Elevated serum C-reactive protein levels in osteoarthritis.

British Journal of Rheumatology 1997;36:140–149

LETTERS TO THE EDITOR

detectable by the standard turbidimetric assay, the difference between the controls and the OA patients may not be revealed by such assays. The modest rise in serum CRP levels in OA may relate to mild synovitis. However, we cannot entirely dismiss the possibility that the elderly OA patients have other intercurrent diseases which cause elevated CRP. CRP levels are regulated by cytokines, particularly interleukin-6, interleukin-1 and tumour necrosis factor-a. These cytokines are produced locally in OA joints [8–11]. It would seem likely, therefore, that it is these cytokines which stimulate CRP production in OA. How synovial inflammation is involved in the pathogenesis of OA remains to be established. In rheumatoid arthritis, CRP concentrations correlate with synovitis, joint erosions and radiographic progression [12–14]. It will be important to determine whether elevated CRP correlates with the severity of radiographic features, particular constellations of clinical symptoms or evidence of disease progression in OA. We acknowledge TOA Medical Electronics Co. Ltd, Japan, who supplied the PAMIA-30 system and all reagents. We thank Miss K. Meadows and Mr L. Shepstone for technical and statistical support.

Elevated Serum C-Reactive Protein Levels in Osteoarthritis S—Despite histological evidence of synovitis in some osteoarthritis (OA) patients [1, 2], most of the established serum markers of systemic inflammation are usually within the normal range. C-Reactive protein (CRP) is recognized as one of the most sensitive measures of inflammation [3, 4], but the standard turbidimetric assay widely used in clinical practice has a lower detection limit of 010 mg/l. However, more sensitive assays showed that the upper limit of the normal range is only 01 mg/l [5]. Consequently, serum CRP levels which are up to 10 times normal remain undetected in standard assays. The present study was thus undertaken to determine whether serum CRP levels are raised in patients with OA as judged by a sensitive assay. Sera from 167 members of a cohort [48 males and 119 females; mean (..) age 65.0 (11.8) yr] of patients with OA of the knee [6] were used. Sera from 51 healthy volunteers [18 males and 33 females; age 35.5 (9.9) yr] who had no knee pain and no history or signs of joint disease nor of recurrent infections were used as controls. Serum CRP levels were measured by an automated PAMIA-30 counting immunoassay (TOA Medical Electronics Co. Ltd, Japan) with a lower detection limit of 0.02 mg/l [7]. Data were analysed following logarithmic transformation (natural log) to allow for their skewed distribution and presented as geometric means with 95% confidence intervals (CI). Between-group differences were compared using Student’s t-test and Spearman correlation was used for testing associations with age and gender. The average intra-assay and inter-assay CVs of the PAMIA-30 counting immunoassay were 6.2 and 13.0%, respectively. Serum CRP concentrations are shown in Fig. 1. They were highly skewed in distribution, but normally distributed after logarithmic transformation. The geometric mean (95% CI) CRP concentration in patients with OA was 3.25 (2.59, 3.91) and that in normal controls was 0.93 (0.11, 1.75) mg/l (unpaired t-test; P Q 0.0001). The patients with OA were older than controls and had a higher male to female ratio. However, no significant correlations were found with age (r = 0.011, 95% CI: −0.271, 0.291 for the control group and r = 0.054, 95% CI: −0.102, 0.207 for patients with OA) or sex [0.84 (−0.36, 2.03) in men and 0.99 (−0.11, 2.09) in women (P = 0.386) for the control group]. The results thus show that serum CRP concentrations are raised in many patients with established knee OA compared to healthy subjects. Since the values obtained for CRP levels in most of the OA patients and almost all of the normal individuals fall below those

F. 1.—Serum C-reactive protein levels measured by a highly sensitive counting immunoassay (PAMIA-30) in normal subjects (controls) and patients with osteoarthritis (OA). The solid horizontal bars represent the absolute mean values in each group.

140


M. S, C. J. E,*, P. A. D, J. R. K

Rheumatology Unit, University Department of Medicine, Bristol Royal Infirmary, Bristol BS2 8HW and *Department of Microbiology and Pathology, University of Bristol, Bristol Accepted 14 June 1996 Correspondence to: M. Sharif. 1. Goldenberg DL, Egan MS, Cohen AS. Inflammatory synovitis in degenerative joint disease. J Rheumatol 1982;9:204–9. 2. Gordon GV, Villanueva C, Schumacher HR, Gohel V. Autopsy study correlating degree of osteoarthritis, synovitis and evidence of articular calcification. J Rheumatol 1984;11: 681–6. 3. Tillett WS, Francis T. Serological reactions in pneumonia with a non-protein somatic fraction of pneumococcus. J Exp Med 1930;52:561–71. 4. Kushner I. The phenomenon of acute phase response. Ann NY Acad Sci 1982;389:38–48. 5. Shine B, de Beer FC, Pepys MB. Solid phase radioimmunoassay for C-reactive protein. Clin Chim Acta 1981;117:13–23. 6. Cushnaghan J, Dieppe PA. Study of 500 patients with limb joint osteoarthritis. I. Analysis by age, sex, and distribution of symptomatic joint sites. Ann Rheum Dis 1991;50:8–13. 7. Ohtake T, Ohtsuka Y, Ishibashi M, Kano S, Iri H. Counting immunoassay for measuring C-reactive protein in serum. Jpn J Clin Chem 1992;21:259–65. 8. Firestein GS, Alvaro-Garcia JM, Maki R. Quantitative analysis of cytokine gene expression in rheumatoid arthritis. J Immunol 1990;144:3347–53. 9. Houssiau FA, Devogelaer J-P, Damme JV, de Deuxchaisnes NC, Sick JV. Interleukin-6 in synovial fluid and serum of patients with rheumatoid arthritis and other inflammatory arthritides. Arthritis Rheum 1988;31:784–8. 10. Westacott CI, Whicher JT, Barnes IC, Thompson D, Swan AJ, Dieppe PA. Synovial fluid concentration of five different cytokines in rheumatic diseases. Ann Rheum Dis 1990;49: 676–81. 11. Holt I, Cooper RG, Denton J, Meager A, Hopkins SJ. Cytokine interrelationships and their association with disease activity in arthritis. Br J Rheumatol 1992;31:725–33. 12. Amos RS, Constable TJ, Crockson RA, Crockson AP, McConkey B. Rheumatoid arthritis: relation to serum C-reactive protein and erythrocyte sedimentation rates to radiographic changes. Br Med J 1977;i:195–7. 13. Larsen A. The relationship of radiographic changes to serum acute phase proteins and rheumatoid factor in 200 patients with rheumatoid arthritis. Scand J Rheumatol 1988;17:123–9. 14. Dawes PT, Fowler PD, Clarke S, Fisher J, Lawton A, Shadforth MF. Rheumatoid arthritis: treatment which controls the C-reactive protein and erythrocyte sedimentation rate reduces radiographic progression. Br J Rheumatol 1986;25:44–9.

