Human leukocyte antigen-DRB1*1104 in the chronic iridocyclitis of pauciarticular juvenile rheumatoid arthritis*

Human leukocyte antigen-DRB4*1104 in the chronic iridocyclitis of pauciarticular juvenile rheumatoid arthritis Hector Melin-Aldana, MD, E d w a r d H. Giannini, MSc, DrPFI, J a n a l e e Taylor, MSN, Daniel J. Lovell, MD, MPH, J o s e p h E. Levinson, MD, Murray H. Passo, MD, J o s e p h Ginsberg, MD, Miles J. Burke, MD, a n d David N. Glass, MD From the Divisions of Rheumatology and Ophthalmology, Children's Hospital Medical Center, Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, Ohio To determine whether g e n e t i c markers for chronic iridocyclitis c o u l d be idenlifled, we used both serologic and o l i g o n u c l e o t i d e dot blot techniques to characterize i m m u n o g e n e t i c a l l y 164 children with early-onset pauciarticular juvenile rheumatoid arthritis. Seventy-eight children (47.6%) had chronic iridocyclitis and 86 (52.4%) had not had e v i d e n c e of e y e disease during a mean follow-up period after the onset of arthritis of 15.8 years (minimum of 5.5 years). Control subjects were 218 healthy, unrelated individuals. The analysis was limited to alleles known to be associated with an increased or d e c r e a s e d risk of early-onset pauciarticular juvenile rheumatoid arthritis or of chronic iridocyclitis in this form of juvenile rheumatoid arthritis. Only one split of human l e u k o c y t e antigen (HLA)-DR5, HLA-DRB1*1104, showed a statistically significant association with a risk of chronic iridocyclitis (chi-square value = 7.52; p = 0.036 adjusted; odds ratio 3.45); HLA-DQA1*0501 and HLA-DQB1*0301, both in linkage disequilibrium with HLA-DRB1*1104, also were significantly associated with eye disease. Patients w i t h both the DRB1*1104 and DPBI*0201 genes had a 7.7-fold increased risk for chronic iridocyclitis c o m p a r e d with that for other patients. The presence of HLADRB1*1104 was a b o u t four times as specific, but only a b o u t one third as sensitive, as a n t i n u c l e a r antibodies in identifying patients at risk for eye disease. Although all children with early-onset pauciarticular juvenile rheumatoid arthritis should u n d e r g o periodic slit-lamp examinations, those with the HLA class II g e n e DRBI*1104 are at particularly high risk for eye disease, and we r e c o m m e n d that they be monitored carefully for its evolution. (J PEDIATR1992;121:56-60) Chronic nongranulomatous inflammation of the anterior uveal tract involving the iris and ciliary body (chronic iridocyclitis) is a prominent clinical complication of juvenile rheumatoid arthritis. Chronic iridocyclitis is most prevalent

Supported by grants from the Arthritis Foundation, the National Institutes of Health (Nos. R01 AR39979 and AR21393), the Schmidlapp Foundation, the Children's Hospital Research Foundation, Cincinnati, Ohio, and the Lion'~ Club of Ohio. Submitted for publication Dec. 16, 1991; accepted Feb. 13, 1992. Reprint requests: Edward H. Giannini, MS, DrPH, Division of Rheumatology, PAV-1, Children's Hospital Medical Center, Elland and Bethesda Boulevard, Cincinnati, OH 45229. 9/20/37099

56

among children in whom arthritis develops in fewer than five joints at the time of articular disease onset; this type is referred to as pauciarticular (oligoarticular) JRA. The frequency of iridocyclitis among those with pauciarticular ANA EOPA-JRA HLA JRA SSC

Antinnclear antibody Early-onset pauciarticular JRA Human leukocyte antigen Juvenile rheumatoid arthritis Standard saline citrate

JRA approaches 20%. A subgroup of patients with pauciarticular JRA at particular high risk for iridocyclitis includes young girls whose serum contains antinuclear an-

