American Journal of Respiratory and Critical Care Medicine

Despite reports of familial clustering of sarcoidosis, little empirical evidence exists that disease risk in family members of sarcoidosis cases is greater than that in the general population. To address this question, we estimated sarcoidosis familial relative risk using data on disease occurrence in 10,862 first- and 17,047 second-degree relatives of 706 age, sex, race, and geographically matched cases and controls who participated in the multicenter ACCESS (A Case– Control Etiology Study of Sarcoidosis) study from 1996 to 1999. Familial relative risk estimates were calculated using a logistic regression technique that accounted for the dependence between relatives. Sibs had the highest relative risk (odds ratio [OR] = 5.8; 95% confidence interval [CI] = 2.1–15.9), followed by avuncular relationships (OR = 5.7; 95% CI = 1.6–20.7), grandparents (OR = 5.2; 95% CI = 1.5–18.0), and then parents (OR = 3.8; 95% CI = 1.2–11.3). In a multivariate model fit to the parents and sibs data, the familial relative risk adjusted for age, sex, relative class, and shared environment was 4.7 (95% CI = 2.3–9.7). White cases had a markedly higher familial relative risk compared with African-American cases (18.0 versus 2.8; p = 0.098). In summary, a significant elevated risk of sarcoidosis was observed among first- and second-degree relatives of sarcoidosis cases compared with relatives of matched control subjects.

Keywords: family; blacks; whites; risk

Sarcoidosis is a multisystem granulomatous disorder of unknown etiology with disease onset generally occurring in young to middle adulthood (between the ages of 20 and 40 yr). Both environmental and hereditary etiologies have been proposed, with the latter mainly supported by reports of familial clustering (1-6) and associations between sarcoidosis and genetic polymorphisms (7-14). Recent findings that link sarcoidosis to the major histocompatibility region on chromosome 6p (15) further support the genetic etiology hypothesis (16, 17).

Despite the low population incidence of sarcoidosis (18), families with two or more affected members are a relatively common occurrence (1, 16, 17, 19-21). For instance, a recently reported survey found a prevalence of familial sarcoidosis of 3.6% and 4.3% in Finnish and Japanese patients, respectively (22). Higher prevalence estimates of familial sarcoidosis have been reported in Irish—9.6% (16), Swedish—6.9% (5), and African-American—19% (1) patient populations. The main limitation of these familial reports is the lack of a comparison group, and therefore it is unclear whether variation in familial sarcoidosis is due to variation in familial aggregation of disease risk, disease prevalence, or both.

Comparisons between disease occurrence in relatives of sarcoidosis cases and a reference population are limited. Headings and coworkers (2) estimated a sarcoidosis prevalence of 1.5% in 523 first-degree relatives of 80 African-American sarcoidosis cases that was appreciably higher than the 0.07% disease prevalence in the New York City comparison population. Buck and McKusick (3) found a 2% prevalence of disease in 125 first-degree relatives of a sample of primarily African-American sarcoidosis cases, but no first-degree relatives of controls with a history of sarcoidosis were discovered, which precluded estimating familial relative risk. Rybicki and coworkers (6) studied 488 parents and sibs of 179 African-American sarcoidosis cases and found a 2.5-fold increased risk of history of sarcoidosis in case relatives over that in the general population. McGrath and coworkers (20) recently reported a sib relative risk for sarcoidosis in a primarily white population of between 36 and 73.

The considerable sample size of case and control relatives in ACCESS (A Case–Control Etiology Study of Sarcoidosis) allowed us to estimate familial relative risk of sarcoidosis in both whites and African-Americans. Our results confirm what has to now be based on mainly anecdotal reports, that sarcoidosis aggregates in families.

Study Population

The study population consisted of 10,862 first-degree and 17,047 second-degree relatives identified by 706 sarcoidosis case–control pairs. Complete data on sibs and parents that could be used in the familial aggregation analyses were availble for 646 of the 706 pairs (91.5%). Sarcoidosis cases were enrolled at 10 clinical centers from November 1996 to June 1999. Cases were enrolled within 6 mo of histological confirmation of disease. One matched control per case was enrolled using random digit dialing sampling techniques. Controls were matched to cases on race, sex, age (± 5 yr), and three-digit phone exchange or zip code for the case. Both cases and controls were administered a written informed consent by the ACCESS study coordinator before any study procedures were performed. Details of the ACCESS study design can be found in an earlier published report (23).

Data Collection

Each case and control was asked to enumerate all his or her first- (parents, full sibs, and children) and second-degree (avuncular and grandparents) relatives. For each first-degree relative enumerated, the case or control was asked the relative's current age (or age at death), sex, race (parents only), number of years lived together, place in the birth order (sibs only), and whether the relative was ever diagnosed with sarcoidosis. For first-degree relatives reported to have a history of sarcoidosis, additional questions about the age at diagnosis of the relative and time lived together with the relative after his or her diagnosis were also asked. Cases and controls who reported a first-degree relative with a history of sarcoidosis were subsequently given a letter to present to the relative or relatives in question asking them to consent to further questioning by study personnel by sending back a self-addressed postcard. Relatives who returned the postcard were subsequently contacted by study personnel and questioned about the history and characteristics of their putative sarcoidosis diagnosis.

