2000 and 2010 was a result of an increase in the Hispanic population (). Yet, in spite of lower socioeconomic status (SES) and higher rates for diabetes mellitus and obesity, several studies have shown equal or lower mortality rates for Hispanics in the US when compared to non-Hispanic whites ([6-11]). Given that the majority of current literature suggests lower SES is correlated with worse health outcomes as well as greater mortality ([12-15]), this comparable mortality finding has been referred to as the “Hispanic paradox” (). There have been several proposed hypotheses to describe these findings. Most notably, the “salmon bias hypothesis” (), cultural influences ([17, 18]), as well as genetic variation () are some of the key theories. The salmon bias hypothesis posits that foreign-born Hispanics return to their country at the end of life, and because we lack access to death records in those countries, their deaths are not recorded in the US, which makes them appear to have lower mortality. The sociocultural characteristics of the Hispanic population, specifically Mexican Americans, have been shown to play a role in alleviating negative effects of low SES, such as worse health outcomes (). Furthermore, recently, genetic factors have been suggested to play a role in the Hispanic paradox in chronic obstructive pulmonary disease (COPD), a partly heritable disease (). However, previous studies have not examined this paradox in RA.
Here, we studied mortality within a cohort of RA patients, adjusting for age, sex, SES, and clinical features of RA as well as comorbidity. We compared mortality rates between Hispanics and non-Hispanic whites. Because of their lower SES and greater disease severity, we hypothesized that the Hispanic patients would have increased mortality compared to the non-Hispanic whites.
Box 1. Significance & Innovations
There is increased mortality in rheumatoid arthritis (RA) patients compared to the general population. Identifying a group less affected by this mortality may suggest strategies for interventions aimed at reducing mortality
Our findings suggest that Hispanics with RA, despite greater disease severity and lower socioeconomic status, experience lower mortality than non-Hispanic whites with RA.
PATIENTS AND METHODS
The patients in this study have been described in other publications ([20-22]). Briefly, from 1996–2000, we recruited consecutive patients who met the 1987 American College of Rheumatology criteria for RA () from private and public rheumatology clinics in San Antonio, Texas. All patients participated in a comprehensive baseline evaluation of their clinical and psychosocial characteristics conducted by a physician and trained research assistants. Afterward, we invited them for annual followup evaluations. We followed patients until the censoring date of May 31, 2010. We learned about deaths from public databases, physicians, relatives, neighbors, and obituaries. In April and May 2010, we made further efforts to ascertain vital status by contacting patients directly. Of those we could not contact, we reviewed medical records to determine the date of last contact with the health system. All deceased were confirmed by death certificate. This study received approval from the institutional review board for the current protocol.
A trained interviewer asked patients for their date and place of birth, sex, and race/ethnicity. For the latter, the interviewer asked patients to self-identify as “white,” “black,” “Asian,” “Hispanic,” or “other.” Hispanic patients were further asked if they were of Mexican, Mexican American, Puerto Rican, Cuban, Central American, or other Hispanic background. In the cohort under study, these ethnic definitions identified culturally and genetically distinct groups ([22, 24]). For this analysis, we only included patients who selected “white” or “Hispanic.”
The level of social support was assessed using the Medical Outcomes Study Social Support Survey () at the baseline evaluation. The results from this survey provide 4 separate scales that describe how often each of the following different kinds of social support was made available to the patient: tangible support, affectionate support, positive interaction support, and emotional/informational support. The survey asked how often the following kinds of support were available, with higher scores indicating a greater amount of support (where 1 = none of the time, 2 = a little of the time, 3 = some of the time, 4 = most of the time, and 5 = all of the time). The tangible support scale was measured by questions such as how often there is “Someone to help you if you were confined to bed” and “Someone to take you to the doctor if you needed it.” Affectionate support was measured by questions such as how often there is “Someone who shows you love and affection” and “Someone who hugs you.” Positive interaction support was measured by questions such as how often there is “Someone to have a good time with.” Emotional/informational support was measured by questions such as how often there is “Someone to give you good advice about a crisis” and “Someone who understands your problems.”
