Arthritis Rheum. 2004;50:1179â86. 3. El-Omar EM, Carrington M, Chow WH, McColl KE, Bream JH, ...... Adam M. Huber, MD. IWK Health Centre and Dalhousie ...
ARTHRITIS & RHEUMATISM Vol. 52, No. 1, January 2005, pp 361–370 © 2005, American College of Rheumatology
LETTERS examined, dosage-dependent, accounted for ⬃20% of the observed variation in the levels of both mediators and persisted in the presence of enhancers (e.g., interferon-␥) or suppressors (e.g, dexamethasone) of the IL-1 response. More recently, we identified a novel SNP within this length polymorphism that modifies a putative Oct-1 binding site and may therefore underlie the observed regulatory function (Vamvakopoulos J, et al: unpublished observations). In my opinion, the real potential in the line of investigation pursued by Hall et al lies in highlighting the interindividual variability in ATP-stimulated secretion as a possible genetic contributor to the regulation of IL-1 bioactivity. Their data on stimulus-induced secretion, which would escape confounding by monocyte count variability, certainly suggest this. It is surprising that those authors attempted to link IL1B promoter polymorphisms (which might only affect the rate of gene transcription) to differential secretion triggered by a stimulus acting at the level of the plasma membrane. It is now well established that ATP acts via the P2X7 receptor to trigger IL-1 release (6,7). Several polymorphisms exist within the P2RX7 locus, and reports suggest that at least 3 of these (R307Q, E496A, and I568N) give rise to loss-of-function phenotypes associated with greatly reduced responses to ATP (8–10). In this context, and despite the low frequencies of the aforementioned loss-of-function alleles, proper interpretation of the data collected by Hall et al would, at the very least, have called for concurrent study of polymorphisms in the P2RX7 locus. In conclusion, existing best evidence argues against a significant regulatory influence of polymorphisms within the IL1B locus on the IL-1 response to LPS stimulation. Conversely, polymorphisms within the IL1RN locus appear to cross-regulate IL-1 by modifying expression of the IL-1Ra. Comprehensive, large-scale studies that would provide definitive answers on this issue are unfortunately still lacking; there is, therefore, a pressing need for thorough and critical review of existing literature before embarking on further work seeking to clarify the genetic regulation of IL-1.
DOI 10.1002/art.20756
ILIRN is a prominent genetic regulator of interleukin-1 release To the Editor: I read with interest the recent report by Hall et al (1). Those authors genotyped 2 populations (n ⫽ 25 and n ⫽ 31) for 2 promoter and 1 coding single-nucleotide polymorphisms (SNPs) of the IL1B gene locus, quantified basal and maximal (⫹ ATP) lipopolysaccharide (LPS)–induced interleukin-1 (IL-1) secretion in a whole blood assay, and applied statistical tests to detect possible associations among these variables. IL-1 bioactivity is tightly regulated by an elaborate network of molecular interactions operating at many levels. Genetic polymorphism lies at the heart of this regulatory mechanism, because it may affect most of the key molecular components. In this respect, the scope of the aforementioned study seems rather limited, as it only examined polymorphisms within a single locus (IL1B). Furthermore, the methodology behind the whole blood assay for determination of IL-1 secretory status is suspect: the authors explain that monocytes are the predominant source of IL-1 in this assay, yet blood monocyte counts were not determined. Considering the high rates of IL-1 secretion by blood monocytes (2) and the expected direct correlation between monocyte count and amount of secreted cytokine, there is little in the overall design of this study that could have accounted for the error introduced here. On top of these fundamental limitations and the small populations involved in this study, the assumption of a recessive (rather than gene dosage) effect made by Hall et al, though apparently supported by their data, has likely produced artificially inflated P values, which exaggerate the importance of their findings but are nevertheless quoted in the abstract. It is also disappointing that, although Hall et al acknowledge in passing that other polymorphisms outside the IL1B locus may affect the IL-1 response to LPS, they make no effort to discuss existing data in this respect. In fact, the early work by Dinarello, Warner, and others showed that IL-1, acting via IL-1 receptor I (IL-1R1), drives its own production and release (3,4). Other authors have shown that the IL-1 receptor antagonist (IL-1Ra) and other extracellular regulators (e.g., IL-1R2) effectively modulate the IL-1 response by interrupting this positive feedback loop and have suggested a regulatory role for polymorphisms within the IL1RN locus (5). Expanding on this model, my colleagues and I examined the relative influence of selected polymorphisms across the IL1 gene cluster (including the IL1B ⫺31 and ⫹3953 SNPs studied by Hall et al) on LPS-stimulated IL-1 and IL-1Ra release in a controlled in vitro environment (2). We reported that none of these polymorphisms appreciably modified the copious synthesis and release of these mediators by LPS-stimulated, freshly isolated monocytes. Conversely, we found that the length polymorphism within intron 2 of the IL1RN locus was a prominent genetic regulator of both IL-1Ra and IL-1 release by differentiating monocytes maintained in culture for at least 24 hours. The effect of this single IL1RN polymorphism was independent of all other polymorphisms
Joannis Vamvakopoulos, MsC, PhD Biomedicum Helsinki Helsinki, Finland 1. Hall SK, Perregaux DG, Gabel CA, Woodworth T, Durham LK, Huizinga TW, et al. Correlation of polymorphic variation in the promoter region of the interleukin-1 gene with secretion of interleukin-1 protein. Arthritis Rheum 2004;50:1976–83. 2. Vamvakopoulos JE, Green C, Metcalfe S. Genetic control of IL-1 bioactivity through differential regulation of the IL-1 receptor antagonist. Eur J Immunol 2002;32:2988–96. 3. Dinarello CA, Ikejima T, Warner SJ, Orencole SF, Lonnemann G, Cannon JG, et al. Interleukin 1 induces interleukin 1: I. Induction of circulating interleukin 1 in rabbits in vivo and in human mononuclear cells in vitro. J Immunol 1987;139:1902–10. 4. Warner SJ, Auger KR, Libby P. Interleukin 1 induces interleukin 1: II. Recombinant human interleukin 1 induces interleukin 1 production by adult human vascular endothelial cells. J Immunol 1987;139:1911–17. 361
362
5. Danis VA, Millington M, Hyland VJ, Grennan D. Cytokine production by normal human monocytes: inter-subject variation and relationship to an IL-1 receptor antagonist (IL-1Ra) gene polymorphism. Clin Exp Immunol 1995;99:303–10. 6. Perregaux DG, McNiff P, Laliberte R, Conklyn M, Gabel CA. ATP acts as an agonist to promote stimulus-induced secretion of IL-1 and IL-18 in human blood. J Immunol 2000;165:4615–23. 7. MacKenzie A, Wilson HL, Kiss-Toth E, Dower SK, North RA, Surprenant A. Rapid secretion of interleukin-1 by microvesicle shedding. Immunity 2001;15:825–35. 8. Gu BJ, Sluyter R, Skarratt KK, Shemon AN, Dao-Ung LP, Fuller SJ, et al. An Arg-307 to Gln polymorphism within the ATPbinding site causes loss-of-function of the human P2X7 receptor. J Biol Chem 2004;279:31287–95. 9. Sluyter R, Shemon AN, Wiley JS. Glu496 to Ala polymorphism in the P2X7 receptor impairs ATP-induced IL-1 release from human monocytes. J Immunol 2004;172:3399–05. 10. Wiley JS, Dao-Ung LP, Li C, Shemon AN, Gu BJ, Smart ML, et al. An Ile-568 to Asn polymorphism prevents normal trafficking and function of the human P2X7 receptor. J Biol Chem 2003;278: 17108–13.