Statistical Methods

To account for possible family correlation structures, the method of Liang (24) was used to calculate odds ratios and p values for covariates in all familial aggregation models. This method differs from traditional case–control logistic regression analyses in that the disease status of each case/control relative, rather than case/control status, is the dependent variable in the analysis. In other words, the unit of analysis is the individual relative rather than the case–control pair. The relative risk for familial aggregation is the ratio of the odds that an individual has a sarcoidosis history and is a relative of a case compared with the odds that an individual has no sarcoidosis history and is a relative of a control. The additional advantage of this modeling scheme is that potential disease risk factors of individual family members can be incorporated into the model so that the familial relative risk estimate is adjusted for possible confounding factors that may be differentially distributed between case and control relatives.

Other analyses included calculating odds ratios for effect modifiers by including interaction terms between the variable for being related to a case or control and the putative effect modifier. Differences between cumulative age-specific probability of ever being diagnosed with sarcoidosis in case and control relatives were tested as a log rank statistic for Kaplan–Meier survival curves.

Confirmation of Sarcoidosis in Relatives

As described above, we attempted to confirm reports of sarcoidosis in first-degree relatives of cases and controls by personal interview. We were able to interview reported sarcoidosis cases of 21 of 51 first-degree relatives of cases and three of 13 first-degree relatives of controls. In these interviewed first-degree relatives, we found that disease history was confirmed by self-report in 20 of 21 relatives of cases and all three relatives of controls. Because a majority of reported affected relatives and all reported unaffected relatives were not contacted, all analyses were based on sarcoidosis history as reported by the case or control.

Crude Sarcoidosis Familial Relative Risk Estimates in Different Family Members

Table 1 summarizes the reported sarcoidosis occurrence in first- and second-degree relatives of cases and controls. The odds ratio of a relative with a history of sarcoidosis being related to a case was approximately four or greater and significant in each class of first- and second-degree relatives, with the exception of children (odds ratio [OR] = 3.3; 95% confidence interval [CI] = 0.3–32.2). The highest OR was observed for sibs with a history of sarcoidosis being related to a case (OR = 5.8; 95% CI = 2.1–15.9). The occurrence of sarcoidosis was 1% in both first- and second-degree relatives of cases with a combined 139 first- and second-degree relatives with a reported history of sarcoidosis among the 14,345 relatives enumerated. The OR for a second-degree relative with a history of sarcoidosis being related to a case was higher (OR = 5.2; 95% CI = 1.5–18.2) than that for first-degree relatives (OR = 3.8; 95% CI = 1.9–7.6). Cases were also three times more likely to report some other blood relative previously affected with sarcoidosis compared with controls (data not shown). In terms of relatives or close contacts with no blood relation to the index case or control, no positive association was observed between sarcoidosis and case status. Spouses of cases were actually less likely to have a history of sarcoidosis compared with spouses of controls (OR = 0.2; 95% CI = 0.04–1.1), but this association did not reach significance.

Table 1.  SUMMARY OF SARCOIDOSIS FAMILIAL ASSOCIATIONS IN 706 ACCESS CASE–CONTROL PAIRS

Relative Type CasesControlsOdds Ratio (95% Confidence Interval)p Value
NNumber Affected (%)NNumber Affected (%)
Parents 1,468 18 (1.2) 1,396 4 (0.3)3.8 (1.2–11.3)0.019
Sibs 2,722 28 (1.0) 2,587 6 (0.2)5.8 (2.1–15.9)0.0007
Children 1,335  5 (0.4) 1,354 3 (0.2)3.3 (0.3–32.2)0.298
All first-degree relatives 5,525 51 (0.9) 5,33713 (0.2)3.8 (1.9–7.6)0.0001
Grandparents 2,936 67 (2.3) 2,79212 (0.4)5.2 (1.5–18.0)0.008
Avuncular 5,884 21 (0.4) 5,435 3 (0.1)5.7 (1.6–20.7)0.008
All second-degree relatives 8,820 88 (1.0) 8,22715 (0.2)5.2 (1.5–18.2)0.009
All first- and  second-degree relatives14,345139 (1.0)13,56428 (0.2)4.6 (2.2–9.6)0.0006
Spouses702  5 (0.7)700 7 (1.0)0.2 (0.04–1.1)0.058