Level of acculturation was assessed using a 4-question language-based scale by Deyo and colleagues (), as described previously (). A value of 0 represents no acculturation, while 4 represents full acculturation to the prevalent US English-speaking mainstream, or “Anglo,” culture. Acculturation was measured in the same way for both Hispanics and non-Hispanic whites. The questions and scoring of possible answers are 1) “Which was the first language you spoke as a child?” (English = 1, English and Spanish interchangeably = 0, Spanish = 0); 2) “What language do you speak at home?” (English = 1, English and Spanish interchangeably = 0, Spanish = 0); 3) “What language do you prefer to speak?” (English or English and Spanish interchangeably = 1, Spanish = 0); and 4) “Do you read any English?” (yes, some, or a little = 1, no = 0).
We classified SES according to Nam and Powers, using years of formal education, inflation-adjusted monthly household income, and current or past occupation to calculate an SES score on an ascending scale between 0 and 100 ().
We defined RA duration as the interval between the date of RA diagnosis and the time of study visits. At the baseline visit and at each of the followup visits, we asked patients to provide a list of their medications, and we reviewed pharmacy and medical records. We asked patients whether they were receiving glucocorticoids, and if so, we asked for the date at which these medications were first prescribed as well as the dose currently in use. We estimated cumulative oral glucocorticoid dose by multiplying the current daily dose by the number of days since glucocorticoids were initiated ().
A physician examined 48 joints for tenderness or pain on motion, for swelling or deformity, and for the presence of extraarticular subcutaneous nodules as previously described (). Joint examination Spearman-Brown reliability coefficients were 0.94 for tenderness/pain on motion, 0.90 for swelling, and 0.98 for deformity. We quantified joint damage on a plain radiograph view of both hands and wrists according to Sharp and colleagues (). In the current study, all radiographs were scored by a single rater (JFR).
For the erythrocyte sedimentation rate (ESR), we used the Westergren technique. If any ESR data were missing, we imputed the most recent adjacent followup visit levels into the missing data points. Approximately 57 baseline ESR values were replaced with second followup visits, and approximately 213 followup visits were replaced with adjacent ESR values. Serum rheumatoid factor (RF) was measured by the latex agglutination technique. Patients were considered seropositive if any determination during the study was positive. HLA–DRB1 genotyping was performed by Bio-Synthesis using polymerase chain reaction sequence–specific primer amplification with Fastype kits (). We classified HLA–DRB1 types using the 1996 World Health Organization Nomenclature Committee for Factors of the HLA System update ().
Hypertension was considered to be present if there was a physician's diagnosis in the medical record or an antihypertensive medication was prescribed, or from the measured blood pressure throughout the study (average systolic blood pressure ≥140 mm Hg). To assess obesity, we measured height and weight to calculate the body mass index (BMI; in kg/m2). Diabetes mellitus was considered to be present if a physician had recorded the diagnosis in the medical record, if the patient had taken antidiabetic medications during the course of the study, or if the fasting blood sugar level at a study visit was ≥126 mg/dl. Hypercholesterolemia was considered to be present if a physician recorded the diagnosis in the medical record, the patient had ever been prescribed lipid-lowering medication, or the fasting plasma cholesterol level measured during a study visit was ≥200 mg/dl. Smoking was divided into 3 categories: current smokers, past smokers, or never smokers. Persons who had ever smoked cigarettes were classified as current smokers if they continued to smoke, or were classified as past smokers if they had quit by the time of their most recent visit.
The comorbidity scale that we used was the Charlson Comorbidity Index (). This index records the presence or absence of 18 health problems, each one weighted for severity according to predefined values. The following health problems receive a weight of 1 in the index: myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease without hemiplegia, dementia, chronic pulmonary disease, connective tissue disease (including RA), peptic ulcer disease, mild liver disease, and diabetes mellitus without end-organ damage. The following health problems receive a weight of 2: hemiplegia, moderate or severe renal disease, diabetes mellitus with end-organ damage, malignant neoplasms, leukemia, and lymphoma. Moderate or severe liver disease receives a score of 3. Metastatic solid tumor and acquired immunodeficiency syndrome receive a score of 6. The final score is provided by the sum of the weights of each patient's health problems. The examining physician used a validated self-report questionnaire to interview the patients about the presence or absence of the predefined health problems (). The physician then reviewed available medical records to verify the problems reported by the patient. If discrepancies between the medical record and the self-reported responses arose, the case was discussed among 3 or 4 physicians in order to decide on the final score. Interrater reliability was 0.94 ().