Volume 121 Number 1

tibody. The prevalence of iridocyclitis among this subtype, referred to as early-onset pauciarticular JRA, is about 70%. This eye disease occurs less frequently among patients with polyarticular onset JRA, and is rare among those with systemic illness. Although eye disease usually develops within the first few years after the appearance of articular symptoms, its onset may be delayed 10 or more years. 1=5 Thus patients with JRA undergo periodic slit-lamp examinations for many years after the onset of joint disease. Approximately 25% of patients with chronic iridocyclitis have prolonged, severe inflammation and vision-threatening complications despite aggressive therapy. 6 Anecdotal data suggest that serious sequelae, including blindness, may be prevented by early detection and treatment in the majority of patients, 1 although definitive, long-term prospective studies have not been done. Recent studies of class I and class II human leukocyte antigen have been aimed at identifying children with the highest risk of eye disease], 8 Subjects with EOPA-JRA who possess HLA-DR5, are about three and one-half times more likely to have eye disease than are other patients with EOPA-JRA. 8 Recent technologic advances, specifically the use of the polymerase chain reaction and gene sequencing methods, have made it possible to subdivide H L A alleles on the basis of their nucleotide sequences. HLA-DR5, for example, has been split into the subtypes H L A - D R B I * l l 0 1 , "1102, "1103, "1104, and "1201. Working with a large sample of patients with EOPA-JRA, we reexamined the role of various class II H L A genes, and especially the splits of HLA-DR5, in susceptibility to iridocyclitis. We used the oligonucleotide probes for alleles that have been shown to be associated with increased or decreased risk of EOPA-JRA or eye disease in patients with EOPA-JRA.

METHODS SubjectS. We studied 164 patients with EOPA-JRA, of whom 78 (47.6%) had chronic iridocyclitis. All were white, the majority with ethnic origins in northern Europe. All met the criteria for pauciarticular J R A as defined by the American College of Rheumatology, 9 and all had onset of disease before 8 years of age. All had undergone periodic slit-lamp examinations at least once a year and, in most cases, every 6 months. Among those with iridocyclitis, the mean age at onset of articular symptoms was 2.1 years; eye disease developed a mean of 2.9 years (median, 1.96 years) after the onset of arthritis. The remaining 86 patients had been followed for a minimum of 5 years (mean, 15.8; range, 5.5 to 42.9 years), during which time they had remained free of eye disease; their mean age at articular disease onset was 2.5 years. Control subjects. The 218 control subjects were healthy adults with no history of rheumatic disease; they were un-

H L A in chronic iridocyclitis o f J R A

57

related to the patients or to each other. All were white, ethnically matched to the patient sample. HLA typing by serologic study. Subjects underwent H L A typing with antisera tested by the eighth, ninth, and tenth histocompatibility workshops. H L A alleles, defined on the basis of D N A sequence, are named in accordance with the World Health Organization Nomenclature Committee for Factors of the H L A System. l~ HLA typing by oligonucleotide dot blot. Genomic DNA was obtained from whole blood cell lysates by either phenol-chloroform or salt extraction methods. Amplification by the polymerase chain reaction employed primers specific for the second exon of HLA-DRB1, -DQA1, -DQB1, and -DPB1 genes. In the case of the HLA-DRB1 genes, only alleles belonging to the HLA-DRw52 group (DR3, 5, w6, w8) were amplified. Negative control samples, substituting water for target genomic DNA, were run with every reaction. For the study of HLA-DRB1 alleles, 11 of the probes described by Fernandez-Vina et al.11 were used: PS-2, PS- 12, PS-21, PS-22, PS-28, PS-29, PS-30, PS-34, PS-36, PS-41, and PS-63. In addition, two probes designed by us (probe sequences available on request) on the basis of the D N A sequence of specific alleles were also used: H M A 4 (DRBI*1101, "1201, "0801, *0802, *0803), and HMA03 (DRB!*0301, *0302). The study of HLA-DQB1 alleles employed some of the specific oligonucleotide probes described by Khalil et a l l 2 (BG1 to BG9) and four additional probes designed by us: HMA62 (DBQl*0601, *0602, "0301, *0302, *0303, "0401, *0402), HMA63 (DQBI*0501, *0502, *0503, *0603, *0604), HMA42 (DQBI*0402), and HMA47 (DQBl*0401). Probes used for the study of HLA-DQA1 alleles were described previously) 3 Typing of HLA-DPB1 genes was done by using the probes and method described by Angelini et al. 14 In selected instances, a second amplification by polymerase chain reaction was performed, this time with primers "PPP2" and "PPP3" as described by FernandezVina et al) 5 The probes were labeled with phosphorus 32 (by using polynucleotide kinase) and added to the hybridization solution at a concentration of 4 X 105 cpm/ml. Briefly, 2/xl of the polymerase chain reaction-amplified material were applied directly onto nylon membranes with a pore size of 0.1/~m (Nytran; Schleicher & Schuell, Inc., Keene, N.H.). As controls, D N A samples from homozygous cell lines with and without the alleles studied were included on each filter. The membranes were then fixed with an ultraviolet cross-linker (Stratagene, La Jolla, Calif.), and prehybridized for 1 hour at 37 ~ C in a solution containing sodium dodecylsulfate reagent, SSC solution, Denhardt solution, and salmon sperm DNA. Hybridization