DOI 10.1002/art.20887
Reply To the Editor: We would like to thank the editors for the opportunity to reply to the observations of Dr. Vamvakopoulos regarding our report. As alluded to in his comments, the regulation of IL-1 bioactivity is a complex and tightly regulated process involving numerous molecular interactions that could be susceptible to genetic influences. Based on this complexity we did not attempt to investigate each individual component for genetic influences impacting regulation of IL-1 bioactivity, but instead focused our efforts on polymorphisms within the IL1B gene itself that may explain a portion of a significant interindividual variability previously observed for LPSstimulated secretion of mature IL-1. Perregaux et al (1) observed, using an ex vivo blood-based IL-1 output assay, that the efficiency at which LPS alone promoted release of IL-1 (relative to that released by a secretion stimulus such as ATP) was remarkably donor dependent. In our recent report, donor variability in the LPS-inducible output of IL-1 was observed across 2 independent study populations, supporting the previous observation. There are an increasing number of reports in the literature associating IL1B gene polymorphisms with diseases having an inflammatory element to their pathogenesis (2–5), and polymorphisms that affect IL-1 levels could certainly be envisioned to contribute to disease processes. Based on these reports and the published functional consequences of IL1B gene polymorphisms, we tested the hypothesis that variation within the IL1B locus may explain a portion of the intersubject variability observed in the ex vivo blood assay. While the sample size in our first study was small, we were able to observe a significant association between a haplotype comprising 3 IL1B polymorphisms (C-511T, T-31C, and C3954T) and LPS-
LETTERS
induced secretion of IL-1 protein. This finding supported our initial hypothesis and suggested a recessive effect for which we reported both gene dosage and recessive effect P values, both of which were significant. More importantly, we were able to replicate this specific effect in an independent population, again supporting a potential recessive effect. While the numbers in each study are small, the fact that we observed the same association in 2 independent studies provides strong evidence that IL1B gene variation contributes to the intersubject variability in IL-1 output following LPS challenge. In both study populations there was extensive linkage disequilibrium (LD) between the C-511T and T-31C polymorphisms and to a lesser extent with the T3954C polymorphism in exon 5. Based on this and the reported LD across this region, it is unclear which allele or combination of alleles is contributing to this variation. Further studies are needed to elucidate this. While it is recognized that variation in the IL1RN gene contributes to variation in IL-1Ra and IL-1 protein production, there is little to suggest that this polymorphism directly influences the mechanism of IL-1 secretion (6,7). It is well established that the extent of LD between the C-511T polymorphism in IL1B gene and the IL1RN variablenumber tandem repeat (VNTR) is minimal; thus, these variants would be expected to segregate independently and perhaps have independent effects on function. In fact, we recently have shown that the C-511T polymorphism acts independently of the IL1RN VNTR allele 2 in conferring risk of osteoarthritis susceptibility, and others have also reported significant associations that are driven by either IL-1 or IL-1Ra (2–5). In the design of our study we utilized an ex vivo bood assay for assessing IL-1 levels. We believe that this type of assay is likely to be representative of the cytokine output capacity existing in vivo, as this approach involves minimal cell manipulation. Thus, we minimized changes in cytokine output capacity resulting from environmental factors introduced during cell culture. As highlighted by Dr. Vamvakopoulos, the number of monocytes is expected to influence the amount of cytokine released. For this very reason we normalized the IL-1 output value obtained with LPS alone to that observed with a combination of LPS and ATP. The combination of LPS and ATP is established to be an efficient mechanism for achieving maximum release of IL-1. Thus, by normalizing to the amount of IL-1 released in the presence of the dual stimulus, which is referred to as the stimulus-induced secretion ratio, we evaluated relative ability of LPS to promote cytokine release in a manner that is independent of monocyte number. Understanding the many genetic factors that influence IL-1 bioactivity will obviously require a great deal of additional work and is outside the scope of our focused study. We have established a link between IL1B gene polymorphisms and the efficiency at which IL-1 is externalized in response to LPS challenge, providing compelling evidence that this atypical secretory process is under the control of genetic variation. Due to the extent of LD through this region, it is not clear at this time whether the observation we have reported is due solely to these variants in the promoter region of the IL1B gene itself or to some other allele or gene within the IL1 cluster. The cytokine output mechanism that we assessed is not expected to be affected by transcription rates, but at this point our knowledge of how transcription, translation, and posttranslational modification reactions are influenced by genetic variation is
LETTERS
363
unclear and warrants further study. It is well understood in the genetics community that for complex pathways and diseases, phenotypic differences are often the result of multiple genes (including but not limited to P2X7 receptor and other genes in the IL1 cluster in this case) and environmental conditions acting in concert to express a given condition. The data presented in our report contribute to the literature by describing a replicated observation that IL1B polymorphism may account for a portion of the genetic component of intersubject variability in IL-1 secretion and providing functional evidence supporting the reported genetic associations between IL1B polymorphisms and human inflammatory disease. Stephanie K. Hall, BS Christopher A. Gabel, PhD Albert B. Seymour, PhD Pfizer Global Research and Development Groton, CT 1. Perregaux DG, McNiff P, Laliberte R, Conklyn M, Gabel CA. ATP acts as an agonist to promote stimulus-induced secretion of IL-1 and IL-18 in human blood. J Immunol 2000;165:4615–23. 2. Meulenbelt I, Seymour AB, Nieuwland M, Huizinga TW, van Duijn CM, Slagboom PE. Association of the interleukin-1 gene cluster with radiographic signs of osteoarthritis of the hip. Arthritis Rheum 2004;50:1179–86. 3. El-Omar EM, Carrington M, Chow WH, McColl KE, Bream JH, Young HA, et al. Interleukin-1 polymorphisms associated with increased risk of gastric cancer. Nature 2000;404:398–402. 4. Grimaldi L, Casadei VM, Ferri C, Veglia F, Licastro F, Annoni G, et al. Association of early-onset Alzheimer’s disease with an interleukin-1a gene polymorphism. Ann Neurol 2000;47:361–8. 5. Timms A, Crane A, Sims AM, Cordell H, Bradbury L, Abbott A, et al. The interleukin 1 gene cluster contains a major susceptibility locus for ankylosing spondylitis. Am J Hum Genet 2004;75:587–95. 6. Carter MJ, Jones S, diGiovine FS, Camp NJ, Lobo AJ, Dugg GW. Allele 2 of the interleukin-1 receptor antagonist gene polymorphism is associated with reduced expression interleukin-1 receptor antagonist in ulcerative colitis. Gastroenterology 1998;114:A97. 7. Santtila S, Savinainen K, Hurme M. Presence of the IL-1RA allele 2 (IL1RN*2) is associated with enhanced IL-1 production in vitro. Scand J Immunol 1998;47:195–8.
