To estimate the prevalences of polyarteritis nodosa (PAN), microscopic polyangiitis (MPA), Wegener's granulomatosis (WG), and Churg-Strauss syndrome (CSS).
To estimate the prevalences of polyarteritis nodosa (PAN), microscopic polyangiitis (MPA), Wegener's granulomatosis (WG), and Churg-Strauss syndrome (CSS).
Cases were collected in Seine–St. Denis County, a northeastern suburb of Paris, which has 1,093,515 adults (≥15 years), 28% of whom are of non-European ancestry. The study period encompassed the entire calendar year 2000. Cases were identified by general practitioners, the departments of all the public hospitals and 2 large private clinics, and the National Health Insurance System. The Chapel Hill nomenclature was used to define MPA, and American College of Rheumatology criteria to define WG and CSS; PAN was diagnosed based on clinical laboratory, histological and/or angiographic findings. Three-source capture–recapture analysis was performed to correct for incomplete case ascertainment.
A total of 75 cases were retained and capture–recapture analysis estimated that 23.8 cases had been missed by any 1 of the 3 sources. Accordingly, prevalences per 1,000,000 adults (95% confidence interval [95% CI]) were estimated to be 30.7 (95% CI 21–40) for PAN, 25.1 (95% CI 16–34) for MPA, 23.7 (95% CI 16–31) for WG, and 10.7 (95% CI 5–17) for CSS. The overall prevalence was 2.0 times higher for subjects of European ancestry than for non-Europeans (P = 0.01).
This study provides the first prevalence estimates for these 4 vasculitides for a multiethnic, urban population. The significantly higher prevalence observed for Europeans may infer a genetic susceptibility of Caucasians. Compared with previous estimates based mostly on rural populations, the higher frequency of PAN and the lower frequency of WG might suggest specific environmental etiologic factors.
Polyarteritis nodosa (PAN), microscopic polyangiitis (MPA), Wegener's granulomatosis (WG), and Churg-Strauss syndrome (CSS) are primary small- and medium-vessel vasculitides. These 4 vasculitides are commonly considered a single entity because of clinical and pathologic similarities and a strong association of antineutrophil cytoplasmic antibodies (ANCA) with WG, MPA, and CSS (1, 2). All 4 entities affect predominantly middle-aged adults and require long-term followup because of the risk of relapse (1–3).
So far, the epidemiologic data for PAN, MPA, WG, and CSS have been scarce. However, since the publication of the 1990 American College of Rheumatology (ACR) classification criteria (4) and the 1992 Chapel Hill Consensus Conference (CHCC) nomenclature of vasculitides (3), an increasing number of epidemiologic studies has focused on this topic and provided incidence or prevalence estimates for these diseases (5–12). The results of those studies suggested latitudinal differences, with WG being more frequent in northern countries (8, 13) and MPA prevailing in the south (13). Ethnic variations and differences between rural and urban areas might also play a role (12, 14).
The previous epidemiologic investigations were carried out in predominantly or exclusively rural populations (5–12). Moreover, no population-based estimate has yet been established for nonwhite populations. We therefore conducted a population-based survey with the aim to estimate the prevalences of PAN, MPA, WG, and CSS in an urban and multiethnic French county.
The study was carried out in Seine–St. Denis County, a northeastern suburb of Paris, France. It covers an area of 236 km2 and is part of the highly urbanized Parisian agglomeration (Ile-de-France region; Figure 1). From an economic point of view, employment in the Ile-de-France region is mainly in the tertiary sectors (services, commerce, transportation), with industry and construction accounting for 81% and 18% of the employment figures, respectively (15).
According to the national census conducted in 1999, the total population of the Seine–St. Denis County was 1,382,928 inhabitants, including 1,093,515 adults (≥15 years). The male-to-female ratio for the adult population was 0.94. Among these adults, 301,102 (28%) were of non-European ancestry. These subjects came from the Maghreb (Morocco, Algeria, Tunisia; 40%), the French overseas counties and territories (19%), and Turkey (5%); the remaining 36% predominantly include descendants from Asian countries and sub-Saharan Africa. The methods used to calculate the size of the non-European population are detailed in Appendix A.