58

Melin-Aldana et al.

The Journal of Pediatrics July 1992

Table I. F r e q u e n c y of class I1 H L A genes in control subjects and in patients with E O P A - J R A , with a n d without iridocyclitis

DRB1 alleles DR1 DRBI*I101 DRBI*1104 DRBI*I201 DRBI*1301 DRBI*I302 DRB 1"0801 DQA1 alleles DQA1 *0101 DQA 1 *0103 DQA1 "0401 DQA1 "0501 DQBI alleles DQBI*0301 DQB 1"0603 DQB 1*0402 DPBI alleles DPB 1"0201

Control subjects

Patients with iridocyclitis

Patients without [ridocyclitis

(n = 175) 28 24 15 4 17 18 11 (n = 183) 43 23 10 87 (n = 172) 56 15 9 (n = 185) 47

(n = 61) 5 10 19 3 17 5 21 (n = 62) 9 20 20 45 (n = 60) 34 12 17 (n = 58) 35

(n = 69) 23 4 8 3 12 10 18 (n = 71) 26 l3 21 30 (n = 69) 17 13 12 (n = 68) 32

Odds ratio*

Chi-square*

pt

p (adjusted)

0.18 3.19 3.45 1.14 1.80 0.53 1.49

12.1 3.78 7.52 ND ND ND ND

<0.001 0.049 0.006 -----

0.003 0.29 0.036 -----

0.29 1.94 1.13 3.62

8.34 ND ND 12.38

0.004 --<0.001

0.024 --0.003

4.00 1.08 1.88

13.77 ND ND

<0.001 ---

0.001 ---

1.71

ND

--

--

ND, Not done. *Based on comparison of patients with and without iridocyclitis. "~Basedon chi-square value with l degree of freedom.

was performed overnight, also at 37 ~ C, in a n o t h e r solution containing S S C solution, D e n h a r d t solution, salmon sperm D N A , and a previously labeled probe. A f t e r that, filters were washed in 6 • S S C solution, twice at room temperature and once for 5 minutes at a specific t e m p e r a t u r e appropriate to the nucleotide sequence of each probe. T h e filters were exposed to Kodak ( X A R ) film ( E a s t m a n K o d a k Co., Rochester, N . Y . ) for an average of 18 hours. Statistical analysis. Proportional data were analyzed by chi-square test with one degree of freedom without Yates correction or, when appropriate, by the Fisher Exact Test. Odds ratios were calculated according to the m e t h o d of Woolf. 16 Values were adjusted for multiple comparisons by using the Bonferroni correction. I f p values b e c a m e nonsignificant after correction, both unadjusted and adjusted values are shown. In all cases, p values less than or equal to 0.05 were considered statistically significant. Because limited amounts of D N A were available from some patients, it was impossible to test all subjects for all alleles. T h e denominators used in the calculations of gene frequencies, therefore, vary somewhat from the total of 164 patients and 218 control subjects. RESULTS

Patients with EOPA-JRA compared with control subjects. A m o n g the splits of H L A - D R 5 , only H L A D R B 1" 1104 showed a statistically significantly increased