DOI 10.1002/art.20778
Programmed death 1 gene polymorphism and systemic lupus erythematosus in different ethnic populations: comment on the article by Lin et al To the Editor: In a recent issue of Arthritis & Rheumatism, Lin et al reported on the identification of a single-nucleotide polymorphism in the gene for programmed death 1 (PD-1) and its association with rheumatoid arthritis but not with systemic lupus erythematosus (SLE) (Lin SH, Yen JH, Tsai JJ, Tsai WC, Ou TT, Liu HW, et al. Association of programmed death 1 gene polymorphism with the development of rheumatoid arthritis, but not systemic lupus erythematosus. Arthritis
Rheum 2004;50:770–5). Investigators at this institution have previously described the association of a polymorphism in the fourth intron of the PD-1 gene with SLE (Prokunina L, Castillejo-Lopez C, Oberg F, Gunnarsson I, Berg L, Magnusson V, et al. A regulatory polymorphism in PDCD1 is associated with susceptibility to systemic lupus erythematosus in humans. Nat Genet 2002;32:666–9). The polymorphism we described, called PD1.3, is a G⬎A change in which the associated allele A leads to the disruption of a transcription factor or repressor binding site for the family of runt-domain proteins, possibly runt-related transcription factor 1. The polymorphism described by Lin et al is the same as the one we describe as PD1.5. It is not functional, and more importantly, is not in tight linkage disequilibrium with PD1.3. This could explain why those investigators did not identify an association with SLE. However, more likely explanations are the small sample size and the use of only 1 polymorphism. Preliminary results show that PD1.3 allele A is extremely rare in the Chinese population, suggesting that polymorphisms in this gene other than PD1.3 or PD1.5 may be important in Chinese SLE patients. Alternatively, it is possible that PD-1 is not important in Chinese SLE patients due to genetic heterogeneity or ethnic differences. Thus, the negative association with SLE observed by Lin et al does not contradict our previous finding of a positive association with SLE in Caucasian patients. Marta E. Alarcon-Riquelme, MD, PhD University of Uppsala Uppsala, Sweden
DOI 10.1002/art.20888
Reply To the Editor: In our article we reported that a nonfunctional singlenucleotide polymorphism (SNP), C872T, in exon 5 of the PD-1 gene is associated with the development of rheumatoid arthritis (RA), but not SLE, in a Chinese population. In contrast to our results, an SNP in intron 4 of the PD-1 gene, named PD-1.3, was found by Prokunina et al to be associated with SLE susceptibility in Europeans (1). There are various possible explanations for these differing results regarding the connection between the PD-1 gene and SLE susceptibility. First, it may be argued that the small sample size accounts for the negative association in our study. However, although it might have been possible to reach a significant association by increasing the sample size, only a weak association with a low odds ratio would be expected, due to the similar genotype distributions between the Chinese SLE population and the Chinese normal populations. In fact, in Prokunina and colleagues’ study, the PD-1.5 SNP, identical to the C872T SNP in our study, was found to be only weakly associated with SLE susceptibility in Europeans, and this result can be considered consistent with ours. Second, it is possible that another functional PD-1 SNP, not in linkage disequilibrium with the C872T SNP, affects SLE susceptibility in the Chinese population. Interestingly, the
364
LETTERS
significant association between the nonfunctional C872T SNP and RA susceptibility in our study also implies the possible existence of another functional PD-1 SNP in linkage disequilibrium with the C872T SNP that could influence RA development. Therefore, if PD-1 truly has an effect on susceptibility to SLE as well as RA in the Chinese, the question is raised as to how 2 different functional genetic variations in a gene can differentially influence the development of 2 autoimmune diseases. Another more likely explanation is that PD-1 contributes to SLE susceptibility in Europeans, but not in the Chinese, due to the influence of genetic background (similar to the finding that deficiency of Fc␥ receptor IIB or PD-1 results in distinct autoimmune phenotypes in different mouse strains) (2–4). In addition, interpretation of the results observed in our study and those reported by Prokunina et al may be complicated by the possibility that there is another pathogenic gene, not the PD-1 gene itself, that is adjacent to the PD-1 gene and affects the development of SLE and/or RA. In conclusion, it appears that the results from the 2 studies show no contradiction, and more work is needed to clarify the actual role of PD-1 in affecting the development of SLE and RA in different ethnic groups. Shih-Chang Lin, MD, PhD Cathay General Hospital Taipei, Taiwan Chung-Jen Chen, MD Chang Gung Memorial Hospital Kaohsiung, Taiwan 1. Prokunina L, Castillejo-Lopez C, Oberg F, Gunnarsson I, Berg L, Magnusson V, et al. A regulatory polymorphism in PDCD1 is associated with susceptibility to systemic lupus erythematosus in humans. Nat Genet 2002;32:666–9. 2. Nishimura H, Nose M, Hiai H, Minato N, Honjo T. Development of lupus-like autoimmune diseases by disruption of the PD-1 gene encoding an ITIM motif-carrying immunoreceptor. Immunity 1999; 11:141–51. 3. Bolland S, Ravetch JV. Spontaneous autoimmune disease in Fc␥RIIB-deficient mice results from strain-specific epistasis. Immunity 2000;13:277–85. 4. Nishimura H, Okazaki T, Tanaka Y, Nakatani K, Hara M, Matsumori A, et al. Autoimmune dilated cardiomyopathy in PD-1 receptor-deficient mice. Science 2001;291:319–22.