Adjusted Familial Relative Risk Estimates

The multivariate analyses presented in Tables 2 3 4 5 were restricted to the sib and parent subset of reported relatives. Second-degree relatives were omitted because their data were limited to disease status. Children were omitted because they were generally too young to be at risk for sarcoidosis. Table 2 shows different models that were fit in sibs, parents, and a combined data set of both sibs and parents. With the reported history of sarcoidosis of the sib or parent as the dependent variable, the first set of models tested included only a covariate for being related to a case to estimate a crude familial relative risk. The familial relative risk of sarcoidosis was greater for sibs (OR = 5.8; 95% CI = 2.1–15.9) compared with parents (OR = 3.8; 95% CI = 1.2–11.3). In the combined sibs and parents data set, the familial relative risk was 4.7 (95% CI = 2.3–9.7). Next, other putative sarcoidosis risk covariates (i.e., age, birth order in sibship of case/control, sib versus parent, female sex, and time in years lived with case or control) were simultaneously introduced into three separate models for each data set. In these set of “Full” models, compared with the simpler set of Family History models the familial relative risk of having had sarcoidosis increased for both sibs (OR = 5.8 to 7.3) and parents (OR = 3.8 to 4.9), but changed little in the combined sibs and parents data set (OR = 4.7 to 4.5). Among the other risk covariates in the Full models, age was a significant risk factor in the sibs (OR = 4.9; 95% CI = 1.7–14.3) and the sibs and parents data sets (OR = 1.7; 95% CI = 1.1–2.6). Female sex was a significant risk factor in the parents data set (OR = 3.2; 95% CI = 1.3–8.1), but not in the sibs (OR = 1.1; 95% CI = 0.2–6.6) or sibs and parents data sets (OR = 1.7; 95% CI = 0.8–3.6). Shared environment was modeled as a covariate that quantified time spent together before disease diagnosis, but this parameter was not significant in any of the three models fit to the sibs, parents, and sibs and parents data sets. In the sibs data set, later birth order had a protective, but nonsignificant effect on disease risk (OR = 0.8; 95% CI = 0.5–1.3).

Table 2.  SARCOIDOSIS FAMILIAL RELATIVE RISK MODELS IN 5,309 SIBS AND 2,864 PARENTS OF 646 ACCESS CASE–CONTROL PAIRS

Model and Associated Covariates Odds Ratio (95% Confidence Interval)
SibsParentsSibs and Parents
Family history model
 Case relative* 5.8 (2.1–15.9)3.8 (1.2–11.3)4.7 (2.3–9.7)
Full model
 Case relative* 7.3 (2.6–20.2)4.9 (1.1–21.5)4.5 (2.1–9.6)
 Age 4.9 (1.7–14.3)1.2 (0.7–2.0)1.7 (1.1–2.6)
 Birth order§ 0.8 (0.5–1.3)
 Sibling2.4 (0.5–11.7)
 Female sex1.1 (0.2–6.6)3.2 (1.3–8.1)1.7 (0.8–3.6)
 Time together 1.0 (0.4–2.9)0.4 (0.8–2.2)1.2 (0.6–2.5)

*Covariate modeling risk associated with being a relative of a sarcoidosis case compared with being a relative of a matched control.

Birth order covariate was fit in sibs data set; sibling covariate was fit in sibs and parents data set.

Covariate modeling incremental risk associated with 1 yr of age.

§Covariate modeling risk associated with decreasing birth order (i.e., from oldest to youngest); OR < 1 → older sibs at increased risk; OR > 1 → younger sibs at increased risk.

  Covariate modeling incremental risk associated with 1 yr of living in the same household with the index case before the year of diagnosis.

Table 3.  THE EFFECT OF CASE CHARACTERISTICS ON FAMILIAL RELATIVE RISK ESTIMATES IN 5,309 SIBS  AND 2,864 PARENTS OF 646 ACCESS CASE–CONTROL PAIRS

Case CharacteristicGroup AGroup BInteraction Coefficient* Standard Errorp Value
RaceAfrican-AmericanWhite−1.861.120.098
SexFemaleMale−0.690.910.444
Age40+< 40−0.750.780.334
CXR stageII-IV0–I−0.400.770.604
Extrathoracic diseasePresentAbsent0.120.850.890
Pulmonary functionBelow medianAbove median−0.330.830.691

*  If the interaction coefficient ≈ 0, then there was no difference in the familial relative risk between Groups A and B; if the interaction coefficient was strongly positive, then the familial relative risk estimate was greater in Group A compared with B; if the interaction coefficient was strongly negative, then the familial relative risk estimate was greater in Group B compared with A.

Table 4.  FAMILIAL RISK MODELS IN 2,358 SIBS AND 1,598 PARENTS OF 355 WHITE ACCESS CASE–CONTROL PAIRS

Model and Associated Covariates Odds Ratio (95% Confidence Interval)
SibsSibs and Parents
Family history model
 Case relative* 7.9 (1.0–62.5)16.6 (2.2–126.1)
Full model
 Case relative* 20.5 (1.8–231.2)27.7 (3.6–215.3)
 Age 1.0 (0.9–1.3) 1.2 (1.1–1.2)
 Birth order§ 0.2 (0.1–0.4)
 Sibling203.0 (31.2–1,320.9)
 Female sex3.6 (0.6–20.0) 1.3 (0.7–2.5)
 Time together 1.1 (1.0–1.3) 1.1 (1.0–1.2)

*Covariate modeling risk associated with being a relative of a sarcoidosis case compared with being a relative of a matched control.

Birth order covariate was fit in sibs data set; sibling covariate was fit in sibs and parents data set.

  Covariate modeling incremental risk associated with 1 yr of age.

§   Covariate modeling risk associated with decreasing birth order (i.e., from oldest to youngest); OR < 1 → older sibs at increased risk; OR > 1 → younger sibs at increased risk.