frequency a m o n g patients with E O P A - J R A ( 2 7 / 1 3 0 ; 21%) compared with control subjects ( 1 5 / 1 7 5 ; 7%; chi-square value = 9.35; p = 0.003). E O P A patients with iridoeyclitis compared with those without the eye disease. The H L A - D R B 1" 1104 gene was the only split of H L A - D R 5 associated statistically (after a d j u s t m e n t ) with increased risk of eye disease (Table I). Thirty-one percent of the children with eye disease h a d this gene, compared with 12% of those with arthritis alone (p = 0.036 adjusted). H L A - D R B I * 1 1 0 1 showed some associated risk (16% of patients with iridocyclitis vs 6% of patients without iridocyclitis), but the results were not statistically significant after a d j u s t m e n t (p = 0.29 adjusted). H L A - D Q A 1 * 0 5 0 1 , in linkage disequilibrium with H L A D R B 1" 1104, was associated with increased risk (p <0.003 adjusted). O t h e r H L A - D Q A 1 genes h a d no statistically significant associations or were so infrequent t h a t no analysis was possible. The H L A - D Q B 1 "0301 gene, also in linkage disequilibrium with H L A - D R B 1 * 1104, was associated with increased risk of eye disease (p <0.001 adjusted). Some protection against eye disease was conferred by H L A - D R 1; only 5 (8%) of 61 patients with iridocyclitis h a d this gene, compared with 23 (33%) of 69 patients without eye disease (p <0.003 adjusted). H L A - D Q A I * 0 1 0 1 , in strong linkage disequilibrium with H L A - D R 1, also showed a protective effect as described previously (p = 0.024 adjusted). 17 H L A - D R B I * 1 3 0 1 , a split of H L A - D R w 6

Volume 121 Number 1

strongly associated with EOPA-JRA, showed some increased risk of eye disease. HLA-DRB 1"0801, also associated with EOPA-JRA, contributed no additional risk of iridocyclitis, again consistent with earlier reports, v, s Gene combinations and haplotypes in patients with EOPA. We created subsets of patients with different gene combinations and observed the risk of iridocyclitis. Among the 124 patients who were tested for HLA-DRB1 and -DPB1 alleles, 13 had both D R B I * l l 0 4 and DPBI*0201; 11 of these had eye disease (odds ratio, 7.77; chi-square value, 8.73; p <0.021 adjusted). Because only five subjects had both D R B I * l l 0 1 and DPBI*0201, a similar interaction between these two genes could not be investigated. The possibility of a hitherto unsuspected linkage disequilibrium between the two genes was tested in our normal population. Of 47 individuals with DPBl*0201, eight also had DRB 1* 1104 genes, yielding positive relative delta value of 0.359 and statistically significant chi-square value (p = 0.021); however, four of the eight subjects were of Italian-American ancestry, suggesting that a genetic founder effect, rather than genetic linkage (i.e., a specific haplotype carrying both DPBl*0201 and DRBI*1104), might be responsible for this association in normal subjects. Some evidence has been presented and recently summarized for a third class II gene in this group of patients. 18 Of 18 HLA-DRBI*1104 patients with eye disease, 16 had an additional HLA-DRw52 haplotype (a super type linked to the HLA-DRB1 locus genes expressing the HLA-DR3, 5, w6, and w8 serologic specificities), which was found in only 4 of 12 control subjects with H L A - D R B I * l l 0 4 (Fisher Exact Test: p = 0.004). In some patients the additional HLA-DRB 1 gene was independently associated with a risk of EOPA-JRA. Either of two of these genes, HLADRBI*0801 or HLA-DRBI*1301, was present in 8 of 18 patients with eye disease, as well as H L A - D R B I * l l 0 4 , compared .with none of 12 control subjects with HLADRBI* 1104 (Fisher Exact Test: p <0.01). In contrast, patients without iridocyclitis had less evidence of a second D R / D Q haplotype, providing additional evidence of a gene dose effect in generating susceptibility to eye disease (data not given). We also found that 12 of 19 subjects with EOPA-JRA and eye disease who had HLA-DRB 1" 1104 also possessed HLA-AZ, a class I gene that is also associated with JRA. 19 Antinuclear antibodies. We investigated the possibility that the presence of ANAs is a more valid indicator of iridocyclitis risk than is the H L A - D R B I * l l 0 4 gene. Only patients with EOPA-JRA whose ANA determinations used human substrate as the antigen were included in this analysis. (Older methods using mouse or rat liver as the substrate are known to be less sensitive and to yield lower titers of ANA.) A total of 65 patients tested for ANA by

HLA in chronic iridocyclitis o f JRA

59

Table II, Diagnostic validity statistics associated with antinuclear antibodies and HLA-DRB 1" 1104 in patients with EOPA-JRA, with and without chronic iridocyclitis

Sensitivity Specificity Positive predictive value Negative predictive value Etiologic fraction*