DOI 10.1002/art.20769
Lack of response to anakinra in rheumatoid arthritis following failure of tumor necrosis factor ␣ blockade: comment on the article by Buch et al To the Editor: I read with interest the article by Buch and colleagues (1), which described the lack of response to anakinra (an interleukin-1 receptor antagonist) in patients with rheumatoid arthritis (RA) that failed to adequately respond to anti–tumor necrosis factor (anti-TNF) therapy. Limitations of this study include a small number of patients (n ⫽ 26) and the fact that
these patients had disease that was more severe than that of patients typically seen at a rheumatology practice, as evidenced by the mean disease duration (⬎15 years) and mean number of prior disease-modifying antirheumatic drugs (DMARDs) (n ⫽ 5). In fact, patients with these disease characteristics are not likely to respond to any treatment (2). Real-world data from a large registry of patients with RA demonstrates that anakinra may benefit patients with RA who have discontinued prior anti-TNF therapy. RADIUS 1 (Rheumatoid Arthritis DMARD Intervention and Utilization Study) is a prospective, multicenter, observational study designed to systematically document the use patterns, effectiveness, and safety of DMARDs currently used in the management of RA (3,4). This study enrolled ⬃5,000 patients with RA from 389 sites across the US between October 2001 and December 2002; all patients required treatment with a new DMARD (addition or switch). An ad hoc analysis was conducted to identify patients in whom anakinra therapy was initiated at the time of entry into RADIUS 1 and had previously received anti-TNF therapy, based on a data collection cutoff date of March 31, 2004. Demographics were tabulated, and efficacy outcomes were summarized at the visit closest to 6 months. Kaplan-Meier analysis was conducted to determine persistence on anakinra. Results were stratified by prior use of 1 or 2 TNF antagonists (etanercept and/or infliximab). A total of 192 patients in whom treatment with anakinra was initiated at the time of entry into RADIUS 1 had previously been treated with anti-TNF therapy; most had discontinued prior anti-TNF therapy because of lack of efficacy (46.9%) and/or adverse events (40.1%). The mean disease duration was 11.2 years, and patients had received a mean of 3.2 prior DMARDs. As expected, more patients who had previously received 2 TNF antagonists were considered by their physicians to have severe disease compared with patients who previously received only 1 anti-TNF agent (70.8% versus 43.8%). Overall, patients with RA who had previously discontinued anti-TNF therapy demonstrated improvements in all efficacy measures after 6 months of anakinra therapy, and, despite the need for daily injections, most continued to receive anakinra at 6 months (Table 1). Patient-reported outcomes may best discriminate the treatment effect of anakinra (5), and patients who had previously discontinued 2 TNF antagonists showed more robust improvement with anakinra therapy in patient-reported outcomes compared with those who had previously received only 1 anti-TNF agent. However, patients who have discontinued therapy with a TNF antagonist for safety reasons may be good candidates for anakinra therapy, based on its excellent safety profile (6) and the current efficacy results. In conclusion, real-world data from the RADIUS 1 registry suggest that patients with RA who have previously discontinued anti-TNF therapy may benefit from anakinra therapy. Given the limitations of the previous study and the results of the current report, it seems appropriate to consider targeting interleukin-1 with anakinra when RA has not responded to anti-TNF therapy.
LETTERS
365
Table 1. Patients with RA receiving anakinra after previously discontinuing anti-TNF therapy* 1 previous anti-TNF agent (n ⫽ 144)
Efficacy end points
n
Baseline
6 months
Mean individual change
HAQ score (0–3 scale) Tender joint count (28 joints) Swollen joint count (28 joints) Morning stiffness, median minutes Physician’s global assessment (0–10 scale) Patient’s global assessment (0–10 scale) Pain score on VAS (0–10 scale) % with mACR20 response of 3/4 % with mACR20 response of 2/4 % with persistence during anakinra therapy
131 127 127 130 130 128 131 – – –
1.5 16.2 13.9 105 6.5 6.5 6.3 – – –
1.5 12.0 9.9 60 5.2 5.8 5.8 12.5 20.1 68.7
⫺0.1 ⫺4.4 ⫺4.0 ⫺5.5 ⫺1.3 ⫺0.7 ⫺0.5 – – –
2 previous anti-TNF agents (n ⫽ 48)
P
n
Baseline
NS ⬍0.0001 ⬍0.0001 0.007 ⬍0.0001 0.002 0.047 – – –
43 42 42 42 42 41 42 – – –
1.6 15.9 14.1 240 7.1 7.0 7.0 – – –
6 months
Mean individual change
P
1.5 12.3 12.3 137 5.3 5.7 5.6 16.7 22.9 80.0
⫺0.1 ⫺3.6 ⫺1.9 ⫺103 ⫺1.8 ⫺1.4 ⫺1.5 – – –
NS 0.012 0.012 0.002 ⬍0.0001 0.005 0.001 – – –
* Except where indicated otherwise, values are the mean. P values were determined by Wilcoxon’s signed rank test: RA ⫽ rheumatoid arthritis; TNF ⫽ tumor necrosis factor; HAQ ⫽ Health Assessment Questionnaire; NS ⫽ not significant; VAS ⫽ visual analog scale; mACR20 ⫽ modified American College of Rheumatology 20% criteria, calculated based on tender and swollen joint counts and the presence of 2 or 3 of the remaining 4 criteria, excluding the erythrocyte sedimentation rate/C-reactive protein level, because laboratory tests were not mandated by the protocol. Dr. Schiff has received grant/research support, is on the Speakers’ Bureau, and/or is a consultant for the following organizations: Angiotech, Abbott, Amgen, Aventis, Bristol-Myers Squibb, Centacor, Genentech, Hoffman-La Roche, IDEC, Novartis, Merck, and Wyeth-Ayerst.
Michael H. Schiff, MD Denver Arthritis Clinic and University of Colorado Denver, CO 1. Buch MH, Bingham SJ, Seto Y, McGonagle D, Bejarano V, White J, et al. Lack of response to anakinra in rheumatoid arthritis following failure of tumor necrosis factor ␣ blockade. Arthritis Rheum 2004;50:725–8. 2. Anderson JJ, Wells G, Verhoeven AC, Felson DT. Factors predicting response to treatment in rheumatoid arthritis: the importance of disease duration. Arthritis Rheum 2000;43:22–9. 3. Keystone EC, Cush JJ, Lautzenheiser JL, Gibofsky A, Spencer-Green G. Rheumatoid arthritis disease-modifying anti-rheumatic drug (DMARD) intervention and utilization (RADIUS): design and implementation [abstract]. Ann Rheum Dis 2002;61 Suppl 1:366. 4. Cush JJ, Lautzenheiser JL, Markenson JA, Spencer-Green G. RADIUS: interim findings from a prospective, multicenter, observational outcomes study in rheumatoid arthritis (RA) patients [abstract]. Arthritis Rheum 2002;46 Suppl 9:S528. 5. Cohen SB, Strand V, Aguilar D, Ofman JJ. Patient- versus physician-reported outcomes in rheumatoid arthritis patients treated with recombinant interleukin-1 receptor antagonist (anakinra) therapy. Rheumatology (Oxford) 2004;43:704–11. 6. Fleischmann RM, Schechtman J, Bennett R, Handel ML, Burmester GR, Tesser J, et al. Anakinra, a recombinant human interleukin-1 receptor antagonist (r-metHuIL-1ra), in patients with rheumatoid arthritis: a large, international, multicenter, placebocontrolled trial. Arthritis Rheum 2003;48:927–34.