HLA-DRB1* Genotypes in Greek Rheumatoid Arthritis Patients: Association with Disease Characteristics, Sex and Age at Onset S—The association of certain HLA-DRB1* genotypes in Greek rheumatoid arthritis (RA) patients with disease characteristics, sex and age at onset was investigated. This study was designed because differences have been reported in the HLA-DR profile between Greek and northern European RA patients [1–4]. The study included 86 unrelated Greek RA patients (13 males, 73 females) fulfilling ACR criteria [5] and 130 controls matched to the cases by ethnic origin and age. The patients’ age ranged from 24 to 80 yr (mean 55.2 yr, .. 12.2). RA patients hospital-

141

ized (35%) or seen as out-patients (65%) in the rheumatology departments of two hospitals located in the Athens metropolitan area over a 1 yr period were asked to participate. No patient refused. HLA typing was performed by polymerase chain reaction (PCR) and SSO [6]. Mantel–Haenszel (M–H) (one-tail), Fisher’s exact and M–H for linear trend tests were used for statistical analyses [7]. The age at onset ranged from 23 to 80 yr (mean 48.9, .. 12.9); in 36.5%, it was below 45 yr. A total of 81% of patients had positive rheumatoid factor (RF) at some stage of the disease, 70% presented radiological erosions and 13% s.c. nodules. Felty’s syndrome was absent from all RA patients. Genotypes HLA-DR4 and DR10 were significantly increased among RA patients; 37.2% of these patients and 16.9% of controls were positive for DR4 (RR = 2.9, P Q 0.001); 10.5% of patients and 4.6% of controls were positive for DR10 (RR = 2.4, P = 0.045). HLA-DR1 was increased (RR = 1.79), though not significantly, after controlling by stratification on DR4 and DR10. Stratification of genotypes into DRB1* alleles showed that *0101, *0401, *0405 and *1001 are associated with a statistically significantly (P Q 0.01) elevated risk of developing RA (RR = 2.60, 13.23, 8.23 and 2.42, respectively). The *0408 allele was absent from RA patients, but present in 1.5% of controls. Alleles with the shared epitope (shared sequence of amino acids within the third hypervariable region of the HLADRB1* molecule [8, 9] SE) *0101, *0102, *0401, *0404, *0405, *0408, *1001 were present in 65% of cases vs 31.5% of controls (RR = 4.05, P Q 0.00001). A total of 47.7% of cases and 28.5% of controls carried a single SE allele, whereas two SE alleles were present in 11.6% of the cases and 2.3% of the controls (P Q 0.001). The presence of one SE allele led to a 2.85-fold increase in the risk for RA, when compared to the absence of both (SE−/SE−), whereas this increase was 8.57-fold when two SE alleles were present. Certain characteristics of a single SE+ were compared to the SE−/SE−. The largest significant effect was with *0401, followed by *0405 and *0101, with RRs of 10.3, 9.43 and 2.40, respectively. The risk of having positive RF and s.c. nodules was higher among subjects carrying two SE alleles: for RF RR = 9.13 (P Q 0.001); for nodules RR = 10, but not significant. Most significant findings (P Q 0.01) persisted after stratification by sex, despite the small number of male cases and corresponding limitations of power. Specifically, the RR for men was double the size of that for women in the single SE allele category (RRs 5.68 and 2.67, respectively) and four times higher in the presence of two SE alleles (RRs 30 and 7.74). Stratification by age at RA onset showed that men younger than 45 yr had a five times higher risk than women when they carried two SE alleles (RR = 60 vs 12, P Q 0.01). Men older than 45 yr had a risk three times higher (but not significantly so) than women (15 vs 4.74). For the over 45 age group, the increase in risk for both sexes was significant in SE+/SE+ vs SE−/SE− (P Q 0.004).

142

BRITISH JOURNAL OF RHEUMATOLOGY VOL. 36 NO. 1

The results of this study indicate that the SE sequence, particularly on both haplotypes, predisposes to seropositive RA in the Greek population, as has been observed in northern Europeans [10]. In addition, SE alleles were positively associated with susceptibility at a young age, particularly in young men. C. S, M. S, Y. K,* I. K,† P. V. K,* A. L,* E. G,* E. K*

National Tissue Typing Centre, ‘George Gennimatas’ General Hospital, Athens, *Department of Hygiene and Epidemiology, University of Athens Medical School and †‘Evangelismos’ General Hospital, Athens, Greece Accepted 25 June 1996 Correspondence to: E. Kaklamani, Department of Hygiene and Epidemiology, University of Athens Medical School, Micras Asias 75, Goudi 115 27, Athens, Greece. 1. Papasteriades CA, Kappou ID, Skopouli FN, Barla MN, Fostiropoulos GA, Moutsopoulos HM. Lack of HLA-antigen association in Greek rheumatoid arthritis patients. Rheumatology Int 1985;5:201–3. 2. Boki KA, Panayi GS, Vaughan RW, Drosos AA, Moutsopoulos HM, Lanchbury JS. HLA class II sequence polymorphisms and susceptibility to rheumatoid arthritis in Greeks. The HLA-DRb shared epitope hypothesis accounts for the disease in only a minority of Greek patients. Arthritis Rheum 1992;35: 749–55. 3. Boki KA, Drosos AA, Tzioufas AG, Lanchbury JS, Panayi GS, Moutsopoulos HM. Examination of HLA-DR4 as a severity marker for rheumatoid arthritis in Greek patients. Ann Rheum Dis 1993;52:517–9. 4. Spyropoulou-Vlachou M, Stavropoulou-Gioka A, Kappou I et al. HLA DNA polymorphism associated with rheumatoid arthritis in Greek patients. Arch Hellen Med 1994;11: 265–73. 5. Arnett FC, Edworthy SM, Bloch DA et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988;31: 315–24. 6. Tiercy JM, Jeannet M, Mach B. A new approach for the analysis of HLA class II polymorphism ‘HLA oligotyping’. Blood Rev 1990;4:9–15. 7. Breslow NE, Day NE. Statistical methods in cancer research. Vol. 1. The analysis of case-control studies. IARC Scientific Publications No. 32. Lyon: IARC, 1980. 8. Winchester R, Dwyer E, Rose S. The genetic basis of rheumatoid arthritis. The shared epitope hypothesis. Rheum Dis Clin North Am 1992;18:761–83. 9. Arnett FC. Histocompatibility typing in the rheumatic diseases. Diagnostic and prognostic implications. Rheum Dis Clin North Am 1994;20:371–90. 10. MacGregor A, Ollier W, Thomson W, Jawaheer D, Silman A. HLA-DRB1* 0401/0404 genotype and rheumatoid arthritis: increased association in men, young age at onset, and disease severity. J Rheumatol 1995;22:1032–6.