   Covariate modeling incremental risk associated with 1 yr of living in the same household with the index case before the year of diagnosis.

Table 5.  FAMILIAL RISK MODELS IN 2,951 SIBS AND 1,266 PARENTS OF 291 AFRICAN-AMERICAN ACCESS CASE–CONTROL PAIRS

Model and Associated Covariates Odds Ratio (95% Confidence Interval)
SibsParentsSibs and Parents
Family history model
 Case relative* 5.1 (1.6–16.4)1.8 (0.5–6.0)3.1 (1.4–7.1)
Full model
 Case relative* 6.1 (1.9–18.9)1.5 (0.3–7.7)2.9 (1.2–6.9)
 Age 1.2 (1.0–1.3)1.0 (0.3–2.9)1.0 (1.0–1.1)
 Birth order§ 0.8 (0.5–1.3)
 Sibling1.8 (0.4–8.5)
 Female sex0.8 (0.1–6.1)4.4 (0.9–21.5)1.9 (0.8–4.4)
 Time together 1.0 (0.9–1.1)0.7 (0.1–3.4)1.0 (1.0–1.1)

*Covariate modeling risk associated with being a relative of a sarcoidosis case compared with being a relative of a matched control.

  Birth order covariate was fit in sibs data set; sibling covariate was fit in sibs and parents data set.

  Covariate modeling incremental risk associated with 1 yr of age.

§Covariate modeling risk associated with decreasing birth order (i.e., from oldest to youngest); OR < 1 → older sibs at increased risk; OR > 1 → younger sibs at increased risk.

  Covariate modeling incremental risk associated with 1 yr of living in the same household with the index case before the year of diagnosis.

Age-specific Cumulative Probability of Sarcoidosis in Sibs and Parents

Figure 1 shows the higher age-specific cumulative probability of being affected with sarcoidosis in sibs of cases compared with the sibs of controls. Beginning around age 20, the two curves begin to separate with an approximate 2-fold difference by age 30 and a 4-fold difference by age 60. A similar plot for parents (Figure 2) shows a less pronounced difference between the case parents and control parents. This is largely due to the lower age-specific probability of sarcoidosis in the case parents (solid line in Figure 2) as compared with the case sibs (solid line in Figure 1). Using the age at diagnosis as the event time for parents and sibs with reported sarcoidosis, and censoring others either at age of death or current age, a Kaplan– Meier analysis of the sarcoidosis probability curves (Figures 1 and 2) showed that the difference between case and control sibs was highly significant (p = 0.0003), but the difference between case and control parents was only marginally significant (p = 0.052).

Heterogeneity of Sarcoidosis Familial Relative Risk

Next, a series of familial relative risk models were fit that added, in each separate model, an interaction term for six separate demographic and clinical characteristics of cases. The interaction coefficients in Table 3 are not risk estimates, but rather statistical tests for heterogeneity of sarcoidosis risk between the relatives of different case–control pairs. The p values in the last column of Table 3 represent the probability that the interaction term is significantly different from zero (i.e., tests the null hypothesis of no heterogeneity between groups). Of the six interactions tested, the interaction term for racial heterogeneity of sarcoidosis had the largest absolute value (−1.86), but it did not reach significance (p = 0.098). None of the other five interaction terms came close to significance.

Given the large interaction term for racial heterogeneity in familial relative risk, and the knowledge that genetic risk factors can differ by race, the models in Table 2 were rerun separately for whites (Table 4) and African-Americans (Table 5). In the white sib data set (Table 4), the familial relative risk estimate was 7.9 (95% CI = 1.0–62.5). Adjustment for other risk covariates further increased the familial relative risk estimate to 20.5. When the parental data were combined with the sib data, the unadjusted familial relative risk estimate was 16.6 (95% CI = 2.2–126.1), and the adjusted familial relative risk estimate was 27.3 (95% CI = 3.6–215.3). A separate model for white parents could not be fit because no control parents had a history of sarcoidosis. Age was a significant independent risk factor in the sibs and parents data set, but not in the sibs only data set. Female sex had an elevated, albeit nonsignificant, relative risk estimate in the sibs data set, but decreased to almost one in the sibs and parents data set. The relative risk estimate for time together (OR = 1.1) was significant (p < 0.05) and of similar magnitude in the sibs and sibs and parents data sets. Aside from the family relative risk estimates, the other notable difference when analyzing the white case–control pairs separately was the stronger protective effect of later birth order (OR = 0.2; 95% CI = 0.1–0.4).

In African-Americans (Table 5), the unadjusted and adjusted familial relative risk estimates in sibs were 5.1 (95% CI = 1.6–16.4) and 6.1 (95% CI = 1.9–18.9), respectively. In the combined sibs and parents data set, these relative risk estimates decreased to 3.1 (95% CI = 1.4–7.1) and 2.9 (95% CI = 1.2–6.9), respectively. The effect of age on familial relative risk was similar in the African-American (Table 5) and white (Table 4) sibs data sets, but contrary to what was observed in whites, the relative risk estimate for age in African-Americans was not elevated in the sibs and parents data set. Another notable difference in the results for African-Americans was the absence of any association between time together or birth order and familial relative risk. For African-Americans, the relative risk estimate for female sex was higher in the sibs and parents data set (OR = 1.9; 95% CI = 0.8–4.4) compared with the sibs data set (OR = 0.8; 95% CI = 0.1–6.1), but the confidence interval for both of these relative risk estimates included one. The relative risk estimate for sibs was also elevated (OR = 1.8; 0.4–8.5) in African-Americans, but was of smaller magnitude than was observed in whites and was not significant.