Compared with the previous census conducted in 1990, the total size of our adult study population remained stable (+0.1%). The annual immigration-emigration rate over this period was 2.4% of the adult population. The highest annual fluctuation was observed for the group of individuals 25–39 years old and non-French citizens, with 4.1% and 3.8% who had immigrated or emigrated, respectively, each year. Compared with the French population, the age distribution of the adult study population is characterized by a higher percentage of subjects 25–59 years old (63% versus 59%) and a lower percentage of subjects ≥60 years old (19% versus 26%). This difference is even more prominent for the non-European population, with 74% of the individuals 25–59 years old and only 12% ≥60 years (Figure 2). All these data were provided by the Institut National des Statistiques et des Études Économiques (National Institute for Statistics and Economic Studies) (16).
The study encompassed the period from January 1 to December 31, 2000. The cases were ascertained through 3 separate sources: 1) letters were sent (with a followup letter in case of non-response) to all the 1,119 general practitioners (GPs) registered in the College for Physicians of the area studied; 2) letters were sent (with a followup telephone call in case of non-response) to the 20 departments of internal medicine, rheumatology, nephrology, pneumology, and hemodialysis of all the public hospitals and 2 large private clinics of the study area; and 3) the computerized database of the National Health Insurance System (Caisse Primaire d'Assurance-Maladie) of Seine–St. Denis County was searched. The French National Health Insurance System provides free access to primary and secondary medical care for individuals with chronic or costly diseases, including the 4 vasculitides considered. Therefore, the National Health Insurance System database contains regularly updated information obtained from applications to exonerate such patients from medical expenses. The GPs and hospital physicians were asked to inform us of any known case of PAN, MP, WG, or CSS living in the study area during the year 2000. For the National Health Insurance System source, we recruited all subjects who had been recorded on December 31, 2000, with one of the following codes of the International Classification of Diseases (ICD): M 30 (PAN group), M 30.0 (PAN), M 30.1 (CSS), M 30.8 (other conditions related to PAN), M 31.3 (WG). Each case was recorded with the first 3 initials of the patient's last name, the first 2 initials of the first name, sex, birth date, residential postal code, and, for the GPs, the date of the patient's last assessment and the address of the hospital where he or she was examined. To obtain the addresses of the hospitals where the patients recruited through the National Health Insurance System source were examined, we phoned the patients after written consent had been obtained in response to a letter from the National Health Insurance System.
All the diagnoses were confirmed by review of the hospital charts. Medical charts were reviewed by the same investigator (AM) with respect to the ACR criteria for WG (17) and CSS (18) and the CHCC definition of MPA (3). PAN was defined as a vasculitis of predominantly medium-sized arteries that may be associated with hepatitis B virus (HBV) infection; angiographically documented microaneurysms were considered highly suggestive of PAN, whereas features of glomerulonephritis, lung hemorrhage, or ANCA positivity excluded the diagnosis of PAN (19). We also required histologic proof of vasculitis. However, histologic documentation of glomerulonephritis (for MPA, WG, and CSS) or extravascular granulomata (for WG and CSS) were considered equal to documented vessel inflammation when associated with ANCA positivity. For PAN, angiographic evidence of microaneurysms was also considered a surrogate for histologic examination. Limb-restricted or other localized forms of PAN were not retained. None of the patients could be classified in more than 1 vasculitis category and, for borderline cases, a consensus was sought among the investigators. For patients who had been diagnosed prior to the routine use of ANCA enzyme-linked assays, ANCA were considered to be present based on the sole result of a positive indirect immunofluorescence test.
Statistical analyses were computed using the SAS Statistical Package, version 8.12 (SAS Institute, Cary, NC). Quantitative variables are expressed as means ± SD. Qualitative variables were compared using chi-square and, when appropriate, Fisher's exact tests. All statistical analyses were 2-tailed and P values <0.05 were considered significant. Confidence intervals were calculated at the 95% level (95% CI).
Capture–recapture analysis was used to estimate the total number of cases. The capture–recapture technique analyzes 2 or more data sets constructed from different sources of case ascertainment and, based on the overlap among these sets, generates an estimate of the number of cases missing from any one. For appropriate application of capture–recapture analysis, the following criteria have to be satisfied: 1) the study population is closed; 2) the individuals can be matched; 3) the probability of ascertainment does not vary from one source to another according to particular characteristics (equal catchability); and 4) the sources of case ascertainment are independent (source independency) (20, 21).