Clubbing in Patients with Human Immunodeficiency Virus Infection S—We read with great interest the paper by Boonen and colleagues describing two patients who developed clubbing after human immunodeficiency virus (HIV) seroconversion [1]. We have previously reported 38 cases of rheumatic diseases obtained from a large cohort of 1277

HIV-infected individuals who were treated in Guipuzcoa (northern Spain) between 1988 and 1992 [2]. One of them was a 28-yr-old man who was referred to us in November 1992 because of acute arthritis in both knees. Since 1991, he had noted enlargement of the tips of his fingers and toes. There was no family history of clubbing or evidence of conditions related to rheumatic diseases. Because of past i.v. drug addiction, serology for HIV was tested in 1985 with a positive result. He developed oesophageal candidiasis in 1987, disseminated tuberculosis in 1988 and visceral leishmaniasis in 1989. Splenectomy was performed in 1992. He received zidovudine between 1988 and 1991, and didanosine between 1991 and 1995. On examination, there was pronounced clubbing of fingers and toes. Both knees were swollen and warm. No abnormalities emerged in laboratory testing apart from a low number of CD4-positive lymphocytes (0.03 × 109/l). The erythrocyte sedimentation rate was 12 mm/h, the values of transaminases, uric acid and thyroxine were normal, and rheumatoid factor, antinuclear antibodies and HLA B-27 negative. Radiographs of the chest, knees and long bones did not show abnormalities, and 99m-Tc diphosphonate scintigraphy failed to show periosteal increased uptake. Right knee arthrocentesis was undertaken, but no synovial fluid was aspirated. Treatment with naproxen (1 g/day) was given with good results; the arthritis of both knees disappeared in 3 weeks. He suffered from recurrences of tuberculosis and leishmaniasis in 1995. At that time, no new changes or clubbing occurred. As Boonen et al. show in their paper [1], the majority of reported patients with HIV infection who developed hypertrophic osteoarthropathy (HOA) had pulmonary infections and there is a recognized relationship between cardiac, pulmonary and gastrointestinal infections and HOA. However, in other cases, there is no clear explanation for the development of HOA [1, 3]; therefore, the authors suggest that both HOA and clubbing might be secondary to HIV infection. In our case, chronic and recurrent infections were present before and after the development of clubbing. Furthermore, it appeared 6 yr after the recognition of HIV positivity. We suggest that recurrent concomitant infections were probably correlated with the presence of clubbing in our patient. The cause of the acute arthritis involving both knees cannot be established because synovial fluid (which usually shows a white blood cell count Q0.5 × 109/l in HOA) was not obtained. There was no evidence of periostitis (the cause of the effusions as a sympathetic reaction in patients with HOA). Symptoms and signs of arthritis disappeared within 3 weeks, and arthritis of unknown origin has been reported in patients with HIV infection [4, 5]. We conclude that clubbing and HOA must be included in the wide spectrum of rheumatic manifestations in patients with HIV infection. Because of the high number of associated infections and neoplasms, the role of HIV in the development of clubbing and HOA is difficult to determine in these patients.

143

LETTERS TO THE EDITOR J. B, C. G, M. F

Rheumatology Unit, Hospital NS Aranzazu, Apartado de Correos 477 , 20014 San Sebastian, Spain Accepted 20 June 1996 Correspondence to: J. Belzunegui. 1. Boonen A, Schrey G, Van der Linden SJ. Clubbing in human immunodeficiency virus infection. Br J Rheumatol 1996;35: 292–4. 2. Belzunegui J, Cancio J, Pego JM et al. Estudio retrospectivo de las manifestaciones reuma´ticas en pacientes infectados por el VIH en Guipu´zcoa. Rev Esp Reumatol 1994;21:274–8. 3. Harris JP. HOA and HIV. Ann Intern Med 1988;109:250. 4. Berman A, Espinoza LR, Dı´ az JD et al. Rheumatic manifestations of human immunodeficiency virus infection. Am J Med 1988;85:59–64. 5. Calabrese LH, Kelley DM, Myers A, O’Connell M, Easley K. Rheumatic symptoms and human immunodeficiency virus: the influence of clinical and laboratory variables. Arthritis Rheum 1991;34:257–63.

Reactive Arthritis in a Patient with Simultaneous Parvovirus B19 Infection and Clostridium difficile Diarrhoea S—We report a case of Clostridium difficile diarrhoea and arthritis after antibiotic treatment for a skin infection. Retrospectively, parvovirus B19 (B19) was suggested as the cause of the primary infection. A 34-yr-old woman was prescribed cefuroxime for impetigo, as diagnosed by a general practitioner. Diarrhoea and urticaria developed within a week; faecal culture of C. difficile was positive. After another week, arthralgias, myalgias and iritis developed. On admission to the hospital on 7 March 1993, arthritis of the right ankle joint and arthralgia of the wrists were observed. Walking was difficult, and the patient complained of mild dysuria. The blood haemoglobin was 103 g/l, the peripheral blood leucocyte count was normal (5.8 × 109/l), but the C-reactive protein (CRP) level was high (126 mg/l). The axial temperature was 38.7°C. The urine sample revealed leucocytosis; the bacterial culture later proved negative. Ciprofloxacin was given i.v. for 3 days for suspected pyelonephritis. Metronidazole treatment started 4 days before hospitalization was also continued for 3 days. Synovial fluid examination revealed leucocytes 3750 × 106/l with 42% mononuclear cells and 58% polymorphonuclear cells. The joint symptoms gradually subsided. However, the diarrhoea began again on 16 March, and with C. difficile positive in the stool sample, metronidazole was readministered. The diarrhoea and fever subsided in a week. Rheumatoid factor was negative. All five blood cultures taken during the hospitalization were negative. No serum antibodies to Chlamydia trachomatis, Campylobacter, Salmonella and Yersinia were detectable. The patient was discharged home symptom free on 30 March 1993. In the last control a year later, this HLA-B27-positive patient was healthy, except for occasional arthralgias of the ankle and knee joints. A blood sample drawn on 11 March 1993 revealed, retrospectively, the presence of B19 DNA sequences,