The association between birth order and sarcoidosis in the white subset (Table 4) and the fact that family size and birth order are closely related compelled us to perform additional analyses that included a variable for family size in an effort to tease apart any birth order effect on sarcoidosis risk. In both the full and race-stratified data sets, family size had a nonsignificant effect on risk of having had disease and the birth order relative risk estimate was unchanged from those reported in Tables 4 and 5 (data not shown). Additional analyses that dichotomized family size and birth order failed to shed more light on the putative relationship between birth order and familial relative risk as the number of familial cases in sibs was too small to calculate precise relative risk estimates.

The results of the present study represent the largest effort to date to quantify the familial aggregation of sarcoidosis. Of all the previous case–control studies of sarcoidosis, only one matched case–control study identified familial cases, but its small sample size (62 case–control pairs) precluded making an accurate estimate of sarcoidosis familial relative risk (3). In our large case–control study, the odds of a first- or second-degree relative with a history of sarcoidosis being related to a case were 4.6 times greater than being related to a control. Among different first- and second-degree relative classes, the familial relative risk of a sarcoidosis history varied little, ranging from 3.3 in children to 5.8 in sibs. The lower familial relative risk for children may reflect an underlying genetic mechanism or changing environment over time, but more likely indicates that most children were too young to develop sarcoidosis.

Because the data on children were not suitable for studying familial relative risk, further multivariate analyses were limited to sibs and parents. The overall familial relative risk in sibs was larger than that in parents (OR = 3.8 in Table 2), suggesting either a recessive mode of inheritance with incomplete penetrance or a shared environmental effect. A surrogate for the latter was a covariate that modeled time spent together before disease onset. The relative risk estimate for this covariate was found to be slightly elevated only in the white subset (see Table 4), but it did achieve significance in both the sibs and sibs and parents data sets that were analyzed for this subgroup. Additional data on exposures to specific environmental risk factors in relatives would be necessary to further examine the reasons for this potential increased risk in sibs.

For African-Americans, the familial relative risk estimate for a sarcoidosis history in sibs and parents was 3.1. This is similar to a recently reported family study that found a 2.5-fold elevated disease risk in sibs and parents of African-American sarcoidosis cases compared with the African-American general population (6). The main difference between the two studies is the risk differential between sibs and parents. In the present study, familial relative risk for African-American sibs was 2- to 3-fold greater than for parents, depending on the modeling approach, whereas the aforementioned family study found that parents had slightly increased relative risk compared with sibs (RR = 2.8 versus 2.2). This difference in relative risk estimates for parents and sibs between the two studies may largely be due to different study designs (case–control versus family study) rather than any etiologic heterogeneity between the two populations. The family study design generally ensures more reliable data among relatives, whereas the case–control design used in ACCESS relied heavily on the reporting accuracy of the case or control. Despite these differences in familial relative risk estimates for sibs and parents, the significant familial relative risks that both studies found, which were of similar magnitude, support earlier reports (2, 3) that familial factors contribute to sarcoidosis susceptibility in African-Americans.

The results for whites in the ACCESS study can be compared with only one previous familial relative risk study recently reported (20). In a sample of sarcoidosis family members primarily of the white race residing in the United Kingdom, McGrath and coworkers (20) found a sib relative risk between 36 and 73, significantly higher than the adjusted familial relative risk of 20.5 (Table 4) we found in the ACCESS white population. Several important differences exist between the two studies, however, that may explain these differences. In the United Kingdom study, the sib risk estimate was not adjusted for race, sex, or age, potentially important cofounders with respect to sarcoidosis risk. Of greater importance, however, is the large potential for bias in comparing the prevalence of familial sarcoidosis with population prevalence estimates. This is because familial sarcoidosis generally is defined as the percent of family members who ever had a diagnosis of sarcoidosis, whereas sarcoidosis prevalence is more likely representative of the percent of the population who currently have active disease. This bias could be especially large for whites who generally have a short self-limiting disease course (25).

Apart from the significant sarcoidosis familial relative risk estimates in the entire ACCESS study sample and the two racial subsets, perhaps our most intriguing result was the large difference in familial relative risk between African-Americans and whites. The familial relative risk estimate for white sibs and parents was 16.6 compared with 3.1 for African-Americans and for sibs alone, race-specific estimates were 203 for whites versus 1.8 for African-Americans. However, the white parameter had a very wide confidence interval, and therefore it is unclear whether white sibs actually have two orders of magnitude greater risk than their African-American counterparts. Although the racial differences we observed in familial relative risk were not significant, they suggest that shared familial factors may have a greater contribution to sarcoidosis susceptibility in whites than in African-Americans.