The matching criteria used in our study were the first 3 initials of the last name, first 2 initials of the first name, date of birth, sex, and the residential postal code. Capture–recapture estimates were established by log-linear modeling (22, 23). Source dependency was examined by comparing the nonsaturated model to models including all the possible combinations of second degree interaction terms between sources. The model providing the best compromise between the least interaction terms and the highest adequacy as measured by the likelihood ratio statistics (deviance G2, Akaike's information criterion [AIC], Bayesian information criterion [BIC]) and the weighted BIC (24) was considered most appropriate. To assess unequal catchability, the representativeness of each source was investigated by comparing the observed and estimated distributions after stratification by parameters potentially influencing the probability of the case being ascertained. Unequal capture was identified when, for a given source, these distributions differed significantly and the estimates were then corrected by introducing the corresponding interaction term into the log-linear model (25). Prevalence rates were calculated with the number of cases as estimated by capture–recapture analysis as the numerator and the adult population as the denominator. The variances of capture–recapture estimates were calculated according to the method described by Hook and Regal (26).
The investigation was approved by the Commission Nationale de l'Informatique et des Libertés (National Commission of Informatics and Freedom; no. 759902).
The response rates to the questionnaires by GPs and the hospital departments were 53% and 100%, respectively. A total of 121 patients were identified through the 3 sources. Among those, we excluded patients with other diagnoses (n = 23), those not fulfilling the diagnostic criteria (n = 19), those living outside the area studied (n = 5), those <15 years old (n = 2), or those diagnosed after the year 2000 (n = 4).
Therefore, we eventually identified 68 patients definitively diagnosed with PAN (n = 23), MPA (n = 16), WG (n = 21), or CSS (n = 8). Histologic confirmation, obtained for 64 patients, was based on vasculitis for 75%, glomerulonephritis for 25%, and/or extravascular granulomata for 22%; for the remaining 4 cases, vasculitis was documented by angiographically detected microaneurysms. For 7 (30%) of the 23 patients with PAN, the vasculitis was HBV related; all these cases were diagnosed prior to 1996, and 5 before 1989. All the patients diagnosed with MPA, WG, and CSS who had no histologically documented vessel inflammation tested positive for ANCA. Among the 16 patients diagnosed with MPA, 3 (5%) presented with kidney-limited disease associated with pauciimmune glomerulonephritis and a perinuclear or antimyeloperoxidase ANCA labeling pattern. The main demographic, clinical, and immunologic data of these 68 cases are summarized in Table 1.
|Disease||No. of patients||Mean age at onset, years ± SD||Disease duration years ± SD||Sex ratio, M/F||ANCA positivity %||HBV infection, %||Non-European ancestry, %|
|PAN||23||47.7 ± 14.7||8.2 ± 6.1†||15/8||0||30||4|
|MPA||16||60.2 ± 14.3||6.0 ± 5.8||5/11||62||0||25|
|WG||21||50.6 ± 17.9||7.2 ± 6.6||12/9||85||0||14|
|CSS||8||43.5 ± 16.8||9.8 ± 8.6||5/3||71||0||38|
|Total||68||50.3 ± 17.6||7.5 ± 6.5||37/31||44||10||16|
The medical files of 16 additional registered cases could not be examined. These patients had been recruited solely by the National Health Insurance System database and had failed to reply or consent to our request for data retrieval from their hospital charts. These patients had been reported to the National Health Insurance System with the ICD codes M 30 and 30.0 (n = 15) and M 31.3 (n = 1). For these patients, we applied a positive diagnostic value of 0.46 (0.13 for PAN, 0.33 for MPA), which had been obtained for the other patients retrieved only from the National Health Insurance System source with the same ICD codes and whose medical files could be reviewed. Consequently, we estimated that the 15 cases recorded with the ICD codes M 30 and 30.0 corresponded to 7 true cases, including 2 cases of PAN and 5 cases of MPA. According to the same algorithm, we considered that the single case with the ICD code 31.3 was not a true case of CSS. Finally, we retained a total of 75 cases (PAN, n = 25; MPA, n = 21; WG, n = 21; CSS, n = 8).
Among the 75 cases retained, 43 (57%) were derived from a single source, 30 (40%) from 2 sources, and 2 (3%) from 3 sources. For most cases, complete concordance of the 5 matching criteria was found; a few cases had 1 discordant criterion concerning the initials of the first name, the day or month of birth, or the postal code. Figure 3 shows the distribution of cases and overlap among them as reported by the different sources.