by using two B19-specific polymerase chain reaction (PCR) assays [1]. Circulating B19-specific IgM and IgG were also present in high concentrations. It seems apparent that our patient suffered initially from a B19 infection and subsequently developed C. difficile diarrhoea. The ‘impetigo’ was probably a rash caused by B19. The clinical picture of the arthritis fits with C. difficile reactive arthritis (ReA). On the other hand, a typical Reiter’s syndrome in association with B19 infection has been reported [2], and our patient had arthritis, iritis and urethritis. It is not possible to distinguish, by clinical or other criteria, whether she had a B19 arthritis or C. difficile-associated arthritis, or both. Clostridium difficile-triggered ReA seems to be associated with HLA-B27; 12 (63%) of the 19 B27-typed cases published have been B27 positive [3–10]. In several previous reports of ReA associated with C. difficile diarrhoea, the primary symptoms are typical for respiratory infections. Paty and Nichols [3] reported a pregnant woman with a flu-like illness and pruritic skin rash occurring before C. difficile-associated arthritis. Of the three patients reported by Hannonen et al. [5], one had an upper respiratory infection and one had acute tonsillitis as the primary infection. A navy sailor reported by Keating and Vyas [10] suffered primarily from sinusitis. In two other reports, C. difficile-associated arthritis was preceded by insect bites [4, 6]. The lack of microbiological diagnosis of the primary infection is common to all these cases; therefore, the possibility that the arthritis was due to the primary infection cannot be ruled out. We conclude that the aetiopathogenesis of C. difficile-associated ReA should be studied more extensively. Finding bacterial antigens or bacterial DNA in the inflamed joint would give more evidence of the true arthritogenic nature of C. difficile. It is possible that the development of ReA is only indirectly caused by C. difficile, by inducing a change in the intestinal flora, a situation resembling ReA developing after intestinal by-pass. However, even in such a case, the bacterial antigens or DNA should occur in the joint tissue. Likewise, the role of B19 and other arthritogenic microbes causing the primary infection should be taken into account. This study was supported by the Academy of Finland and the Sigrid Juse´lius foundation. S. N, U. Y-K,* P. T

Department of Medical Microbiology, Turku University, FIN-20520 , Turku and *Division of Rheumatology, Department of Medicine, Tampere University Central Hospital, FIN-36280 Pikonlinna, Finland Accepted 20 June 1996 Correspondence to: S. Nikkari. 1. Nikkari S, Torvinen A, Hannonen et al. Persistence of parvovirus B19 in synovial fluid and bone marrow. Ann Rheum Dis 1995;54:597–600. 2. Kandolf R, Kirschner P, Hofschneider PH, Vischer TL. Detection of parvovirus in a patient with ‘reactive arthritis’ by in situ hybridization. Clin Rheumatol 1989;8:398–401.

144


3. Paty JG, Nichols RE. Arthritis and non-antibiotic-associated pseudomembranous colitis. Arthritis Rheum 1987;30:1075–6. 4. Atkinson MH, McLeod BD. Reactive arthritis associated with Clostridium difficile enteritis. J Rheumatol 1988;15:520–2. 5. Hannonen P, Hakola M, Mo¨tto¨nen T, Oka M. Reactive oligoarthritis associated with Clostridium difficile colitis. Scand J Rheumatol 1989;18:57–60. 6. Cope A, Anderson J, Wilkins E. Clostridium difficile toxininduced reactive arthritis in a patient with chronic Reiter’s syndrome. Eur J Clin Microbiol Infect Dis 1992;11:40–3. 7. Putterman C, Rubinow A. Reactive arthritis associated with Clostridium difficile pseudomembranous colitis. Semin Arthritis Rheum 1993;22:420–6. 8. Sensini A, Marroni M, Bassotti G et al. Clostridium difficile-associated reactive arthritis in an HLA-B27 negative male. J Clin Gastroenterol 1993;16:354–5. 9. Boice JL. Reactive arthritis induced by Clostridium difficile. West J Med 1994;160:171–2. 10. Keating RM, Vyas AS. Reactive arthritis following Clostridium difficile colitis. West J Med 1995;162:61–3.

F. 1.

Effective Treatment of Anti Jo-1 Antibody-Positive Polymyositis with Cyclosporin S—We read with interest the letter from Tellus and Buchanan [1] on the treatment of Jo-1 antibodypositive polymyositis with cyclosporin. Currently, we are treating a patient with Jo-1 antibody-positive polymyositis, similarly in whom remission was induced eventually with prednisolone and cyclosporin. In contrast to the case described, our patient has not been satisfactorily maintained in remission on this regimen. In 1993, Mrs SM, a 31-yr-old nurse, presented 2 months postpartum with a seronegative inflammatory polyarthritis affecting the wrists, knees, ankles and small joints of the hands and feet. Examination revealed, in addition to synovitis, that she had fine inspiratory crackles at her lung bases and proximal muscle weakness. Investigations showed creatinine phosphokinase (CPK) of 7815 IU/l (normal 23– 150 IU/l) and aldolase of 296 IU/l (0–7.5 IU/l); electromyography studies confirmed acute polymyositis. Autoantibodies and rheumatoid factor were negative. Anti Jo-1 antibodies were positive. Haemoblobin was 11.2 g/dl with normal indices and plasma viscosity was normal. C-reactive protein was 75.9 mg/l (normal Q0–10 mg/l). Pulmonary function tests showed a restrictive ventilatory pattern [forced vital capacity (FVC) 2.43 l, 67% of predicted; forced expiratory volume in 1 s (FEV1) 2.07 l, 66% of predicted; ratio FEV1/FVC 85%] and reduced carbon monoxide transfer factor 4.2 mmol/kPa·min, 51% of predicted. High-resolution computerized tomography (HRCT) showed interstitial pneumonitis. Echocardiogram was normal. Figure 1 shows the response of muscle enzyme levels to treatment. She was initially treated with 3 g of i.v. methyl prednisolone; this produced an inadequate response and she was then commenced on i.v. pulses of 500 mg of cyclophosphamide in addition to prednisolone 10 mg/day. The cyclophosphamide had to be discontinued due to an allergic reaction to it. Weekly i.m. methotrexate was introduced. Her disease relapsed on this regime. Pulmonary function tests were FVC