Even in this large study of sarcoidosis, the frequency of familial events was low. Although statistical models were fit to the data and odds ratio estimates were obtained, the high ratio of number of parameters to number of events led to risk estimates that have large standard errors. The complexity of the models that were fit to the data was also limited. For instance, our interaction tests were limited to the addition of a single term in the model, and the number of terms included in a model was generally held to less than six variables.

One potential major limitation of our study is differential recall of affected family members between cases and controls. The observed number of sarcoidosis cases in parents and sibs of controls was much lower than the expected number based on cumulative sarcoidosis incidence estimates from a recent study in whites and African-Americans (18). In cases, the expected and observed numbers of reported affected parents and sibs were similar. This would suggest recall bias did play a role in our results, but the significant variation of sarcoidosis incidence in different populations also makes it precarious to use one population as a standard for expected sarcoidosis rates. Nevertheless, this empirical evidence for potential recall bias should be taken into account when interpreting the familial risk estimates we report in this study. It is also possible that detection bias may have caused an artificial difference in the reported frequency of case/control relatives affected with sarcoidosis. A 50-yr incidence study in Rochester, Minnesota showed that individuals with greater access to health care had a higher incidence of sarcoidosis (26). Close family members of sarcoidosis cases may be under greater diagnostic scrutiny than those without a family member affected with sarcoidosis resulting in a case detection differential. Unfortunately, we have no way of testing for this bias in the ACCESS study.

Despite the limitations of the case–control design for measuring familial relative risk, the ACCESS study had a number of strengths that add credence to our results. These include a large sample size (706 case–control pairs), a standardized protocol used by all study sites, and an incident case group that had an extensive clinical characterization. These aspects provided a large amount of consistently collected data on a well-defined study population. The ACCESS study also restricted enrolled cases to those with positive histology within 6 mo of enrollment, which decreased case misclassification, and the probability of oversampling cases with chronic sarcoidosis. By randomly sampling controls from the general population, our reported familial relative risk estimates should reflect true population differences. All three control relatives and 20 of 21 case relatives with a reported history of sarcoidosis that we were able to interview confirmed that they were diagnosed with sarcoidosis. This would suggest that the false positive percentage of sarcoidosis history as reported by the case or control was low and had a nominal effect on our results.

In summary, in the ACCESS study, cases were almost five times more likely than controls to report a sib or parent with a history of sarcoidosis. It should be stressed that despite this significant increased risk for first- and second-degree relatives of sarcoidosis cases, both the absolute risk of sarcoidosis in these family members and the attributable risk for a sib or parent of a sarcoidosis case were approximately 1%. In other words, our findings suggest that increased surveillance of relatives of sarcoidosis cases is probably not warranted given the small percentage that will eventually develop disease. Nevertheless, familial factors should not be discounted in the search for etiologic factors for sarcoidosis, particularly with respect to race. For instance, the disproportionately higher familial relative risk we found in whites compared with African-Americans suggests genetic factors may exert a more detectable effect in white families with multiple affected members. The elevated familial relative risk estimates we found in both races support the separate genetic linkage studies currently underway in whites in Europe and African-Americans in the United States. Our race-specific familial relative risk estimates should be useful in determining the proportion of familial risk explained by any putative sarcoidosis susceptibility gene(s) that are identified by these or future genetic studies of sarcoidosis.

Clinical Centers

Beth Israel Deaconess Medical Center: Steven E. Weinberger,  M.D.; Patricia Finn, M.D.; Erik Garpestad, M.D.; Allison  Moran, R.N.

Georgetown University Medical Center: Henry Yeager, Jr.,  M.D.; David L. Rabin, M.D.; Susan Stein, M.A.

Case Western Reserve University–Henry Ford Health Sciences  Center: Michael C. Iannuzzi, M.D.; Benjamin Rybicki, Ph.D.;  Marcie Major, R.N.; Mary Maliarik, Ph.D.; John Popovich,  Jr., M.D.

Johns Hopkins University School of Medicine: David R. Moller,  M.D.; Carol J. Johns, M.D.; Cynthia Rand, Ph.D.; Joanne  Steimel, R.N.

Medical University of South Carolina: Marc A. Judson, M.D.;  Susan D'Alessandro, R.N.; Nancy Heister, R.N.; Theresa  Johnson, R.N.; Daniel T. Lackland, Dr.P.H.; Janardan Pandey,  Ph.D.; Steven Sahn, M.D.; Charlie Strange, M.D.

Mount Sinai Medical Center: Alvin S. Teirstein, M.D.; Louis DePalo, M.D.; Sheldon Brown, M.D.; Marvin Lesser, M.D.;  Maria L. Padilla, M.D.; Marilyn Marshall.

National Jewish Medical and Research Center: Lee S. Newman,  M.D., M.A.; Cecile Rose, M.D., M.P.H.; Juli Barnard, M.A.

University of Cincinnati Medical Center: Robert P. Baughman,  M.D.; Elyse E. Lower, M.D.; Donna B. Winget.

University of Iowa College of Medicine: Geoffrey McLennan,  M.D., Ph.D.; Gary Hunninghake, M.D.; Chuck Dayton, B.S.  Pharm.; Linda Powers, M.S.

