RHEUMATOLOGY

Letters to the Editor Rheumatology 2014;53:374–376 doi:10.1093/rheumatology/ket396 Advance Access publication 27 November 2013

L ET T E R

SIR, The BASDAI and BASFI are frequently used for the assessment of disease activity and function, respectively, in patients with axial SpA [1, 2]. However, the calculation of BASDAI and BASFI scores is only defined for cases when all the questions (6 for BASDAI and 10 for BASFI) are answered. No official instructions are available on how to deal with missing items (MIs). Theoretically even a single MI precludes calculation of the score. Our aim was to investigate the extent of missing answers in BASDAI and BASFI and to select the best strategy to substitute them. Data from the Outcome in Ankylosing Spondylitis International Study (OASIS), a 12-year follow-up prevalence cohort of patients with AS, were used for this analysis [3]. BASDAI (0–10) and BASFI (0–10) questionnaires were regularly applied [1, 2]. First, the number of MIs per questionnaire was evaluated. Further analyses were restricted to questionnaires with no MIs. Varying numbers of MIs were randomly generated per questionnaire: 1, 2, 3 and 4 MIs (and additionally 5 and 6 for BASFI), which were replaced by five strategies: (i) lowest value of the possible score, i.e. value of 0; (ii) middle value of the possible score, i.e. value of 5; (iii) highest value of the possible score, i.e. value of 10; (iv) mean of the remaining items and (v) median of the remaining items. Additionally, for the BASDAI, one MI was randomly generated in one of the morning stiffness items (question 5 or 6) and replaced by the other. The scores resulting from the different imputation techniques were compared with the original scores, taking various levels of agreement into account. The levels of agreement were derived from the smallest detectable change (SDC), i.e. 1.4 (obtained by dividing the smallest detectable difference of 2 cm [4] by ˇ2 [5]). The best imputation techniques were conservatively chosen as those achieving an agreement >90% within a difference between original and imputed scores of 40.7 (half of SDC). This cut-off was chosen in analogy with the European League Against Rheumatism (EULAR) response criteria in RA: moderate response (improvement 50.6), defined as half of the good response (improvement 51.2) [6, 7]. A total of 216 patients were included [mean age 44 (S.D. 13) years, mean symptom duration 21 (S.D. 12) years, 71% males, 85% HLA-B27 positive]. Over the 12 years, 1747 BASDAI questionnaires were obtained: 1719 (98.4%) had no MI, 24 (1.4%) had one MI, 1 (0.1%) had two MIs, 1 (0.1%) had three MIs and 2 (0.1%) had five MIs. The mean BASDAI, calculated with the complete questionnaires

! The Author 2013. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: [email protected]

Downloaded from http://rheumatology.oxfordjournals.org/ at SUNY Health Science Center at Brooklyn - Medical Research Library on March 24, 2015

How to deal with missing items in BASDAI and BASFI

only, was 3.4 (S.D. 2.3). Over the 12 years, 1741 BASFI questionnaires were obtained: 1715 (98.5%) had no MI, 12 (0.7%) one MI, 1 (0.1%) had two MIs, 11 (0.6%) had four MIs and 2 (0.1%) had six MIs. The mean BASFI was 3.6 (S.D. 2.6). All patients gave their informed consent and the ethics committees from all participating hospitals approved the study. For BASDAI, the agreement between original and imputed scores was highest when one of the morning stiffness items was substituted by the other, with an agreement of 99.7% for a difference between scores 40.7 (Table 1). In practice, this is the same as averaging the remaining items in case of one MI in one of the morning stiffness questions. Further, in the imputation of BASDAI, an agreement >90% for a difference 40.7 could only be obtained in questionnaires with one MI (in any of the questions). For these, the best imputation technique was substitution by the mean of the remaining items, reaching an agreement of 94%. For a simplification, we further evaluated the agreement of the average of the remaining five items in case of any MI and it was 91%. Imputing MIs in BASFI resulted in high agreement between the original and imputed scores in up to three MIs imputed (Table 1). Similar to the BASDAI, the best imputation technique was substitution by the mean of the remaining items, which again is the same as averaging the remaining items. For a difference between original and imputed scores 40.7, an agreement of 100% was obtained for imputation of one MI, 97% for imputation of two MIs and 93% for imputation of three MIs. To our knowledge, this is the first study proposing instructions on how to handle MIs in BASDAI and BASFI based on data. Up to one MI for the BASDAI and three for the BASFI can reliably be imputed. For both the BASDAI and BASFI, we propose averaging the remaining items, as this is easy to perform and makes this substitution strategy suitable for calculation of scores on an individual level and in clinical practice, and independent of the level of disease activity or functional disability. A BASDAI of 4 is frequently used to define eligibility for biologic therapy [8]. In case of MI in the BASDAI, a score can still be calculated and contribute to guide clinical decisions. A commonly used technique to replace missing data is multiple imputation. We decided not to use this simulation-based procedure because its purpose is not to recreate the individual missing values as close as possible to the true ones, but to handle missing data in a way that results in valid statistical inference [9]. In this study we did not want to make inferences from relationships between variables, but derive imputation rules that can be followed at a questionnaire level and be of use for clinical practice. In conclusion, the BASDAI and BASFI can be reliably