The estimates established by log-linear modeling are summarized in Table 2. The nonsaturated model had the best fit. This model had a low G2 value (G2 = 5.56; P = 0.14) and negative AIC (–0.44) and BIC (–8.75) values; furthermore, the estimated number of cases missed (n = 23.8) was close to the weighted BIC estimate (n = 22.4).
|Log linear model||df||G2||P||AIC||BIC||x||N||95% CI|
|GP × HD, GP × NHIS, HD × NHIS||0||0||1.0||0||0||7.9||82.9||65–101|
|GP × NHIS, HD × NHIS||1||0.11||0.74||−1.89||−4.66||11.0||86.0||74–98|
|GP × HD, HD × NHIS||1||3.55||0.06||1.55||−1.22||32.7||107.7||59–156|
|GP × HD, GP × NHIS||1||1.18||0.28||−0.82||−3.59||23.7||98.7||74–124|
|GP, HD × NHIS||2||4.83||0.09||0.83||−4.71||18.1||93.1||75–111|
|GP × NHIS, HD||2||2.14||0.34||−1.86||−7.40||18.2||93.2||78–108|
|GP × HD, NHIS||2||3.56||0.17||−0.44||−5.98||33.4||108.4||82–135|
|GP, HD, NHIS||3||5.56||0.14||−0.44||−8.75||23.8||98.8||81–116|
|Weighted BIC estimates||22.4||97.4|
|GP, NHIS, HD × followup site||8||10.70||0.22||−5.3||−27.45||23.8||98.8||67–131|
|Stratification for diagnosis|
|Stratification for geographic origin|
The representativeness of the sources was investigated for the 68 patients whose medical records had been reviewed with respect to the specific diagnoses, year of diagnosis (according to 2 periods defined by the median value), and the geographic origin (European or non-European ancestry). Because we suspected that the cases followed in hospitals outside the study area were unequally captured, we also evaluated this variable (followup inside or outside the county). Indeed, it was found that these cases were significantly less captured by the hospital departments source (P = 0.005; Table 3). Consequently, we included the corresponding interaction term (hospital departments × followup location) in the log-linear model by making the assumption that the 7 additionally retained cases were all followed outside the county. The model obtained was slightly more adequate but estimated the same number of cases missed (Table 2). Therefore, we finally selected the nonsaturated model including no interaction term. Accordingly, the total number of estimated cases was 98.8 (95% CI 81–116) and, after stratification, a total number of cases of 33.6 (95% CI 23–44) for PAN, 27.5 (95% CI 18–37) for MPA, 25.9 (95% CI 18–34) for WG, and 11.7 (95% CI 5–19) for CSS (Table 2). Table 4 summarizes the derived prevalence estimates for PAN, MPA, WG, CSS, and all the vasculitides combined for the adult population of the study area.
|Parameter||Observed distribution, %||Expected distribution, %|
|GP n = 26||HD n = 51||NHIS n = 25|
|Year of diagnosis|
|1995 or before||58||45||40||47|
|1996 or later||42||55||60||53|
|Within the county||65||94†||68||75|
|Outside the county||35||6†||32||25|
|Prevalence, per 1,000,000 adults||95% CI|
Among the 68 cases ascertained, 11 were of non-European ancestry (Maghreb, n = 4; Turkey and Asian countries, n = 3; sub-Saharan Africa, n = 2; French Antilles, n = 2). With respect to the geographic origin of the cases recorded by the National Health Insurance System source, we hypothesized that 1 additional case of non-European ancestry figured among the 7 cases whose medical files had not been reviewed. The stratified capture–recapture analysis estimated that a total of 83.0 (95% CI 67–99) patients were of European ancestry and 15.8 (95% CI 9–23) cases were of non-European ancestry. Thus, the overall prevalence estimates for the European population was twice that of the non-European population (Table 4).
Our study provides the first population-based prevalence rates for PAN, MPA, WG, and CSS determined for an exclusively urban and multiethnic area. Accordingly, we estimated the prevalence per 1,000,000 adults to be 90.3 for all 4 vasculitides, and 30.7 for PAN, 25.1 for MPA, 23.7 for WG, and 10.7 for CSS. The overall prevalence was 2.0 times higher for the subjects of European background compared with non-Europeans.
Although the ACR classification criteria and the CHCC definitions are now widely used for epidemiologic purposes, accurate classification of PAN, MPA, WG, and CSS still presents a problem. All previous studies of MPA (7, 9, 11) used the CHCC nomenclature, which is the only one available at present, whereas diagnoses of PAN, WG, and CSS were based either on the ACR criteria (8–11) or CHCC definitions (7). However, the concepts of classification of the ACR criteria and CHCC definitions differ markedly, as previously noted (27). The CHCC definitions are biopsy-dependent and usually are considered more restrictive (28, 29), whereas the ACR criteria, when improperly applied for diagnostic purposes, may also be met by nonvasculitic diseases (30). In agreement with other authors (10), we consequently added thoroughly defined histologic features to the ACR criteria, but no consensus has been reached on this matter.