ANA

HLA-DRBI"1404

(%)

(%)

85.3 23.4 44.6 68.8 37.1

31.1 88.4 70.4 59.2 22.1

*Etiologicfractionis the percentageof cases in the populationthat can be attributed to the presenceof the factor (see reference24).

using human substrate had titers that were considered definitely positive (according to the standards of the laboratory in which the test was performed), including 29 (85%) of 34 patients with eye disease, and 36 (77%) of 47 patients without eye disease (chi-square value = 0.94; p = 0.67). Table I1 summarizes the diagnostic validity statistics associated with ANA and HLA-DRBI*1104. The presence of HLADRB 1* 1104 was about four times as specific, but only about one third as sensitive, as ANAs in estimating risk of eye disease. DISCUSSION These data suggest that the major HLA-DR5(wl 1) split that predisposes children with EOPA-JRA to chronic iridocyclitis is carried on HLA-DRB 1" 1104 haplotypes. Our findings support earlier observations that the DRB 1* 1104 effect in this disease is dependent on the presence of at least one other major histocompatibility class II gene, HLA-DPBI*0201.2~ 21 In a substantial proportion of the patients with DRB 1" 1104, there is also evidence of an additional major histoeompatibility class II (DR or DQ) gene effect, again in accordance with data published earlier involving many of these same patients, z~ Comparison of patients with iridocyclitis versus those without iridocyclitis suggested that the DRB 1* 1101 gene carries some increased risk, although the overall data do not favor this gene as a major factor in disease association. Whether the DRBI*1104 gene effect is a direct consequence of DRB1 *1104 or linked genes, especially of HLADQ, is unresolved at this time. A study regarding patients with a more erosive form of pauciarticular JRA and less eye disease yielded evidence that the HLA-DQ locus contributes the important gene, a finding that requires confirmation in other series. Iv In addition to the DRB 1* 1104 interaction with DPBI*0201 predicted by earlier studies, 2~ strong interactions with DRBI*1301 were also noted; DRBl*0801 haplotypes appear to generate a disease effect with less interaction. Our data supporting the need for the

60

Melin-Aldana et al.

D R / D Q gene in addition to D R B I * 1104 indicate a very complex role for H L A in the pathogenesis of E O P A - J R A . Homozygosity of H L A * D R B I * 1 1 0 4 was not seen. This striking interaction between D R B I * 1104 and DPB1 "0201 is consistent with previous studies of the serologically defined H L A - D R 5 specificity in this population carried out before the availability of oligonucleotide dot blot techniques. 22 The H L A - D R B 1" 1301/DPB 1 *0201 interaction is also in agreement with recently published data. 23 The extent to which these observations can be used to establish clinical predictors for the development of iridocyclitis in patients with J R A remains unclear. Actuarial analysis shows that HLA-DR5 haplotypes (including D R B 1" 1104) impose an ongoing higher risk of eye disease than that imposed by H L A - D R B I * 0 8 0 1 or H L A D R B 1* 1301, both of which carry minimal or no additional risk beyond the initial disease association. 7 The H L A D R B 1* 1104/DPB 1 *0201 association is particularly pertinent in that eye disease developed in all but 2 of the 13 patients with this combination of alleles. Frequent slit-lamp examination of all patients with E O P A - J R A is the only means currently available for prompt detection of this potentially blinding eye disease; neither the young patients nor their parents are usually aware of the development of the disease. Further studies of this gene combination as a possible predictor of iridocyclitis are indicated. Combining results of genetic testing and A N A data should lead to higher prognostic sensitivity and specificity, as suggested here. Such studies might also include evaluation of the value of A N A in serum in comparison with other H L A gene markers. The overall prevalence of A N A in more than 80% of patients with E O P A - J R A suggests that the A N A cannot be explained solely by a specific immune response associated with H L A - D R B I * 1104 gene or haplotype. We thank Louise Kittredge for editorial assistance and Pat Donnelly for HLA-DP typing. REFERENCES

1. Chylack LT Jr. The ocular manifestations of juvenile rheumatoid arthritis. Arthritis Rheum 1977;20(suppl):217-23. 2. Cassidy JT, Sullivan DB, Petty RE. Clinical patterns of chronic iridocyclitis in children with juvenile rheumatoid arthritis. Arthritis Rheum 1977;20(suppl 2):224-7. 3. Rosenberg AM. Uveitis associated with juvenile rheumatoid arthritis. Semin Arthritis Rheum 1987;16:158-73. 4. Cassidy JT, Brody GL, Martel W. Monarticular juvenile rheumatoid arthritis. J PEDIATR 1967;70:867-75. 5. Bywaters EGL, Ansell BM. Mona,~r~icular,. arthritis in children. Ann Rheum Dis 1965;24:116-22. 6. Kanski JJ. Uveitis in juvenile chronic arthritis. Clin Exp Rheum 1990;8:499-503.