DOI 10.1002/art.20889
Reply To the Editor: We thank Dr. Schiff for his interest in our recent article. We agree that our study included only a small number of patients,
and that they did have severe disease. However, in favor of this study is the fact that it was prospective, with clearly defined outcomes and very well-defined patients. Furthermore, the study participants were specifically chosen not only because TNF blockade had previously failed, but also because of primary inefficacy of TNF blockade in the majority of cases (23 of 26). The hypothesis being tested was that these patients with pure nonresponse to TNF antagonists could reasonably have been expected to have a more interleukin-1 (IL-1)–driven disease. The fact was that they had virtually no measurable response to anakinra (and, in fact, had an outcome with anakinra that was worse than that of comparable patients not selected for TNF blockade resistance). Therefore, contrary to expectations, it appears that patients who would have responded to IL-1 are included in the population of TNFresponsive patients. In Dr. Schiff’s series, it would be of interest to know how the improvements observed in the patients reported correlated with the reason for discontinuation of anti-TNF therapy (i.e., safety or inefficacy), and whether the inefficacy was what we have termed primary or secondary (Buch MA, Seto Y, Bingham SJ, Bejarano V, Bryer D, White J, et al. C-reactive protein as a predictor of infliximab treatment outcome in patients with rheumatoid arthritis: defining subtypes of nonresponse and subsequent response to etanercept. Arthritis Rheum 2005;52:42–8). One would have expected secondary TNF nonresponders (who initially respond to a TNF antagonist) to have had better responses. We concur that patients who have discontinued TNF antagonist therapy for safety reasons may respond to anakinra therapy. We still strongly believe that patients with primary failure to TNF blockade are not those with an IL-1–driven disease. M. H. Buch, MBChB, MRCP S. J. Bingham, MA, MBBChir, MRCP P. Emery, MA, MD, FRCP University of Leeds Leeds, UK
366
DOI 10.1002/art.20738
Fetal first-degree heart block, or where to set the confidence limit: comment on the article by Sonesson et al To the Editor: Sonesson et al have showed that first-degree heart block was strongly related to the transplacental passage of anti–Ro 52-kd antibodies into the fetus (1). They suggest that the risk for a fetus to develop first-degree heart block in such conditions is 33%, well above the largely reported risk of 2–5%. However, from the data provided by the authors (in their Table 2 and Figure 1), it seems that most of the fetuses with first-degree heart block had atrioventricular (AV) conduction times very close to the upper 95% confidence limit. Setting the upper confidence limit at ⬃135–140 msec (2,3) instead of 130 msec would put 6 of these 8 fetuses within the normal range for AV conduction. The confidence interval for normal AV conduction presented by the authors is surprisingly very narrow, and questionable for 2 different reasons. 1) Using echo machines, the inherent temporal resolution of the Doppler signal varies from 10 msec to 20 msec (4). This itself would bring the 95% confidence interval of even a perfect digitization to ⫾20–40 msec (versus the 16–20 msec reported by the authors). 2) The digitization itself carries a non-negligible inherent error as pointed out by Glickstein et al (2), who also found the 95% confidence interval for measuring mechanical PR interval to be 22 msec using left ventricular inflow/outflow Doppler methodology (1 of the 2 methods used by Sonesson et al). In fact, the authors of this present study (1) used previously published normal data (3) with a 95% confidence interval of 20 msec. In addition to the question regarding the width of the confidence interval, the mean PR interval reported by Sonesson et al is 10 msec lower than the previously published data (2). Thus, it seems that the high first-degree heart block incidence reported by Sonesson et al, which has not been found by any other group, may be explained by their low mean and narrow range of normal data. The spontaneous return to normal AV conduction would then be a mere fluctuation around the upper 95% limit. Another important issue relates to the feasibility of the methods. Sonesson et al use the previously published data on 284 normal fetuses (3) to serve as controls, to which they added 20 more subjects ages 17–22 weeks. However, looking at the data in Table 2 of their article, the control group is not 304 but only 142 by the inflow/outflow method, and is even lower (75) by the superior vena cava and aorta (SVC-Ao) method. This suggests that these methods are not highly feasible. Data were obtained using the SVC-Ao method in only 14 of 24 patients in the autoantibody-exposed group (58%), a low feasibility similar to the 63% reported by Andelfinger et al (3). We would also like to indicate that bradycardia in the non–heart block group is a finding which has not been observed by us or by others. Although Sonesson et al do not provide the raw data on which Table 2 is based, the difference between the 2 populations with widely overlapping standard deviations is probably not statistically significant. In conclusion, we agree that introducing the novel method of utilizing SVC-Ao recordings as a methodology for
LETTERS
measuring atrioventricular conduction in the fetus is important. However, we suggest caution in interpreting the results in view of the questionable definition of first-degree heart block. The risk for heart block is probably still 2–5%, and not 33%. The presumed evidence of spontaneous fetal first-degree heart block recovery is consequently doubtful. Azaria J. J. T. Rein, MD Dror Mevorach, MD Zeev Perles, MD Amiram Shovali Uriel Elchalal, MD Hebrew University Jerusalem, Israel 1. Sonesson SE, Salomonsson S, Jacobsson LA, Bremme K, WahrenHerlenius M. Signs of first-degree heart block occur in one-third of fetuses of pregnant women with anti–SSA/Ro 52-kd antibodies. Arthritis Rheum 2004;50:1253–61. 2. Glickstein J, Buyon J, Kim M, Friedman D, PRIDE investigators. The fetal Doppler mechanical PR interval: a validation study. Fetal Diagn Ther 2004;19:31–4. 3. Andelfinger G, Fouron JC, Sonesson SE, Proulx F. Reference values for time intervals between atrial and ventricular contractions of the fetal heart measured by two Doppler techniques. Am J Cardiol 2001;88:1433–6. 4. Rein AJ, O’Donnell CP, Geva T, Nir A, Perles Z, Hashimoto I, et al. Use of tissue velocity imaging in the diagnosis of fetal cardiac arrhythmias. Circulation 2002;206:1827–33.