In the present study, MPA was defined according to the CHCC nomenclature, and WG and CSS according to the ACR criteria. We only retained cases with biopsy-proven vasculitis or histologic proof of glomerulonephritis or extravascular granulomata when associated with ANCA positivity. Accurate definition of PAN was even more difficult. The ACR criteria have poor specificity and do not discriminate PAN from MPA (31). Conversely, the CHCC nomenclature, which defines PAN as an exclusively medium-sized vessel disease, is generally considered too restrictive (29). Therefore, we redefined PAN as a vasculitis affecting predominantly—but not exclusively—medium-sized arteries and also took into account clinical, immunologic, virologic, and angiographic findings. This situation highlights the fact that a clear-cut classification of these diseases has not yet been fully achieved. In the meantime, the heterogeneity of the case definitions applied by the various investigators might be partly responsible for the differing results.
The strategy of case ascertainment also proved to be important. In studies conducted in geographically isolated rural populations, case retrieval has been based purely on data from referral hospitals covering the whole area surveyed (9, 10). However, in our population, this method would have led to incomplete case ascertainment because of the proximity of hospitals to neighboring counties and because patients can choose to be evaluated outside the county. In accordance with other studies (7, 8, 11, 12), we used multiple case-ascertainment sources and the data provided by the GPs and the National Health Insurance System enabled us to also identify patients seen in hospitals outside the study area. The 3 different methods of recording patient data should have assured that no patient subgroup was completely overlooked.
Capture–recapture analysis was performed to correct for potential undercounts. The estimates were established by log-linear modeling to take into account possible violations of the assumption of source independency and equal catchability (22, 23). The model we chose suggested that the 3 sources were independent. As expected, we identified unequal capture of the patients seen in hospitals outside the study area who were reported to us significantly less often by the hospital departments. Because the assumption of equal catchability is verified on the basis of intuitively selected factors, it is possible that some factors not fulfilling this assumption remained undetected. However, emphasis was placed on the contribution, albeit small, of variable catchability to the bias of capture–recapture estimates (32). Capture–recapture analysis estimated that 23.8 cases had been missed by any 1 of the 3 sources, thereby indicating an uncorrected completeness of case ascertainment of 76%.
Among the 4 vasculitides studied, WG is the best documented, as far as its epidemiology is concerned. Our prevalence estimate of 23.7/1,000,000 adult inhabitants is the lowest hitherto reported and would support the hypothesis of a decreasing north–south gradient for WG (13, 14). A 5-year prevalence rate of 26–32/1,000,000 (12) was reported for the United States and prevalence rates of 42 and 58/1,000,000, respectively, were reported for southern and northern areas of Germany (7). A prevalence of 53/1,000,000 was observed for Norway (9). The highest prevalence rates of 63/1,000,000 adults and 95/1,000,000 were found, respectively, in the United Kingdom (11) and northern Norway (8). The latter studies reported the highest incidences being observed for subjects ≥65 years (8, 11). The lower prevalence in our study could therefore be partly explained by the smaller percentage of individuals ≥60 years old in our population compared with the general French population.
Few epidemiologic data are available for MPA and CSS. The estimated MPA prevalence rate per 1,000,000 was 9.0 in a northern area of Germany, whereas no cases were found in the southern part (7). Conversely, Watts et al reported an annual incidence of 8/1,000,000 (11), which would be in agreement with our much higher prevalence estimate of 25.1/1,000,000 adults. As for CSS, our prevalence estimate of 10.7/1,000,000 adults falls within the bounds of previously reported figures, 2 and 13/1,000,000 (7, 9).
Finding PAN to be the most frequent vasculitis in our population is at variance with statements of PAN being less common than WG, MPA, or CSS (14). Our prevalence estimate of 30.7/1,000,000 adults would suggest that classical PAN is not as rare as once thought. In previous studies, the PAN prevalence was estimated to be 2 and 9/1,000,000 according to the CHCC definition (7) and 33/1,000,000 based on the ACR criteria (9). These differences may be largely due to the above-mentioned heterogeneity of the definitions used for PAN. Our criteria were certainly more sensitive than the CHCC definition, but probably more restrictive than the ACR criteria, particularly with regard to the discrimination of PAN from MPA. In a substantial subset of cases, however, PAN was clinically indistinguishable from MPA with classification relying solely on the size or type of primarily involved vessels in a biopsy specimen, thus highlighting that the differentiation of these 2 entities is often a delicate undertaking. A high annual PAN incidence rate of 77/1,000,000 was observed in a small population with high rates of HBV infection (33) but in our study, not more than 30% of the PAN cases appeared to be HBV-related. Most of these patients were diagnosed during the 1980s, thus suggesting that the incidence of HBV-associated PAN is currently decreasing, possibly as a consequence of vaccination campaigns and the improved safety of blood products.