The Journal of Pediatrics July 1992

7. Giannini EH, Malagon C, Van Kerckhove C, et al. Longitudinal analysis of HLA-associated risks for iridocyclitis in juvenile rheumatoid arthritis. J Rheumatol 1991;18:1394-7. 8. Malagon C, Van Kerckhove C, Giannini EH, et al. The iridocyclitis of early-onset pauciarticular JRA: outcome in immunogenetically characterized patients. J Rheumatol 1992; 19:160-3. 9. Cassidy JT, Levinson JE, Bass JC, et al. A study of classification criteria for a diagnosis of juvenile rheumatoid arthritis. Arthritis Rheum 1986;29:274-81. 10. Bodmer JG, Marsh SGE, Albert ED, et al. Nomenclature for factors of the HLA system, 1990. Hum Immunol 1991;31:18694. 11. Fernandez-Vina M, Shumway W, Stastny P. DNA typing for class II HLA antigens with allele-specific or group-specific amplification. II. Typing for alleles of the DRw52-associated group. Hum Immunol 1990;28:51-64. 12. Khalil I, d'Auriol L, Gobet M, et al. A combination of HLADQB asp57-negative and HLA DQA arg52 confers susceptibility to insulin-dependent diabetes mellitus. J Clin Invest 1990;85:1315-9. 13. Van Kerckhove C, Melin-Aldana H, Elma MS, et al. A distinct HLA-DRw8 haplotype characterizes patients with juvenile rheumatoid arthritis. Immunogenetics 1990;32:304-8. 14. Angelini G, Bugawan TL, Delfino L, Erlich HA, Ferrara GB. HLA-DP typing by DNA amplification and hydrization with specific oligonucleotides. Hum Immunol 1989;26:169-77. 15. Fernandez-Vina M, Moraes ME, Stastny P. DNA typing for class II HLA antigens with allele-specific or group-specific amplification. III. Typing for 24 alleles of HLA-DP. Hum Immunol 1991;30:60-8. 16. Woolf B. On estimating the relationship between blood groups and disease. Ann Hum Genet 1955;19:251-3. 17. Van Kerckhove C, Luyrink L, Taylor J, et al. HLADQAI*0101 haplotypes and disease outcome in early-onset pauciarticular juvenile rheumatoid arthritis. J Rheumatol 1991;18:874-9. 18. Nepom DS, Glass DN. Juvenile rheumatoid arthritis and HLA: report of the Park City III Workshop. J Rheumatol (in press). 19. Oen K, Petty RE, Schroeder ML. An association between HLA-A2 and juvenile rheumatoid arthritis in girls. J Rheumatol 1982;9:916-20. 20. Van Kerckhove C, Luyrink L~ Elma MS, et al. HLA-DP/DR interaction in,children with juvenile rheumatoid arthritis. Immunogenetics 1990;32:364-8. 21. Begovich AB, Bugawan TL, Nepom BS, Klitz W, Nepom GT, Erlich HA. A specific HLA-DPB allele is associated with pauciarticular juvenile rheumatoid arthritis but not adult rheumatoid arthritis. Proc Natl Acad Sci USA 1989;86:9489-93. 22. Hoffman RW, Shaw S, Francis LC, et al. HLA-DP antigens in patients with pauciarticular juvenile rheumatoid arthritis. Arthritis Rheum 1986;29:1057-62. 23. Forre O, Ploski R, Vinje O, Ronningen KS, Sorskaar D, Vartdal F. Genomic typing of juvenile rheumatoid arthritis patients: a preliminary report [Abstract]. Arthritis Rheum 1991;34(suppl):S 139. 24. Kramer MS. Clinical epidemiology and biostatistics. New York: Springer-Verlag, 1988:105-6.