2.83 l, FEV1 2.27 l, transfer factor 4.7 mmol/kPa·min. She was then treated with oral prednisolone 60 mg (1 mg/kg/day) and cyclosporin 150 mg/day (2.5 mg/kg/ day). To maintain disease control and facilitate prednisolone reduction, cyclosporin was gradually increased. Prednisolone was reduced by 10 mg/month to 40 mg/day, then by 5 mg/month to 20 mg/day with a plan to reduce by 2.5 mg/month thereafter. At a dosage of 17.5 mg of prednisolone and 225 mg of cyclosporin, her disease flared (CPK 1345 U/l) and so prednisolone was increased to 30 mg/day and cyclosporin up to 250 mg/day. Symptoms and CPK levels improved, and prednisolone was reduced over 6 weeks to 20 mg and then reduced by 2.5 mg/month. On 15 mg/day of prednisolone, she was asymptomatic and the CPK level was normal. At 12.5 mg/day of prednisolone, CPK started to rise, so cyclosporin was increased up to 275 mg/day and prednisolone back to 15 mg/day. After 3 months of this dosage of cyclosporin, serum creatinine levels were 30% above baseline and the dosage of cyclosporin had to be reduced, which resulted in increasing muscle enzyme levels and an increase of prednisolone dosage to 20 mg/day to control the disease. Currently, pulmonary function tests are FVC 3.19 l (90% of predicted), FEV1 2.41 (77% of predicted) and transfer factor 5.6 mmol/ kPa·min (61% of predicted). Our case report illustrates the problem of not only obtaining a remission, but maintaining it in patients with polymyositis. In addition, we present evidence that cyclosporin at least prevented continued deterioration in the associated interstitial pneumonitis, and produced significant improvement in pulmonary function tests and respiratory symptoms. We support the suggestion that cyclosporin is a useful therapeutic option in inducing remission in Jo-1 antibody-positive polymyositis. The time to remission was significantly longer than in Tellus and Buchanan’s case report, and relapses occurred, necessitating increasing doses of cyclosporin which resulted in renal dysfunction. Clearly, a formal prospective trial of cyclosporin in relapsing polymyositis is needed with


at least 1 yr follow-up after remission has been produced. J. K. D, V. E. A, M. P. L

Department of Rheumatology, St Helens Hospital, Marshalls Cross Road, St Helens, Merseyside Accepted 20 June 1996 1. Tellus MM, Buchanan RC. Effective treatment of anti-Jo 1 antibody positive polymyositis with cyclosporin. Br J Rheumatol 1995;34:1187–8.

NSAIDs and Fertility S—I read with interest the case report of Akil et al. [1] regarding a possible association between nonsteroidal anti-inflammatory (NSAID) drug use and anovulation, which the authors suggested may have resulted from the luteinized unruptured follicle syndrome (LUF). Edinburgh researchers recently monitored three ovarian cycles among three patients with inflammatory arthritis [2]. These women used unspecified NSAIDs during two cycles and discontinued the NSAIDs over the periovulatory period during the third cycle. Baseline serum luteinizing hormone (LH), folliclestimulating hormone (FSH) and oestradiol levels were taken to exclude other ovulatory distortions. Urine samples were monitored for oestrone, LH and pregnanediol as ovulation markers. From day 10 until (a) ovulation or (b) a LUF was diagnosed, daily transvaginal ultrasound scans were performed, and serum oestradiol and LH were measured. In all three cases, NSAID therapy was associated with LUF, except for the third cycle where normal ovulation occurred following drug withdrawal. Prostaglandin inhibitors are frequently prescribed for women suffering with endometriotic pain [3], although clinical trials using NSAIDs for endometriotic symptom relief have not been overly impressive [4]. A small clinical trial did, however, find naproxen more successful than placebo in relieving endometriotic pelvic pain [5]. Unlike primary dysmenorrhoea, where the pain often begins immediately before or at menses onset and is gone within 48 h, dysmenorrhoeic endometriosis pain may start several days before and persist throughout the menses and even a few days thereafter [6]. As endometriosis progresses, pain can extend over the entire luteal phase, leaving only a few pain-free days post-menstrually [6]. The dyspareunia experienced by many women with endometriosis is often more painful in the perimenstrual phase [7]. A sonographic study of uterine contractions among non-pregnant women noted that women with endometriosis experienced significantly increased uterine peristalsis during the early and midfollicular phases [8]. Therefore, it is entirely possible that women with endometriosis use NSAIDs, intermittently, at varied menstrual cycle phases, based upon their symptoms—including the periovulatory phase. Although the majority of research on endometriosis has focused on infertility over the last 60 yr, the

145

association of infertility with endometriosis remains unclear. Certainly, the anatomic distortion secondary to invasive endometriosis and related adhesions is a well-understood deterrent to fertility. However, the lower fecundity rates of endometriotic women with mild endometriosis indicates that endometriosis per se may not be a direct cause of infertility and that infertility experienced by women with mild endometriosis should be considered ‘unexplained’ [7]. The capacity of endometriotic implants to synthesize prostaglandin F (PGF) has been demonstrated [9]. Alterations in fallopian tube prostaglandin production have been noted among endometriotic women [10]. Several researchers have reported increased levels of PGs in peritoneal fluid among endometriotic women [11, 12] and in endometriotic tissue [13]. Lastly, a recent study among six healthy women found that 50 mg indomethacin three times a day over the periovulatory period can delay follicular rupture—in five of six women from 2 to 12 days—with a reduction in intrafollicular blood flow, and without any apparent effect on hormonal and menstrual status [14]. While Akil et al. [1] wish to alert their fellow rheumatologists to this potential NSAID impact among childbearing women wishing to conceive, I would further suggest that all practising clinicians, including family practitioners, gynaecologists and endocrinologists, need to be more cognizant of this phenomenon. S. T

The Endometriosis Alliance of Greater New York, Inc., PO Box 326 , Cooper Station, New York, NY 10276 -0326 , USA Accepted 16 May 1996 1. Akil M, Amos RS, Stewart P. Infertility may sometimes be associated with NSAID consumption. Br J Rheumatol 1996;35:76–8. 2. Roberts R, Smith GW, Hall C, Nuki G. Luteinised unruptured follicles in women with inflammatory arthritis taking NSAIDs. Br J Rheumatol 1995;34(Abstracts Suppl. 2):22 (Abstract 37). 3. Shaw RW. Treatment of endometriosis. Lancet 1992;340: 1267–71. 4. Kauppila A, Puolakka J, Ylikorkala O. Prostaglandin biosynthesis inhibitors and endometriosis. Prostaglandins 1979;18:665–7. 5. Kauppila A, Romberg L. Naproxen sodium in dysmenorrhea secondary to endometriosis. Obstet Gynecol 1985;65:379–83. 6. Dawood MY. Endometriosis. In: Gold JJ and Josimovich JB, eds. Gynecologic endocrinology. New York: Plenum Publishing, 1987:387–404. 7. Endometriosis. In: Am Col Obstet Gynecol Tech Bull 1993: 184:1–7. 8. Kunz G, Leyendecker G. Uterine peristalsis throughout the follicular phase of the menstrual cycle: physiological and pathophysiological aspects. Human Reprod 1995;10(Abstract Book 2):149. 9. Vernon MW, Beard JS, Graves K, Wilson EA. Classification of endometriotic implants by morphologic appearance and capacity to synthesize prostaglandin F. Fertil Steril 1986;46:801–6. 10. Nabekura H, Koike H, Ohtsuka T, Yamaguchi M, Miyakawa I, Mori N. Fallopian tube prostaglandin production with and without endometriosis. Int J Fertil 1994;39:57–63. 11. Drake TS, O’Brien WF, Ramwell PW, Metz SA. Peritoneal fluid