DOI 10.1002/art.20890
Reply To the Editor: We want to thank Dr. Rein and colleagues for their interest in our report, and for highlighting the question of first-degree heart block in the neonatal lupus syndrome. Having reassessed all our data, performed new analyses, and taken the issues raised into account, our interpretation of the data remains the same; fetal heart block is considerably more common than previously understood, with an incidence of first-degree heart block approaching one-third in Ro 52 autoantibody–positive pregnant women. We believe this finding is an important step forward in elucidating the pathogenic mechanism, as well as having clinical implications in terms of patient assessment and followup. To clarify this statement and support our conclusions, we would like to make the following comments in response to the issues raised. The risk of having a fetus with congenital heart block (CHB) is generally considered to be 2–5% in women with anti-Ro and/or anti-La antibodies. However, in a subgroup of anti–Ro 52-kd–positive women a higher risk is suggested. In our study, 2 of 24 (8%) had second- or third-degree heart block. In addition, the risk of having a fetus with first-degree heart block has to our knowledge not been previously evaluated, and accordingly is not included in previous risk estimations of fetal CHB. In 8 fetuses in our study, we could demonstrate AV time intervals exceeding the 95% reference limits during at
LETTERS
least 2 examinations. Seven of these 8 also had SVC-Ao time intervals ⬎135 ms, as shown in Figure 1A of our article, and MV-Ao time intervals ⬎140 ms as shown in Figure 1B of our article, a comparison suggested by Rein and colleagues. Accordingly, the values in these patients also exceed the 99th percentile in normal controls. Furthermore, blocks that had not spontaneously reverted before birth were verified by electrocardiography; one baby had a third-degree block and 4 babies had a first-degree block. In total, 284 women with fetuses with gestational ages of 16–39 weeks were used to construct the reference ranges. Of these, results from 264 were previously reported (1). To ensure correct readings, another 20 control recordings were performed intermixed with the study recordings. The newly included 20 controls showed no deviation from previous controls, and the results of statistical analysis did not differ. After including these controls, the standard error of the estimate was 9.3 msec with the MV-Ao and 8.3 msec with the SVC-Ao approach, which is close to the original data. With the MV-Ao approach, the mean ⫾ SD for all observations was 115 ⫾ 9.6 msec, which compares well with the 120 ⫾ 10 ms obtained with 60 normal fetuses by Glickstein et al (2). One explanation for the 5 msec difference in mean reference values could be that we do not use exactly the same events on the Doppler tracing to represent the start of the A and Ao waves. We also want to stress the fact that our reference material, based on recordings made during ⬎4 times as many pregnancies, demonstrates a direct relationship between AV time intervals and gestational age. This appears logical, considering the age-dependence of PR intervals in both childhood and later years, and makes the reference values ⬇10 msec lower in early gestation. In our previous experience, measurements obtained with both methods (especially the MV-Ao approach) were also inversely related to heart rate (1). Our interpretation was that this difference in heart rate dependence explained why the confidence interval with the SVC-Ao approach was smaller than with the MV-Ao method. Individual raw data from all examinations of all our cases are plotted in Figures 1 and 2 of our original article. Looking at Figure 2, it can be seen that most observations were within the normal range, but with a clear tendency toward lower values. To rule out the possibility that this slight reduction in heart rate was the reason our cases had prolonged AV time intervals, we also plotted all of our observations of the cases on the regressions between AV time intervals and heart rate obtained from controls (Figure 3 of our article). In the statistical analysis Ro 52–positive women were represented by average values of repeated examinations within 3 gestational age periods (17–19, 20–22, and 23–25 weeks). The mean ⫾ SD and significance levels for these average values are presented in Table 2. We now have recalculated the data and found exactly the same numbers and significance levels as reported within the original article. Only observations made on normal pregnancies during the same period of gestation served as controls in this analysis. The report by Andelfinger et al (1) does not include data on feasibility, and the study was not conducted in a way to permit that kind of evaluation. Performing such an analysis from the number of patients in our Table 2 would not give an accurate estimate of the method success rate either, since all 24 patients were not recruited at 18 weeks and followed up every
367
week as initially intended. This is obvious from Figures 1 and 2 of our article, where all examinations made are presented. Another reason is that time for examination was occasionally limited. However, an AV time interval recording was obtained at every visit using at least one of our two methods. When comparing our Figures 1A and B it can be seen that our success rate with the SVC-Ao approach, especially in early gestation, was less than that with the MV-Ao method. The SVC-Ao approach is definitely more demanding technically, but in expert hands both methods have a high success rate (3). One advantage with the SVC-Ao technique is actually in cases of AV time prolongation, when identifying the intersection between the A and E wave on the MV-Ao recordings becomes difficult (Figures 4 and 5, original article). We accordingly use both methods to complement each other. The manufacturer has not been able to provide the exact specification of temporal resolution with our ultrasound system, transducer, and settings (Acuson Computed Sonography, Mountain View, CA). We have, however, not had problems with temporal resolution with the echo machine. Using a sweep speed of 100 mm/second, the resolution of the calipers is 3 msec. Making repeated measurements on consecutive Doppler waveforms, even with both methodologic and biologic variation, provides results that usually fall within a range of 10 msec. Using the average of 3 waveforms further reduces the range of error. The values obtained fit well with a variation (SD) of time delay measurements of ⬍4ms, reported in a validation study of ultrasound systems (4). There are many factors, including the ultrasound system and the Doppler mode, that will affect the temporal resolution (4). Temporal resolution for velocities of tissue on the GE system used by Rein et al (5) may, accordingly, not necessarily be the same as when recording blood flow velocities with our system. Still, Rein et al have caused us concern by raising the issue of temporal resolution, and we will further evaluate it in an ongoing methodologic study conducted at our laboratory. In conclusion, by using serial Doppler echocardiography in a selected risk population, we have demonstrated that fetuses of Ro 52–positive mothers frequently show indirect signs of first-degree heart block. The prolonged AV conduction may normalize before or after birth, but may also progress to a higher degree of block. In addition to giving new insights into the pathologic process of congenital heart block, these findings indicate that serial Doppler echocardiography is a useful instrument for surveillance of pregnancies with risk for fetal heart block. Marie Wahren-Herlenius, MD, PhD Sven-Erik Sonesson, MD, PhD Karolinska Institutet Stockholm, Sweden 1. Andelfinger G, Fouron JC, Sonesson SE, Proulx F. Reference values for time intervals between atrial and ventricular contractions of the fetal heart measured by two Doppler techniques. Am J Cardiol 2001;88:1433–6. 2. Glickstein JS, Buyon J, Friedman D. Pulsed Doppler echocardiographic assessment of the fetal PR interval. Am J Cardiol 2000;86: 236–9. 3. Fouron JC, Proulx F, Miro J, Gosselin J. Doppler and M-mode ultrasonography to time fetal atrial and ventricular contractions. Obstet Gynecol 2000;96:732–6.