University of Pennsylvania and MCP–Hahnemann University  Medical Centers: Milton D. Rossman, M.D.; Eddy A. Bresnitz,  M.D.; Ronald Daniele, M.D.; Jackie Regovich, M.P.H.;  William Sexauer, M.D.

National Heart, Lung, and Blood Institute

National Heart, Lung, and Blood Institute: Robert Musson,  Ph.D.; Joanne Deshler; Paul Sorlie, Ph.D.; Margaret Wu, Ph.D.

Study Chairman

 Reuben Cherniack, M.D.

Study Co-Chairman

 Lee Newman, M.D.

Clinical Coordinating Center

Clinical Trials & Surveys Corp.: Genell L. Knatterud,  Ph.D.;  Michael L. Terrin, M.D.; Bruce W. Thompson,  Ph.D.;  Kathleen Brown, Ph.D.; Margaret Frederick,  Ph.D.;  Frances LoPresti, M.S.; Patricia Wilkins, B.S.;  Martha Canner,  M.S.; Judy Dotson.

Central Repository

McKesson Bioservices (September, 1996 to November, 1998):  Steve Lindenfelser

BBI-Biotech Research Laboratories (December, 1988 to  present): Mark Cosentino, Ph.D.

Central Laboratories

DNA Core Laboratory: Mary Maliarik, Ph.D.

BAL Central Laboratory: Robert Baughman, M.D.

HLA Class II Typing Laboratory: Milton Rossman, M.D.;  Dimitri Monos, Ph.D.; Chung Wha Lee, Ph.D.

Etiologic Antigen in Kveim Reagent Laboratory: David Moller,  M.D.

Immunogenetics Laboratory: Janardan Pandey, Ph.D.

L-Forms Core Laboratory: Peter Almenoff, M.D.; Ian Brett;  Sheldon Brown, M.D.; Marvin Lesser, M.D.

Pathogenic T Cells Laboratory: Lee Newman, M.D.; Brian  Kotzin, M.D.

Ribosomal DNA Core Laboratory: Geoffrey McLennan,  M.D., Ph.D.; Gary Hunninghake, M.D.

RNA Core Laboratory: Patricia Finn, M.D.

Random Digit Dialing Interview Group

Telesurveys Research Associates: Richard D. Jaffe, M.A.

Executive Committee

 Reuben Cherniack, M.D. (Chair)

 Robert P. Baughman, M.D. (9/1/98–8/31/99)

 Joanne Deshler

 Michael C. Iannuzzi, M.D. (9/1/96–3/31/97; 9/1/00–6/30/01)

 Marc A. Judson, M.D. (9/1/96–8/31/97; 9/1/00–6/30/01)

 Genell L. Knatterud, Ph.D.

 Geoffrey McLennan, M.D. (9/1/97–8/31/98)

 David R. Moller, M.D. (9/1/95–3/31/96; 9/1/99–8/31/00)

 Robert A. Musson, Ph.D.

 Lee S. Newman, M.D.

 Milton D. Rossman, M.D. (8/1/95–3/31/86; 9/1/99–8/31/00)

 Alvin S. Teirstein, M.D. (9/1/97–8/31/98)

 Michael L. Terrin, M.D., M.P.H.

 Steven E. Weinberger, M.D. (9/1/97–8/31/98)

 Henry Yeager, Jr., M.D. (9/1/98–8/31/99)

Data Safety and Monitoring Board

 William Martin, M.D. (Chair)

 Takamaru Ashikaga, Ph.D.

 David B. Coultas, M.D.

 Gerald S. Davis, M.D.

 Fred Gifford, Ph.D.

 James J. Schlesselman, Ph.D.

 Diane Stover, M.D.

Ex Officio:

 Reuben Cherniack, M.D.; Genell L. Knatterud, Ph.D.;    Robert Musson, Ph.D.; Lee Newman, M.D.

Supported by contracts (N01-HR-56065, 56066, 56067, 56068, 56069, 56070, 56071, 56072, 56073, 56074 and 56075) with the National Heart, Lung, and Blood Institute.