146


thromboxane B2 and 6-keto-prostaglandin F1a in endometriosis. Am J Obstet Gynecol 1981;140:401–4. 12. DeLeon FD, Vijayakumar R, Brown M, Rao V, Yussman MA, Schulz G. Peritoneal fluid volume, estrogen, progesterone, prostaglandin and epidermal growth factor concentrations in patients with and without endometriosis. Obstet Gynecol 1986;68:189–94. 13. Willman EA, Collins WP, Clayton SG. Studies in the involvement of prostaglandins in uterine symptomatology and pathology. Br J Obstet Gynaecol 1976;83:337–41. 14. Athanasiou S, Bourne TG, Khalid A et al. Effects of indomethacin on follicular structure, vascularity, and function over the periovulatory period in women. Fertil Steril 1996;65:556–60.

Reply S—We agree with Thylan that the possible effect of NSAIDs on fertility deserves wider appreciation and note that our report has been supported by Roberts et al. [1], in an article of which we were not aware when submitting our paper. The thrust of Thylan’s letter seems to be that NSAIDs may be responsible for some of the infertility associated with endometriosis. This may or may not be the case, but we agree that an assessment of infertility should now include obtaining a history of the use of NSAIDs and it is our practice now to counsel in advance young women who require NSAIDs and who may not yet be contemplating a pregnancy. M. A, R. S. A, P. S

Nether Edge Hospital, Osborne Road, Sheffield S11 9EL 1. Roberts R, Smith GW, Hall C, Nuki G. Luteinised unruptured follicles in women with inflammatory arthritis taking NSAIDs. Br J Rheumatol 1995;34(Abstracts Suppl. 2):22 (Abstract 37).

Re: Measuring Outcomes in Rheumatoid Arthritis—Which Measures are Suitable for Routine Use? S—I note Professor Blake’s comments with interest and agree that there is considerable scope for debate on whether we should measure outcomes and, if so, which outcomes, how we should measure outcomes and how frequently. Professor Blake rightly suggests that we may shoot ourselves in the foot if all we do is record deterioration. We do in fact already record outcomes, but poorly and in an unreproducible and usually non-quantitative way. At the crudest level, we record in the notes ‘patient better, same or worse than last time’. The problem with this method, and this is essentially I think what Professor Blake is suggesting, is that it is unreliable, unquantifiable and fails most tests of validity. I agree with him that we must capture the impact of clinical events and the effect of our therapy—the question is how can this be done without simply recording gradual deterioration over time? Part of the answer, but not the whole answer, is to use an instrument which measures health related-

TABLE I Standardized response means (SRM) for EuroQol and diseasespecific measures in patients (n = 56) reporting improvement over 3 months SRM Disease activity—doctor Pain VA scale (10 cm) EuroQol–thermometer EuroQol–utility Joint swelling Joint tenderness Disease activity–patient HAQ ESR

1.0 0.85 0.75 0.71 0.64 0.59 0.5 0.4 0.31

Reliability coefficients (RC) for each instrument in patients (n = 89) reporting no change in RA over 3 months HAQ EuroQol—thermometer EuroQol—utility VA pain scale Tender joint score Disease activity—patient assessed Swollen joint score Disease activity—physician assessed

RC

(95% CI*)

0.9 0.61 0.55 0.5 0.48 0.42 0.35 0.34

(0.8–1.0) (0.51–0.71) (0.45–0.65) (0.4–0.6) (0.38–0.58) (0.32–0.52) (0.25–0.45) (0.24–0.44)

Quality of Life (HR-QOL) [1]. This allows one to avoid the trap of comparing the patient with external criteria, against which they will inevitably fail, and allows one to measure the patient against their own standards, which adapt with age and duration of disease. The task then is to demonstrate (using a valid and reliable instrument) that over the lifetime of the patient we improve QOL when things go wrong and that the area under the QOL curve over time is maximized for as long as possible. There are candidate instruments for achieving this [2]. For example, we have undertaken a large study comparing the performance of ACR disease activity measures with two generic measures—the MOS-SF36 and EuroQol [3]—which will shortly be submitted for publication. One of our main conclusions, however, is that EuroQol, which can be used either as a very simple health profile or as a health index, is valid, more sensitive to change than most of the ACR measures and is reliable (Table I) on test–retest over 3 months. EuroQol satisfies the basic requirements of an HR-QOL instrument [1], is very simple to use and is a candidate instrument for routine clinical use—it could certainly be used on multiple and frequent occasions. In conclusion, Professor Blake is only partly right; although more work is still needed, simple instruments like EuroQol are candidates for routine use. We will shortly be trialling it in routine practice. The debate should continue. N. P. H

Western General Hospital NHS Trust, Crewe Road South, Edinburgh EH4 2XU Accepted 12 June 1996

LETTERS TO THE EDITOR 1. Calman KC. Quality of life in cancer patients—an hypothesis. J Med Ethics 1984;10:124–7. 2. Carr AJ, Thompson PW, Kirwan JR. Quality of life measures. Br J Rheumatol 1996;35:275–81. 3. Hurst NP, Kind P, Ruta D. Validity and responsiveness of EuroQol—a generic health index in rheumatoid arthritis. APLAR, Melbourne, April 1996.

Reply S—If I measure the area under a curve of a Euro-QOL, will this tell me that I can influence usefully events in a rheumatoid’s experience in the short term, that I am fully justified in seeing my patients often and that I could do with extra staff to deliver an optimal service? Or will it tell me that rheumatoid arthritis is a chronic condition, that like Dr Hurst and I, deteriorates with time—with exasperations and remissions?!! D. R. B

Bone and Joint Research Unit, St Bartholomew’s and the Royal London School of Medicine & Dentistry, Turner Street, London E1 2AD Re: Quality of Life Measures S—I enjoyed the review by Carr et al. [1]. They quite rightly draw attention to the lack of definition of the term Quality of Life (QOL) or health relatedQOL (HR-QOL). Many of the measures referred to as HR-QOL instruments in the literature, such as the MOS-SF36, include measures of ‘performance’ or ‘health status’ and may fail to measure the subject’s perception of their health. For example, mobility may be measured in relation to an external and arbitrary standard, such as ability to walk a mile or climb a flight of stairs, rather than measuring the subject’s satisfaction or adaptation to their own particular performance and environment—after all we cannot all run a marathon! Calman [2] has suggested that QOL is ‘the extent to which an individual’s hopes and expectations are matched and fulfilled by experience’; this, I would argue, fulfils our requirement for an operational definition of QOL. The definition highlights the fact that QOL is the deficit (or lack of it) between an individual’s perception of their health and how they would like it to be. For example, a disabled individual who is well adapted to using a wheelchair and living a fulfilled life may well perceive their QOL to be excellent. Conversely, clinicians will be familiar with the patient who is devastated by relatively minor (to us) impairment or disability and who perceives their QOL to be poor. With time and adaptation to their new circumstances, such individuals may again find fulfilment and an improved QOL even though the impairment and disability remain. There are instruments, such as EuroQol and SEIQoL, which to some extent satisfy this definition. I agree that the question of whose weightings should be used in such instruments is important. While the notion of using the subject’s valuations is attractive, it must not be forgotten that QOL measures are being