368
LETTERS
4. Walker A, Olsson E, Wranne B, Ringqvist I, Ask P. Time delays in ultrasound systems can result in fallacious measurements. Ultrasound Med Biol 2002;28:259–63. 5. Rein AJ, O’Donnell CP, Geva T, Nir A, Perles Z, Hashimoto I, et al. Use of tissue velocity imaging in the diagnosis of fetal cardiac arrhythmias. Circulation 2002;106:1827–33.
measures in children with juvenile dermatomyositis. Rheumatology (Oxford) 2003;42:591–5.
DOI 10.1002/art.20891
Reply DOI 10.1002/art.20757
Childhood Myositis Assessment Scale and muscle strength testing in patients with juvenile dermatomyositis: comment on the article by Huber et al To the Editor: We read with interest the recent report by Huber et al (1) on the validation and clinical significance of the Childhood Myositis Assessment Scale (CMAS). Those authors recommend the inclusion of both CMAS and muscle strength testing in research and clinical studies. We strongly endorse these recommendations. In 2 publications from our group (2,3) we showed differences between CMAS results and muscle strength. In the first report (2) we described the differences in changes over time in a patient with juvenile dermatomyositis. Data on this patient indicated that CMAS and muscle strength have different disease activity–tracking characteristics. In the second study (3) we found that the CMAS score was predicted not by muscle strength, but by age and maximal oxygen uptake during an exercise test. These data provide physiologic evidence that the CMAS is not a valid indicator of muscle strength. We believe the data reported by Huber et al should be corrected for age. As noted above, we found a significant association between CMAS score and age (3), and the broad age range in their study population (3.1–18.8 years) might explain some of the variance in the associations they observed. It is well accepted that older children have more muscle strength, which would allow them to lose a greater percentage of muscle strength before reaching a critical threshold under which they would be unable to perform a certain item on the CMAS. Tim Takken, MSc, PhD Janjaap van der Net, PhD Paul J. M. Helders, PhD Wilhelmina Children’s Hospital University Medical Center Utrecht Utrecht, The Netherlands 1. Huber AM, Feldman BM, Rennebohm RM, Hicks JE, Lindsley CB, Perez MD, et al. Validation and clinical significance of the Childhood Myositis Assessment Scale for assessment of muscle function in the juvenile idiopathic inflammatory myopathies. Arthritis Rheum 2004;50:1595–603. 2. Van der Net J, Kamphuis SS, Helders PJ. The Childhood Myositis Assessment Scale to assess muscle function in a patient with juvenile dermatomyositis. Arthritis Rheum 2002;47:694–5. 3. Takken T, Elst E, Spermon N, Helders PJ, Prakken AB, van der Net J. The physiological and physical determinants of functional ability
To the Editor: We appreciate the interest of Dr. Takken and colleagues in our recent publication, and we also appreciate the opportunity to respond to their comments. Takken et al first comment on the importance of distinguishing between CMAS results and the assessment of muscle strength. We agree that the CMAS is not solely a measure of muscle strength, and it is for this reason that we have recommended that it be used in conjunction with more direct measures of muscle strength, such as manual muscle testing (MMT). The CMAS is intended for the assessment of overall muscle function, and as such, the construct addresses not only muscle strength, but also functional performance and endurance. However, the high correlation between CMAS scores and results of MMT that we observed in our study indicates a substantial relationship between these 2 measures. As stated in our report, we believe the 2 measures complement one another when used together in the assessment of children with idiopathic inflammatory myopathy. With regard to Takken and colleagues’ second comment, we agree that age and maximal oxygen uptake are likely to be important predictors of CMAS score. As previously reported by Hicks et al, CMAS scores correlate strongly with peak oxygen uptake during aerobic exercise (Hicks JE, Drinkard B, Summers RM, Rider LG. Decreased aerobic capacity in children with juvenile dermatomyositis. Arthritis Rheum 2002;47:118–23). However, we disagree with the statement that muscle strength is not a predictor of CMAS score. Our data showed a Spearman’s rank correlation of 0.73 between muscle strength (as measured by MMT) and the CMAS score. This represents a substantial and important relationship between the 2 measures. Finally, Takken et al indicate that our data should have been corrected for age. We agree that age influences the ability to perform some items in the CMAS, as has recently been shown by Rennebohm et al (Rennebohm RM, Jones K, Huber AM, Ballinger SH, Bowyer SL, Feldman BM, et al. Normal scores for nine maneuvers of the Childhood Myositis Assessment Scale. Arthritis Rheum 2004;51:365–70). Unfortunately, at this time there are no age-matched normal values for the CMAS. We used data from the report by Rennebohm et al to compare scores on 9 items of the CMAS in 4–9-year-old juvenile dermatomyositis patients and age- and sex-matched healthy control children. Our study showed that children with juvenile dermatomyositis who had abnormal scores on the 9-item CMAS had worse physical function, muscle strength, and global disease assessments. Future work will be needed to determine what constitutes a normal CMAS score in children of varying ages, particularly in younger children. The notion that older children have a greater reserve of muscle strength to
LETTERS
369
lose before they will develop a detectable loss in physical function is unproven. Adam M. Huber, MD IWK Health Centre and Dalhousie University Halifax, Nova Scotia, Canada Frederick W. Miller, MD, PhD Lisa G. Rider, MD National Institutes of Health Bethesda, MD
4. Tomsic M, Logar D, Grmek M, Perkovic T, Kveder T. Prevalence of Sjo ¨gren’s syndrome in Slovenia. Rheumatology (Oxford) 1999; 38:164–70. 5. Bjerrum KB, Keratoconjunctivitis sicca and primary Sjo ¨gren’s syndrome in a Danish population aged 30–60 years. Acta Ophthalmol Scand 1997;75:281–6. 6. Bowman SJ, Ibrahim GH, Holmes G, Hamburger J, Ainsworth JR. Estimating the prevalence among caucasian women of primary Sjo ¨gren’s syndrome in two general practices in Birmingham, UK. Scand J Rheumatol 2004;33:39–43. 7. Thomas E, Hay EM, Hajeer A, Silman AJ. Sjo ¨gren’s syndrome: a community-based study of prevalence and impact. Br J Rheumatol 1998;37:1069–76. 8. Vitali C, Bombardieri S, Moutsopoulos HM, Balestrieri G, Bencivelli W, Bernstein RM, et al. Preliminary criteria for the classification of Sjo ¨gren’s syndrome: results of a prospective concerted action supported by the European community. Arthritis Rheum 1993;36: 340–7.