15 6 2 20 27 39 6 3 1 13 16 3 2 0 11 10 2 2 0 8 8 – – – – – – – – – –

45 37 11 76 75 91a 24 20 2 57 58 15 14 1 45 45 10 12 1 36 36 – – – – – – – – – –

68 76 30 94a 92a 100a 43 45 6 85 83 28 31 3 71 70 20 24 1 59 59 – – – – – – – – – –

81 100a 55 98a 98a 100a 53 64 13 93a 91a 38 43 5 83 81 28 33 2 72 72 – – – – – – – – – –

Dif 4 1.0 Equal 24 9 7 32 39 – 14 6 3 25 28 10 3 1 19 22 8 3 1 18 19 6 2 1 16 15 5 2 1 14 13

Dif 4 1.4 91a 100a 68 100a 99a 100a 67 84 23 98a 96a 49 60 10 92a 89 38 47 4 84 84 – – – – – – – – – – 54 52 23 90a 90a – 35 32 8 81 79 28 23 4 70 67 22 17 3 62 60 19 13 2 56 49 17 11 2 53 47

84 100a 54 100a 99a – 55 64 22 97a 96a 43 47 10 93a 90a 35 37 6 88 84 31 31 4 81 75 27 25 3 76 72

Dif 4 0.7

100a 100a 100a 100a 100a – 70 100a 36 100a 99a 52 63 19 97a 96a 43 51 11 96a 93a 36 43 7 91a 86 32 35 5 87 84

Dif 4 1.0

Dif 4 0.3

Dif 4 0.7

Dif 4 0.3

100a 100a 100a 100a 100a – 85 100a 55 100a 100a 66 89 31 99a 99a 54 67 19 99a 97a 45 58 13 97a 94a 40 48 9 94a 92a

Dif 4 j 1.4

Agreement of >90%, which was the cut-off conservatively chosen by the authors as acceptable for reliable imputation. Dif: difference; Value 0, 5, 10: means that the values 0, 5 or 10 were the values imputed in case of a missing value, respectively. aLevel of agreement: equal:original and imputed scores identical; dif 4 0.3: a difference of a maximum of 0.3 between the original and imputed scores; dif 4 0.7: a difference of a maximum of 0.7 between the original and imputed scores; dif 4 1.0: a difference of a maximum of 1.0 between the original and imputed scores; dif 4 1.4: a difference of a maximum of 1.4 between the original and imputed scores. bMorning stiffness imputed: a missing value was generated for question 5 or 6 of the BASDAI (i.e. morning stiffness) and the value of the remaining (i.e. non-missing) item was imputed.

a

6 missing

5 missing

4 missing

3 missing

Value 0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items Morning stiffness imputedb Value 0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items Value 0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items Value0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items Value 0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items Value 0 Value 5 Value 10 Mean remaining questionnaire items Median remaining questionnaire items

Equal

Downloaded from http://rheumatology.oxfordjournals.org/ at SUNY Health Science Center at Brooklyn - Medical Research Library on March 24, 2015

www.rheumatology.oxfordjournals.org

2 missing

1 missing

Imputation technique

Level of agreement for BASFI (%)a

Level of agreement for BASDAI (%)a

TABLE 1 Percentage of agreement between the original and imputed BASDAI and BASFI scores

Letters to the Editor

375

Letters to the Editor

calculated by averaging the remaining items in case of a maximum of one or three MIs, respectively. Rheumatology key message .

Up to one missing item for BASDAI and three for BASFI can reliably be imputed when assessing patients with AS.

S.R. was supported by Fundac¸a˜o para a Cieˆncia e Tecnologia (FCT) grant SFRH/BD/68684/2010. Disclosure statement: The authors have declared no conflicts of interest.