147

incorporated into decisions on resource allocation, and it is society as a whole which pays for the resources, not the individual subject. Thus, there is a political dimension which cannot be ignored and there are dangers in taking too subjective a view of health. May I also point out one or two errors of fact in Carr’s article? They refer to the standard gamble (SG), time trade off (TTO) and quality of life adjusted years (QALY) as being of use in cost–benefit analysis (CBA). This is not correct and may cause some confusion to the uninitiated. CBA explicitly compares the cost of treatment with the benefit of treatment measured in monetary terms rather than in units of health. SG and TTO are used to derive a ‘utility’ for use in calculating a QALY in cost–utility analysis (CUA). The EuroQol and Quality of Wellbeing scale to which they refer also provide a ‘utility’ value for CUA, and the EuroQol has a set of weightings which were obtained by using TTO in a population survey [3]. It is important to appreciate the methodological relationships between SG, TTO and QOL instruments such as EuroQol. Finally, Carr et al. have omitted to mention one crucial difference between health profiles and utilitybased measures such as health indices, TTO etc. Utility-based measures explicitly value death and perfect health as 0 and 1, respectively, providing anchor points for their scales, whereas profiles do not and are therefore limited in their application. N. P. H

Western General Hospital NHS Trust, Crewe Road South, Edinburgh EH4 2XU Accepted 12 June 1996 1. Carr AJ, Thompson PW, Kirwan JR. Quality of life measures. Br J Rheumatol 1996;35:275–81. 2. Calman KC. Quality of life in cancer patients—an hypothesis. J Med Ethics 1984;10:124–7. 3. Dolan P, Gudex C, Kind P, Williams A. A social tariff for EuroQol: results from a UK general population survey. Discussion paper 138. York: University of York, 1995.

Reply S—Dr Hurst raises some interesting issues in response to our review of Quality of Life (QoL) measures. He suggests that Calman’s [1] definition of quality of life as ‘the extent to which an individual’s hopes and expectations are matched and fulfilled by experience’ should be adopted as the operational definition of QoL. This is an interesting definition, implying, as we have argued, that quality of life should be judged by the individual rather than health professionals or society. Indeed, use of this definition would strengthen the argument for using individualized weightings when assessing QoL for clinical purposes. Where QoL assessments are incorporated into decisions about resource allocation, we acknowledge the political dimension of such decisions and the need to take a more objective overview of health and need, but would still argue that the individual’s weightings are important. The underlying assumption in arguments against using individualized weightings in

148


resource allocation seems to be that subjective views of health/illness will necessarily result in demands for more resources. We would and have [2] argued that the converse may occur. Consider Dr Hurst’s example of a disabled individual who has adapted to using a wheelchair and rates his/her quality of life as excellent. Society’s assessment of that individual’s quality of life, based on what is considered to be ‘normal’ and/ or desirable for someone of the same age, would be much lower than the individual’s. Resource allocation, based on society’s weightings, might involve the provision of resources aimed at increasing mobility, e.g. surgery, physiotherapy, extra aids, specially adapted vehicles and so on, for a patient who does not want such interventions. Anecdotal reports of the number of Zimmer frames used as clothes horses, rather than mobility aids, provide some evidence that inappropriate allocation of resources based on a desire to return patients to a society-defined norm does occur in practice. Adoption of Calman’s definition would help to clarify the classification of various health status measures, some of which are described as QoL measures. Use of such a definition would make inclusion of patient preferences and aspirations a prerequisite for QoL measures. Thus, of the existing measures discussed in the review, only the SEIQoL could claim to be a measure of general health-related QoL. The MACTAR would qualify as a measure of QoL related to physical and social function. The Disease Repercussion Profile measures patient-perceived handicap, a concept similar to the negative dimension of Calman’s definition, i.e. the disparity between an individual’s preferences or aspirations and their experience which is due to disease. In response to Dr Hurst’s point about the econometric methods of assessing QoL, we accept the criticism that use of the term cost–benefit analyses may have been misleading. We are aware of the distinctions between cost–effectiveness, cost–benefit and cost–utility analyses and issues relating to the use of health indices such as the EUROQoL in cost–utility analyses, but felt that this complex area would be better dealt with in a review of health economic analysis. Indeed, Dr Hurst’s final point highlights the complexity of this area. He lists one of the benefits which utility-based measures (including health indices) have over health profiles as being the scale anchors of death and perfect health. Whilst a number of health indices and econometric methods (e.g. the Rosser Index [3] and the Quality of Well-being Scale [4]) do include these anchors, others do not. The descriptive sections of the EUROQoL, for example, which is recommended for use in clinical trials [5], measures self-rated health status on a continuum between the worst and best imaginable health states. Death as an anchor is only introduced in the version of the instrument specifically intended for use in the valuation of different health states. In those instruments which do include death as an

anchor, the problem arises of how to deal with those health states which are rated as worse than death. Instruments either rate these as the same as death, i.e. 0, the ‘zero option’ [6], thus biasing the valuation of health states, or assign negative values. Assigning negative values results in scales that are negatively skewed and asymmetric, and for which there is disagreement about the best method of analysis [7]. The inclusion of death and perfect health as anchors for scales which are used to value different health states in economic analyses is intuitively attractive as it facilitates the mathematical manipulation of comparative data. However, where the object of measuring QoL is to identify problems in clinical practice or to assess the effectiveness of different interventions on a number of dimensions of individuals’ lives, such anchors are not necessary. In this situation, measures which reflect the multidimensionality of quality of life, such as health profiles, are preferable to those which combine its various components to produce a single score (health indices) in a way which obscures the different weights which individuals attach to those components. Thus, whilst health indices may have an advantage over health profiles in economic analyses, health profiles are more useful in clinical situations to describe the impact of disease and/or treatment. A. J. C, P. W. T, J. R. K