Although ethnic factors have also been widely incriminated in the etiology of primary systemic vasculitides, their role has not yet been confirmed (14). The high number of Maghrebian, sub-Saharan African, Asian, and Caribbean descendants made our study suitable to establish estimates of prevalence in relationship to geographic origin. The percentage of non-Europeans in our study population (28%) was determined based on citizenship and places of birth, which might have caused an underestimation. Notably, we may have failed to identify descendants of families that immigrated several generations ago. It is all the more pertinent that the overall prevalence in individuals of non-European ancestry was half that in Europeans. It is unlikely that these differences could be explained only by underdiagnosis or by the lower percentage of non-Europeans older than 65 years. Our findings would rather suggest that whites are more likely to develop PAN, MPA, WG, or CSS. The low number of cases did not allow further stratification of the capture–recapture estimates for each of the 4 vasculitides. However, we observed, particularly for PAN and WG, low numbers of patients of non-European ancestry (Table 1), thereby suggesting that these ethnic differences could be particularly relevant for these 2 entities.
Because environmental factors have also been incriminated as triggers of vasculitis (11), it was considered worthwhile to compare our exclusively urban population-derived results with respect to potential variations with rural populations. The high frequencies of PAN and MPA and low WG prevalence in our urban population could suggest possible risk factors associated with the environment or lifestyle. However, no final conclusions can be drawn because methodologic differences limit the comparability of the studies. In the future, case-control studies, ideally based on incident cases, should be carried out and might provide further insight into potential environmental or occupational factors triggering the onset of PAN, MPA, WG, or CSS.
The authors would like to thank the following physicians for their help in case identification or diagnostic confirmation: P. Ureno-Torres (Aubervilliers), D. Clerc, D. Malbec (Aulnay-sous-Bois), M. H. André, M. Bentata, M. C. Boissier, P. Cohen, P. Godmer, O. Lortholary, L. Mouthon, J. Ramanoelina, D. Sadoun, D. Valeyre (Bobigny), O. Fain (Bondy), N. Belmatoug, B. Fantin (Clichy), M. Echard, R. Ghnassia, L. Marie (Montfermeil-Le Raincy), X. Belenfant, J. Laederich (Montreuil), Z. Amoura, S. Aractingi, H. Bachelez, F. Bani-Sadr, P. Cacoub, S. Dautheville, G. Hayem, C. Jacquot, F. Mignon, L. Mercadal, E. Palazzo, T. Papo, B. Wechsler (Paris), P. Babinet, F. Lhote (St. Denis), A. M. Piette (Suresnes), L. Moulonguet-Doleris (Villepinte), and all the general practitioners of Seine–St. Denis County. We are also grateful to Dr. M. Carzon and Dr. F. Chinaud (Service Médical, Régime Général, Ile-de-France) for their help in providing the data from the Service Médical de Seine–St. Denis; to A. Martin (Bobigny), M. Baudrimont, J. Mikol, and M. Polivka (Paris) for histologic reevaluations; to Dr. A. Gallay (St. Maurice) for advice on methodology; and to Ms. J. Jacobson for editorial assistance.
The number of adults of non-European ancestry in our study area was determined using the data from the 1999 national census of the population (16). This database provides information on the nationalities and the place of birth of the inhabitants. Three items were used to define subjects of non-European ancestry: 1) individuals with nationalities not belonging to European Community countries (n = 170,244 adults); 2) naturalized French citizens (n = 111,861 adults); and 3) French citizens born in the French overseas counties and territories (n = 39,132 adults). Based on the data provided by the Ministère de l'Emploi et de la Solidarité (Ministry for Employment and Solidarity), 82% of the naturalized French citizens (item 2) were from non-European countries (n = 91,726 adults) (34). Consequently, the total size of the adult non-European population was estimated to be 301,102 adults. The French overseas counties and territories include the following: French Guiana, Guadeloupe, Martinique, Réunion, French Polynesia, New Caledonia and Dependencies, Southern and Arctic Lands, Wallis and Futuna Islands, Saint Pierre, Miquelon, and Mayotte.