We plotted age-adjusted Kaplan-Meier survival curves for Hispanic and non-Hispanic white patients. We used Cox proportional hazards regression to compare survival, adjusting for confounders, and t-tests or chi-square tests, as indicated, to test differences at baseline between Hispanic and non-Hispanic white patients (). We used generalized estimating equations to compare clinical characteristics between Hispanics and non-Hispanic whites over the course of the study, adjusting for confounders and accounting for repeated measures within patients (). After comparing Hispanics to non-Hispanic whites, we divided Hispanics into 2 groups, foreign born and US born, and compared mortality using Cox proportional hazards regression. We tested the assumption of proportional hazards for all variables in the Cox models by testing the hypothesis that the slope of their Schoenfeld residuals over time was zero (), and by visually inspecting plots of the survival probability against time. All analyses were conducted using a personal desktop computer with the Stata statistical software package, version 9.0.
We studied 706 patients who participated in a total of 4,155 observations. Time from enrollment until last followup, death, or the censoring date was 6,639 patient-years. There were 434 Hispanic patients and 272 non-Hispanic white patients.
The demographic and baseline clinical characteristics across ethnic groups are shown in Table 1. Hispanics were younger at the disease onset and had lower SES scores (Table 1). At baseline, Hispanics were significantly more likely to have diabetes mellitus and higher BMI, but lower cholesterol, incidence of hypertension, and Charlson comorbidity scores. In regard to the Charlson Comorbidity Index, non-Hispanic white patients showed statistically higher comorbidity scores at baseline, although this was found to be because of their older age. Furthermore, a higher percentage of non-Hispanic whites was found to have been past smokers, although we found no association between smoking status at the time of enrollment and comorbid conditions.
Table 1. Demographic and clinical characteristics of 706 RA patients at baseline according to ethnic group*
Non-Hispanic white (n = 272)
Hispanic (n = 434)
Values are the mean ± SD unless indicated otherwise. RA = rheumatoid arthritis; SES = socioeconomic status; BMI = body mass index; ESR = erythrocyte sedimentation rate.
aThere were 49 patients for whom no radiograph was obtained. The values shown are for a sample of 657 patients at baseline (252 non-Hispanic whites and 405 Hispanics).
bThere were 16 patients who were not HLA typed. The values shown are for 690 patients at baseline with corrected percentages (265 non-Hispanic whites and 425 Hispanics).
Regarding baseline clinical RA characteristics, the mean tender and swollen joint counts and ESR were higher in Hispanics when compared to non-Hispanic whites who had greater mean deformed joint counts and disease duration. Hispanics were more likely to be positive for RF, and non-Hispanic whites were more likely to be positive for the HLA–DRB1 shared epitope (Table 1).
To evaluate sociocultural differences, we analyzed levels of acculturation and social support across ethnic groups at baseline. Interestingly, non-Hispanic whites had significantly greater positive interaction support and emotional/informational support (Table 2).
Table 2. Acculturation and social support scores of 706 rheumatoid arthritis patients at baseline according to ethnic group
Non-Hispanic white (n = 272)
Hispanic (n = 434)
aThere was 1 patient for whom we were unable to obtain data on social support. The values shown are for 705 patients (271 non-Hispanic whites and 434 Hispanics).
We also performed comparisons over time between Hispanics and non-Hispanic whites. Table 3 shows age- and sex-adjusted comparisons between the 2 ethnic groups. Over time, Hispanic patients had significantly more tender and deformed joints, higher ESR, and higher modified Sharp scores. There was no significant difference in the number of swollen joints, prednisone use, presence of nodules, RF positivity, or Charlson comorbidity scores (Table 3).