University of Bristol, Department of Social Medicine, Canynge Hall, Whiteladies Road, Bristol BS8 2PR 1. Calman KC. Quality of life in cancer patients—an hypothesis. J Med Ethics 1984;10:124–7. 2. Carr AJ, Thompson PW. Towards a measure of patientperceived handicap in rheumatoid arthritis. Br J Rheumatol 1994;33:378–82. 3. Rosser RM, Watts VC. The measurement of hospital output. Int J Epidemiol 1972;1:361–8. 4. Kaplan RM, Bush JW, Berry CC. Health status: Types of validity and the Index of Well-being. Health Serv Res 1976;11:478–507. 5. EuroQol Group. EQ-5D User Guide: A measure of health-related quality of life developed by the EuroQol Group. EuroQol Group, 1996. 6. Torrance GW, Zhang Y, Feeny D, Furlong W, Barr W. Multi-attribute preference functions for a comprehensive Health Status Classification System. McMaster University Technical Document (Centre for Health Economics and Policy Analysis Working Paper Series, 92-18). Canada: Hamilton, 1992. 7. Patrick DL, Cain KC, Ullmann RF, Pearlman RA. Measuring preferences for health states worse than death. Med Decision Making 1994;14:9–18.

Pulmonary Toxicity, Methotrexate and X-Rays S—I read with interest Beyeler et al.’s [1] excellent study addressing the problem of pulmonary toxicity with methotrexate. The authors conclude that routine assessment of pulmonary function is unhelpful. I was a little disappointed that they did not comment on the value of routine chest radiographs since Drs Richards and Helliwell, in recent correspondence, appear to agree with BSR guidelines which suggest that a pre-treatment chest radiograph in patients commencing methotrexate is of value [2, 3]. I am not at all convinced of this and the study of Beyeler et al.


suggests that a more selective approach is indicated. The argument in favour of pre-treatment radiographs is, I assume, that it is necessary to document existing abnormalities so that, if respiratory difficulties arise in the future, the availability of previous films aids in assessment. There is no evidence to support this contention and there is no justification for the practice of routinely obtaining baseline chest films [4]. If the BSR guidelines are applied universally, a large number of individuals with no respiratory symptoms and signs will have chest radiographs. Further, the concern that routine pre-treatment chest radiographs are being used as a substitute for thorough clinical evaluation must be a real one. I fear that this practice reflects defensive medicine rather than evidence-based medicine. P. J

Rheumatic Diseases Unit, University of Edinburgh, Western General Hospital, Crewe Road, Edinburgh EH4 2XU Accepted 3 June 1996 1. Beyeler C, Jordi B, Gerber NJ, Im Hof V. Pulmonary function in rheumatoid arthritis treated with low-dose methotrexate: a longitudinal study. Br J Rheumatol 1996;35:446–52. 2. Richards AJ. Re: Audit of shared care between hospital and general practice in the management of rheumatoid arthritis. Br J Rheumatol 1996;35:398. 3. Helliwell PS. Reply. Br J Rheumatol 1996;35:398. 4. Tape TG, Mushlin AI. The utility of routine chest radiographs. Ann Intern Med 1986;104:663–70.

Reply S—We thank Dr Jobanputra for his valuable comments. The aim of our study was to assess lung volumes and gas exchange during methotrexate (MTX) treatment and not to analyse the value of routine chest X-rays before MTX treatment. Within our study, chest X-rays were needed for the interpretation of changes in pulmonary function. However, in daily practice based on our clinical experience, we perform costly chest X-rays (before or during MTX treatment) only in case of a history of lung disease, the presence of respiratory symptoms or abnormal clinical signs. This approach demands a thorough patient history and careful clinical examination, and results in a small risk of missing a lung disease in an asymptomatic patient. C. B, N. J. G, V. I H

Department of Rheumatology and Department of Medicine, Division of Respiratory Diseases, University Hospital, 3010 Berne, Switzerland Iliopsoas Bursa Enlargement S—In the recent case report by Doctors Byrne, Rees and Williams [1], I was surprised to see that Fig. 4 has been reproduced from the paper by Meaney et al. [2] on iliopsoas bursa enlargement. This figure appears without acknowledgement. I was less surprised to see another case report on iliopsoas bursa enlargement so soon after the last one carried by your journal [3] in

149

view of its underrecognition as a cause of hip pain and inguinal swellings. A colleague of mine who recently described another case found that several references did not appear on a Medline search. The apparent underreporting of this condition may therefore reflect, in part, a failure of indexing. P. B. B. J

Queen Elizabeth Hospital for Rheumatic Disease & Rehabilitation, PO Box 1342 , Whakaue Street, Rotorua, New Zealand Accepted 29 May 1996 1. Byrne PAC, Rees JLS, Williams BD. Iliopsoas bursitis—an unusual presentation of metastatic bone disease. Br J Rheumatol 1996;35:285–8. 2. Meaney JF, Cassar-Pullicino VN, Etherington R, Ritchie DA, McCall IW, Whitehouse GH. Ilio-psoas bursa enlargement. Clin Radiol 1992;45:161–8. 3. Jones PBB, Economou G, Adams JE, Bernstein RM. Iliopsoas bursa presenting as deep vein thrombosis in rheumatoid arthritis. Br J Rheumatol 1993;32:832–4.

Reply S—Thank you for the opportunity of making the following reply. We agree that it is important to emphasize the different modes of presentation and the diagnostic difficulties associated with this condition. Our case is unusual since the bursitis was the presenting feature of metastatic carcinoma. Permission to reproduce Fig. 4 was received, but unfortunately not acknowledged. In view of the comments on missing references, we conducted our own Medline search on Ovid CD and found the following. Between 1992 and May 1996, five references were found with a textword search of ‘ilio psoas bursitis’. An extra three references were found with a textword search of ‘bursitis’. A textword search of ‘ilio psoas burs$’ (to include ‘itis’/‘a’/‘ae’) revealed an extra seven references. A subject heading search using the medical subject headings (MeSH) term ‘bursa’ (:synovial bursa) found an extra reference. Use of the subject heading ‘cyst’ (:synovial cyst) located an extra two references. (Between 1966 and May 1996, there are 23 references found as ‘ilio psoas bursitis’ (textword) and yet only three of these are MeSH indexed as ‘synovial cyst’.) We feel this example illustrates well the problems associated with Medline indexing and searching, which are not well appreciated [1]. There is a case to be made for researchers quoting the method of searching and specifying, e.g. which database, which CD Rom, which textwords, and which subject headings were used. P. A. C. B, J. L. S. R, B. D. W* East Glamorgan Hospital, Pontyprydd CF38 1AB and *University Hospital of Wales, Cardiff 1. Lowe HJ, Barnett OG. Understanding and using the medical subject headings (MeSH) vocabulary to perform literature searches. J Am Med Assoc 1994;271:1103–8.