Sofia Ramiro1,2, Astrid van Tubergen3,4, De´sire´e van der Heijde5, Filip van den Bosch6, Maxime Dougados7 and Robert Landewe´1,8

6 van der Heijde DM, van’t Hof M, van Riel PL et al. Development of a disease activity score based on judgment in clinical practice by rheumatologists. J Rheumatol 1993;20:579–81. 7 van Gestel AM, Prevoo ML, van’t Hof MA et al. Development and validation of the European League Against Rheumatism response criteria for rheumatoid arthritis. Comparison with the preliminary American College of Rheumatology and the World Health Organization/International League Against Rheumatism Criteria. Arthritis Rheum 1996;39:34–40. 8 Zochling J, van der Heijde D, Burgos-Vargas R et al. ASAS/EULAR recommendations for the management of ankylosing spondylitis. Ann Rheum Dis 2006;65:442–52. 9 Rubin DB. Multiple imputation for nonresponse in surveys. New York: Wiley.

1

Department of Clinical Immunology and Rheumatology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands, 2Department of Rheumatology, Hospital Garcia de Orta, Almada, Portugal, 3Department of Medicine, Division of Rheumatology, Maastricht University Medical Center, 4School for Public Health and Primary Care (CAPHRI), University of Maastricht, Maastricht, 5Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands, 6 Department of Rheumatology, Ghent University Hospital, Ghent, Belgium, 7Rheumatology B Department, Cochin Hospital, Paris-Descartes University, Paris, France and 8Department of Rheumatology, Atrium Medical Center, Heerlen, The Netherlands. Accepted 18 October 2013. Correspondence to: Sofia Ramiro, Department of Clinical Immunology and Rheumatology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands. E-mail: [email protected]

References 1 Calin A, Garrett S, Whitelock H et al. A new approach to defining functional ability in ankylosing spondylitis: the development of the bath ankylosing spondylitis functional index. J Rheumatol 1994; 21:2281–5. 2 Garrett S, Jenkinson T, Kennedy LG et al. A new approach to defining disease status in ankylosing spondylitis: the bath ankylosing spondylitis disease activity index. J Rheumatol 1994;21: 2286–91. 3 Spoorenberg A, van der Heijde D, de Klerk E et al. Relative value of erythrocyte sedimentation rate and C-reactive protein in assessment of disease activity in ankylosing spondylitis. J Rheumatol 1999;26:980–4. 4 Auleley GR, Benbouazza K, Spoorenberg A et al. Evaluation of the smallest detectable difference in outcome or process variables in ankylosing spondylitis. Arthritis Rheum 2002;47:582–7.

376

Rheumatology 2014;53:376–377 doi:10.1093/rheumatology/ket386 Advance Access publication 4 December 2013

Further confirmation that digital ulcers are associated with the severity of abnormality on nailfold capillaroscopy in patients with systemic sclerosis SIR, In a recent pilot study in a cohort of patients with SSc, Smith et al. [1] demonstrated an association between nailfold videocapillaroscopy (NVC) patterns and the severity of peripheral vascular involvement, as assessed after 18–24 months. These findings lend further support to the proposal that the severity of abnormality on NVC may serve as a useful biomarker for the prediction of digital ulceration in SSc [2, 3]. As part of a prospective, singlecentre study examining the prevalence of SSc-related digital ulcers (DUs) [4], we used quantitative NVC to assess whether microvascular abnormalities were associated with the presence of current DUs. The North West Greater Manchester National Research Ethics Service Committee approved the study. As previously reported [4], patients attending specialist SSc clinics between January and December 2009 were invited to participate and, following informed consent, each was assessed by a specialist nurse who documented the presence or absence of active DUs. An active DU was defined as a distinct lesion with loss of epidermis. Data on point prevalence and physical function are reported elsewhere [4]. Following acclimatization in a temperature-controlled laboratory, patients underwent quantitative NVC of the non-dominant ring finger as previously described [5]. In brief, panoramic, high-magnification (300) images of the distal nailfold capillary row were acquired. A vascular technician manually marked up the

www.rheumatology.oxfordjournals.org

Downloaded from http://rheumatology.oxfordjournals.org/ at SUNY Health Science Center at Brooklyn - Medical Research Library on March 24, 2015

Acknowledgements

5 Bruynesteyn K, Boers M, Kostense P et al. Deciding on progression of joint damage in paired films of individual patients: smallest detectable difference or change. Ann Rheum Dis 2005;64:179–82.