1. Harrington DW, Major M, Rybicki B, Popovich J, Maliarik M, Iannuzzi MCFamilial analysis of 91 families. Sarcoidosis111994240243
2. Headings VE, Weston D, Young RCJ, Hackney RLJFamilial sarcoidosis with multiple occurrences in eleven families: a possible mechanism of inheritance. Ann NY Acad Sci2781976377385
3. Buck AA, McKusick VAEpidemiologic investigations of sarcoidosis. III. Serum proteins; syphilis; association with tuberculosis: familial aggregation. Am J Hyg741961174188
4. Moura M, Carre P, Larios-Ramos L, Didier A, Leophonte PSarcoidosis and heredity. 3 familial cases. Rev Pneumol Clin4619902830
5. Wiman LGFamilial occurrence of sarcoidosis. Scand J Respir Dis Suppl801972115119
6. Rybicki BA, Kirkey KL, Major M, Maliarik MJ, Popovich J, Chase GA, Iannuzzi MCThe familial risk-ratio of sarcoidosis in African-American sibs and parents. Am J Epidemiol1532001188193
7. Arbustini E, Grasso M, Leo G, Tinelli C, Fasani R, Diegoli M, Banchieri N, Cipriani A, Gorrini M, Semenzato G, et al.. Polymorphism of angiotensin-converting enzyme gene in sarcoidosis. Am J Respir Crit Care Med1531996851854
8. Rybicki BA, Maliarik MJ, Malvitz E, Sheffer RG, Major M, Popovich J, Iannuzzi MCThe influence of T cell receptor and cytokine genes on sarcoidosis susceptibility in African Americans. Hum Immunol601999867874
9. Ishihara M, Ohno S, Ishida T, Naruse T, Kagiya M, Mizuki N, Maruya E, Saji H, Inoko HAnalysis of allelic variation of the TAP2 gene in sarcoidosis. Tissue Antigens491997107110
10. Lympany PA, Petrek M, Southcott AM, Newman Taylor AJ, Welsh KI, du Bois RM. HLA-DPB polymorphisms: Glu 69 association with sarcoidosis. Eur J Immunogenet 1996;23:353–359.
11. Maliarik MJ, Rybicki BA, Malvitz E, Sheffer RG, Major M, Popovich J, Iannuzzi MCAngiotensin-converting enzyme gene polymorphism and risk of sarcoidosis. Am J Respir Crit Care Med158199815661570
12. Maliarik MJ, Chen KM, Major ML, Sheffer RG, Popovich J, Rybicki BA, Iannuzzi MCAnalysis of HLA-DPB1 polymorphisms in African-Americans with sarcoidosis. Am J Respir Crit Care Med1581998111114
13. Martinetti M, Tinelli C, Kolek V, Cuccia M, Salvaneschi L, Pasturenzi L, Semenzato G, Cipriani A, Bartova A, Luisetti M“The sarcoidosis map”: a joint survey of clinical and immunogenetic findings in two European countries. Am J Respir Crit Care Med1521995557564
14. Pasturenzi L, Martinetti M, Cuccia M, Cipriani A, Semenzato G, Luisetti MHLA class I, II, and III polymorphism in Italian patients with sarcoidosis. The Pavia-Padova Sarcoidosis Study Group. Chest104199311701175
15. Schurmann M, Lympany PA, Reichel P, Muller-Myhsok B, Wurm K, Schlaak M, Muller-Quernheim J, du Bois RM, Schwinger EFamilial sarcoidosis is linked to the major histocompatibility complex region. Am J Respir Crit Care Med1622000861864
16. Brennan NJ, Crean P, Long JP, Fitzgerald MXHigh prevalence of familial sarcoidosis in an Irish population. Thorax3919841418
17. Rybicki BA, Harrington D, Major M, Simoff M, Popovich J, Maliarik M, Iannuzzi MCHeterogeneity of familial risk in sarcoidosis. Genet Epidemiol1319962333
18. Rybicki BA, Major M, Popovich J, Maliarik MJ, Iannuzzi MCRacial differences in sarcoidosis incidence: a 5-year study in a health maintenance organization. Am J Epidemiol1451997234241
19. Henke CE, Henke G, Elveback LR, Beard CM, Ballard DJ, Kurland LTThe epidemiology of sarcoidosis in Rochester, Minnesota: a population-based study of incidence and survival. Am J Epidemiol1231986840845
20. McGrath DS, Daniil Z, Foley P, du Bios JL, Lympany PA, Cullinan P, du Bios RMEpidemiology of familial sarcoidosis in the UK. Thorax552000751754
21. Parkes SA, Baker SB, Bourdillon RE, Murray CR, Rakshit M, Sarkies JW, Travers JP, Williams EWIncidence of sarcoidosis in the Isle of Man. Thorax401985284287
22. Pietinalho A, Ohmichi M, Hirasawa M, Hiraga Y, Lofroos AB, Selroos OFamilial sarcoidosis in Finland and Hokkaido, Japan—a comparative study. Respir Med931999408412
23. ACCESS Research Group. Design of a case control etiologic study of sarcoidosis (ACCESS). ACCESS Research Group. J Clin Epidemiol 1999;52:1173–1186.
24. Liang KYExtended Mantel-Haenszel estimating procedure for multivariate logistic regression models. Biometrics431987289299
25. Edmondstone WM, Wilson AGSarcoidosis in Caucasians, Blacks and Asians in London. Br J Dis Chest7919852736
26. Hennessy TW, Ballard DJ, DeRemee RA, Chu CP, Melton LJThe influence of diagnostic access bias on the epidemiology of sarcoidosis: a population-based study in Rochester, Minnesota, 1935–1984. J Clin Epidemiol411988565570
Correspondence and requests for reprints should be addressed to Benjamin A. Rybicki, Ph.D., Department of Biostatistics and Research Epidemiology, Henry Ford Health System, 1 Ford Place, 3E, Detroit, MI 48202. E-mail:

†  Deceased.

Related

No related items
American Journal of Respiratory and Critical Care Medicine
164
11

Click to see any corrections or updates and to confirm this is the authentic version of record