DOI 10.1002/art.20774
Concerns about the prevalence of primary Sjo ¨gren’s syndrome: comment on the article by Theander et al To the Editor: We read with interest the recent report by Theander et al (1), which supports the need for a further followup of patients with primary Sjo ¨gren’s syndrome (primary SS) and hypocomplementemia, due to an increased risk of death in this subgroup, largely from the onset of lymphoproliferative malignancy. Reading the article, we could not overlook the statement that, using the new American–European Consensus Group criteria (AECC) (2), the prevalence of primary SS is ⬃0.5% of the adult population. The asserted prevalence is supported by a number of studies (3–5), most recently by Bowman et al (6). In their article, Theander and colleagues refer only to a study in which Thomas et al (7) estimated the prevalence of primary SS at 3.3%, with a minimum prevalence estimate of 1.8%, though they offer no explanation for how they derived the much lower prevalence estimate. Thomas et al used a simplified version of the European Study Group Criteria for SS (8), in which they excluded the rose bengal score and histopathologic study of minor salivary glands and performed a 5-minute, rather than the customary 15-minute, unstimulated salivary flow test. However, Thomas et al added antinuclear antibodies and rheumatoid factor to their test criteria. In light of these modifications, we feel a transparent application of the AECC to their results is difficult at best. Ziga Rotar, MD Sonja Praprotnik, MD, PhD Matija Tomsic, MD, PhD University Medical Center Ljubljana Ljubljana, Slovenia 1. Theander E, Manthorpe R, Jacobsson LT. Mortality and causes of death in primary Sjo ¨gren’s syndrome: a prospective cohort study. Arthritis Rheum 2004;50:1262–9. 2. Vitali C, Bombardieri S, Jonsson R, Moutsopoulos HM, Alexander EL, Carsons SE, et al. Classification criteria for Sjo ¨gren’s syndrome: a revised version of the European Criteria proposed by the American-European Concensus Group. Ann Rheum Dis 2002;61:554–8. 3. Dafni UG, Tzioufas AG, Staikos P, Skopouli FN, Moutsopoulos HM. Prevalence of Sjo ¨gren’s syndrome in a closed rural community. Ann Rheum Dis 1997;56:521–5.
DOI 10.1002/art.20892
Reply To the Editor: Rotar et al correctly point out that in our article on mortality and causes of death in primary SS, we only cited the study by Thomas et al (1) in order to support our assumption of a prevalence of primary SS as low as 0.5%, according to the AECC (2). As Rotar et al point out, the cited study was published years before the publication of the AECC, and Thomas et al did not apply the complete set of classification items in their investigation. There are a number of studies (3–7), most recently the one reported by Bowman et al (8), dealing with prevalence in primary SS in different subsets of the populations, using different methodology and criteria. As the statement on prevalence was only an introductory remark, and not the topic of our report itself, reviewing these reports would have exceeded the scope of our article. To date, the report by Bowman et al (8) is the only one explicitly using the AECC as the basis for a population-based prevalence study in SS, but that report was not yet published when we were preparing our manuscript. Bowman’s study supports the low prevalence of SS, estimated to be 0.3% in their article. The components of the AECC (2) are derived from the European criteria, published and validated in different versions since 1993 (9–11). In 1997 it was proposed that only SSA/SSB autoantibodies be used as serologic markers for SS and either these autoantibodies or pathologic biopsy be required for classification (11). Fox et al (12) calculated the prevalence of primary SS to be ⬃0.5% when using this version of the European criteria, which is very similar to the AECC. In summary, when we were writing our report, we judged the study by Thomas et al (2) as the most solid epidemiologic investigation of the prevalence of primary SS. In that study, 32 individuals had SSA and/or SSB autoantibodies (a surprisingly high number), and 6 patients had autoimmune SS. If some or all of these 6 patients had anti-SSA and/or anti-SSB antibodies, fulfilling the requirements for the AECC, this would yield a point prevalence of 0.6% at most, using conservative assumptions regarding nonresponders (6 of
370
LETTERS
1,000). On the other hand, performance of lip biopsies might have resulted in more patients fulfilling the AECC and thus a higher prevalence rate. Elke Theander, MD Lennart Jacobsson, MD, PhD Malmo ¨ University Hospital Malmo ¨, Sweden 1. Thomas E, Hay EM, Hajeer A, Silman AJ. Sjo ¨gren’s syndrome: a community-based study of the prevalence and impact. Br J Rheumatol 1998;37:1069–76. 2. Vitali C, Bombardieri S, Jonsson R, Moutsopoulos HM, Alexander EL, Carsons SE, et al. Classification criteria for Sjo ¨gren’s syndrome: a revised version of the European criteria proposed by the American-European Consensus Group. Ann Rheum Dis 2002;61:554–8. 3. Bjerrum KB. Keratoconjunctivitis sicca and primary Sjo ¨gren’s syndrome in a Danish population aged 30–60 years. Acta Ophthalmol Scand 1997;75:281–6. 4. Jacobsson LT, Axell TE, Hansen BU, Henricsson VJ, Larsson A, Lieberkind K, et al. Dry eyes or mouth: an epidemiological study in Swedish adults, with special reference to primary Sjo ¨gren’s syndrome. J Autoimmun 1989;2:521–7. 5. Tomsic M, Logar D, Grmek M, Perkovic T, Kveder T. Prevalence of Sjo ¨gren’s syndrome in Slovenia. Rheumatology (Oxford) 1999; 38:164–70. 6. Dafni UG, Tzioufas AG, Staikos P, Skopouli FN, Moutsopoulos
7.
8.
9.
10.
11.
12.
M. Prevalence of Sjo ¨gren’s syndrome in a closed rural community. Ann Rheum Dis 1997;56:521–5. Drosos AA, Andonopoulos AP, Costopoulos JS, Papadimitriou CS, Moutsopoulos HM. Prevalence of primary Sjo ¨gren’s syndrome in an elderly population. Br J Rheumatol 1988;27:123–7. Bowman SJ, Ibrahim GH, Holmes G, Hamburger J, Ainsworth JR. Estimating the prevalence among Caucasian women of primary Sjo ¨gren’s syndrome in two general practices in Birmingham, UK. Scand J Rheumatol 2004;33:39–43. Vitali C, Bombardieri S, Moutsopoulos HM, Balestrieri G, Bencivelli W, Berstein RM, et al. Preliminary criteria for the classification of Sjo ¨gren’s syndrome results of a prospective concerted action supported by the European community. Arthritis Rheum 1993;36:340–7. Vitali C, Bombardieri S, Moutsopoulos HM, Coll J, Gerli R, Hatron PY, et al. Assessment of the European classification criteria for Sjo ¨gren’s syndrome in a series of clinically defined cases: results of a prospective multicentre study. Ann Rheum Dis 1996;55:116–21. Vitali C, Bombardieri S. The European classification criteria for Sjo ¨gren’s syndrome (SS): proposal for modification of the rules for classification suggested by the analysis of the receiver operator characteristics (ROCS) curve of the criteria performance. J Rheumatol 1997;24 Suppl 50:38. Fox RI, Tornwall J, Michelsen P. Current issues in the diagnosis and treatment of Sjo ¨gren’s syndrome. Curr Opin Rheumatol 1999;11:364–71.