Table 3. Age- and sex-adjusted comparisons between Hispanics and non-Hispanic whites in RA characteristics over time in 706 RA patients*
Values are the regression coefficient (95% confidence interval [95% CI]) unless indicated otherwise. RA = rheumatoid arthritis; ESR = erythrocyte sedimentation rate; OR = odds ratio.
aThere were 49 patients for whom no radiograph was obtained. The values shown are for a sample of 657 patients (252 non-Hispanic whites and 405 Hispanics) with 2,713 observations.
bValues are from generalized estimating equation regression models in which the clinical characteristic was the dependent variable. In these models, Hispanic = 1, non-Hispanic white = 0. Therefore, a positive coefficient or an OR >1.0 signifies greater mean or prevalence in the Hispanic patients.
By the censoring date, 229 deaths occurred, for a mortality rate of 3.4 per 100 patient-years (95% confidence interval [95% CI] 3.0, 3.9). Among Hispanics, there were 120 deaths in 4,204 patient-years for a rate of 2.8 deaths per 100 patient-years (95% CI 2.3, 3.3). Non-Hispanic whites had 109 deaths per 2,305 patient-years or 4.7 deaths per 100 patient-years (95% CI 3.9, 5.7). The unadjusted hazard ratio (HR) for mortality among Hispanics versus non-Hispanics was 0.58 (95% CI 0.45, 0.75) (Table 4). This difference in mortality lost significance after adjusting for age and sex, with an HR of 0.96 (95% CI 0.73, 1.26; P = 0.8) for Hispanics. Figure 1 shows age-adjusted Kaplan-Meier curves illustrating the probability of survival, comparing Hispanics to non-Hispanic whites. The lack of difference in mortality between the 2 ethnic groups persisted after adjusting for sex, SES, acculturation, social support, and RA duration. However, when we added RA manifestations such as ESR and tender and swollen joints, as well as comorbidities, to the model, the difference in mortality regained significance, with an HR of 0.56 (95% CI 0.38, 0.84; P = 0.004) favoring Hispanics.
Table 4. Mortality risk in Hispanics versus non-Hispanic whites in 706 RA patients*
RA = rheumatoid arthritis; 95% CI = 95% confidence interval; SES = socioeconomic status; ESR = erythrocyte sedimentation rate.
aCox proportional hazards models.
bThese models included 705 RA patients.
Age and sex adjusted
Age, sex, SES, acculturation level, and social support adjustedb
Age, sex, SES, acculturation level, social support, and duration adjustedb
Age, sex, SES, acculturation level, social support, duration, tender joints, swollen joints, and ESR adjustedb
Age, sex, SES, acculturation level, social support, duration, tender joints, swollen joints, ESR, and Charlson comorbidity adjustedb
We further hypothesized that we would observe a lower mortality rate among foreign-born Hispanics because, as suggested by the salmon bias hypothesis, foreign-born Hispanics may tend to return to die in their country of origin, where we lack access to their death records. When comparing foreign-born to US-born Hispanics, the US-born Hispanics accounted for 96 deaths in 3,157 patient-years for a rate of 3.04 deaths per 100 patient-years (95% CI 2.5, 3.7). Foreign-born Hispanics had 24 deaths in 1,047 patient-years for a rate of 2.3 deaths per 100 patient-years (95% CI 1.5, 3.4). Although no significant difference in mortality was observed between these 2 groups after age adjustment, foreign-born Hispanics showed an HR of 0.79 (95% CI 0.51, 1.24; P = 0.310) compared to US-born Hispanics, suggesting a tendency toward lower mortality. Furthermore, the non-Hispanic whites still showed no significant difference in age-adjusted mortality compared to both foreign-born (P = 0.316) and US-born (P = 0.996) Hispanics.
When we compared survival rates and clinical characteristics between Hispanics and non-Hispanic whites in a large cohort of patients with RA, we found non-Hispanic whites to have greater mortality. This was true even though, on average, Hispanics in our cohort had lower SES scores and worse RA manifestations than non-Hispanic whites. This association remained significant (P = 0.004) after adjusting for age, sex, SES scores, level of acculturation, clinical features of RA, and comorbidity.
A Hispanic survival paradox was first noted in 1986, when Markides and Coreil studied the current literature on mortality within southwestern Hispanics, most of whom were Mexican American (). These authors reviewed mortality in diseases that are leading causes of death, such as cancer, diabetes mellitus, and cardiovascular diseases, and found notable differences across ethnic groups, with lower mortality in Hispanics in almost all of these, with the exception of diabetes mellitus. Since then, a survival advantage among Hispanics has been documented in multiple studies ([6-11, 16]), although some have not been able to confirm this ([39-42]).
Our results seem to support the existence of a Hispanic mortality paradox in RA patients. Although the Hispanics in this RA sample were younger, age adjustment eliminated the initially significant mortality difference. When other variables that indicated greater disease severity in Hispanics, such as the joint counts and ESR, were added to the model, the Hispanic mortality advantage resurfaced. Each of these variables as well as the Charlson Comorbidity Index were significantly associated with mortality in age- and sex-adjusted Cox models in our sample, suggesting that the Hispanic patients were not affected by the negative effect of these variables on survival in the same way as non-Hispanic whites.
One hypothesis given to elucidate this paradox is known as the salmon bias hypothesis, in which foreign-born Hispanics are more likely to return to their country of origin at the end of life, resulting in an “immortal” population of Hispanics in the US (). We were able to provide a preliminary test of this hypothesis. If the salmon bias hypothesis were correct, we would have expected to observe lower mortality among foreign-born Hispanics, since we did not have access to any death records in foreign countries. Although we did not detect a significant difference between US- and foreign-born Hispanics, the direction of the effect suggests lower mortality in foreign-born Hispanics. Therefore, even though our findings do not support the salmon bias hypothesis, they did not exclude it either.
Another prevalent hypothesis suggests that culture affects mortality due to its influences on sociocultural, lifestyle, and behavioral factors that act as a buffering effect (). Several studies have suggested there are health disparities among minorities with RA (), and some authors have attributed these to differences in treatment preferences, education, and other ethnic and cultural influences, which would contribute to poorer outcomes for minorities ([44, 45]). Although these variables could have an effect on mortality differences (), data suggest cultural effects cannot fully explain the paradox (). Within our study, we adjusted for the level of acculturation using a 4-item language-based scale that has been shown to be reliable in distinguishing subsets of Mexican Americans (). Our models revealed that after adjusting for acculturation in models already accounting for comorbidity and RA severity, the lower mortality in Hispanics became more significant. However, adding the results from all of the 4 social support surveys did not affect the model or mortality difference, suggesting the survival advantage in Hispanics was not explained by differences in social support.
One example of a disease reported to be less severe in Hispanics that may have a genetic basis is COPD. Hispanics in New Mexico may have lower odds of COPD and lower decline in lung function, consistent with the paradox (). Because lung function is partly heritable (), these authors proposed that genetic ancestry could be a factor in their findings. The severity of joint destruction in RA may also be heritable (); therefore, it is possible that genetic differences could explain why the clinical manifestations of the disease were greater among Hispanics in the present study. There is a wide variety of genetic diversity within Hispanic subgroups in the US, with varying proportions of European, Native American, and African admixture ([48-50]). Although the focus of this study did not include these data, pinpointing genetic protective factors in Hispanics is an area of considerable scientific interest.
In conclusion, our findings are consistent with the Hispanic paradox, in that after adjusting for age and sex, mortality in Hispanics was not increased, despite lower SES and more severe RA. Several hypotheses have been proposed to explain this finding, such as migration effects of foreign-born Hispanics (), sociocultural effects ([17, 18]), or genetic influences (). Although we cannot support either the salmon bias hypothesis or sociocultural effects, it should be of note that after adjusting for RA manifestations associated with greater mortality, the Hispanic mortality advantage increased, despite having worse outcomes on these measures. Further research is needed in order to determine the underlying factors responsible for this phenomenon, including potential genetic, environmental, or sociocultural factors.
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. Escalante 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. Haas, del Rincon, Escalante.
Acquisition of data. Haas, del Rincon, Battafarano, Restrepo.
Analysis and interpretation of data. Molina, Escalante.