RHEUMATOLOGY
Letters to the Editor
Rheumatology 2010;49:1200–1201
doi:10.1093/rheumatology/kep445
Advance Access publication 27 January 2010
Do snapshot statistics fool us in MTX
pharmacogenetic studies in arthritis research?
SIR, Recently, an interesting discussion was published in
Rheumatology on discrepant literature results concerning
MTX pharmacogenetics in RA [1–3]. This discussion was
triggered by the paper of Lee et al. [2] introducing the
concept of false-positive report probability (FPRP) in
the field of arthritis research. The discussion focused
on the discrepant results observed for single nucleotide
polymorphisms (SNPs) in the ATIC gene (rs4673993 and
rs2372536, both in linkage disequilibrium): the 347 C-allele
[4, 5] and the G-allele [2, 6] were both associated
with increased efficacy of MTX. Similar discrepancies
for SNPs in the methylenetetrahydrofolate reductase
(MTHFR) gene were reported in a meta-analysis earlier
this year [7]. In trying to explain the discrepancy,
Dervieux [1] pointed out the challenges and difficulties
that researchers face when validating associations
between low-penetrance genetic polymorphisms and
complex phenotypes such as drug response. The discussion focused on differences between studies in the FPRP,
differences in sample size or power, demographic dissimilarities among cohorts, environmental factors such as
folate status, duration of disease and treatment duration.
We would like to argue that one of the most important
reasons for discrepant studies is because of crosssectional analysis, also called the snapshot approach.
Most pharmacogenetic studies examine only one time
point during (MTX) treatment. For instance, MTX response
was assessed at 6 months in the European studies [4, 5]
and after 50 months in the US cohorts [2, 6]. The snapshot
approach suffers from several methodological flaws. First,
the snapshot approach may not reflect the true response
characteristics over the whole treatment phase. To illustrate this, we have plotted the typical treatment response
patterns of patients with juvenile idiopathic arthritis (JIA;
Fig. 1). From Figure 1 it becomes clear that treatment
response can be roughly divided into three profiles: (A)
patients who will respond to treatment at any time point
between start of treatment and 1-year follow-up and will
stay in remission (47%); (B) patients who shift back and
forth from responder to non-responder (31%); and (C),
patients who do not show any response during the first
year of treatment (22%). This study was performed in the
University Medical Centre Utrecht (UMCU), Wilhelmina
Children’s Hospital, The Netherlands. Patients with a
LETTERS
Fig. 1 Responders and non-responders in 183 JIA patients following Paediatric American College of Rheumatology
30% (ACRped30) criteria in 3-month intervals up to the most recent visit after start of treatment with MTX.
Response is divided into three profiles: (A) patients who will respond towards treatment at any time point between start of
treatment and 1-year follow-up and will stay in remission (47%); (B) patients who shift back and forth from responder to
non-responder (31%); and (C) patients who do not show any response during first year of treatment (22%). t3, t6, t9,
t12 ¼ time points 3, 6, 9 and 12 months, respectively, after start of MTX treatment; mrv ¼ most recent visit.
! The Author 2010. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: [email protected]
Letters to the Editor
confirmed JIA diagnosis according to the ILAR criteria
were included. All included patients had started MTX therapy between 1990 and 2006. All patients gave their
informed
consent.
The
study
was
approved
by the Medical Ethics Committee of the UMCU. Patients
had been systematically followed every 3 months using
a standardized report form on disease activity. Similar
profiles were observed in adult RA patients. From a clinical point of view, prediction of treatment response at only
one time point (e.g. 6 months) is less informative because,
at the next hospital visit, a substantial number of patients
may become non-responders and vice-versa. Second, the
snapshot approach only evaluates patients that are still
available at the analysed time point and hence, ignoring
dropouts or missing data. Often, missing data are not
missing completely at random (MCAR) and could be
related to the primary outcome, i.e. toxicity or intolerance.
As a consequence, the estimators will be biased for the
investigated SNP on treatment response.
Assessing the FPRP in snapshot approach pharmacogenetic studies may be helpful in detecting spurious findings. However, future pharmacogenetic studies in arthritis
research should preferably evaluate the treatment
response in a longitudinal way. Longitudinal analysis will
allow us (i) to better characterize the different response
profiles of patients (Fig. 1) and (ii) to perform sophisticated
repeated measurement statistics that are not affected by
the disadvantages of snapshot statistics. This method
allows estimating the occurrence of response for a
group as a whole over a certain period of time. This
approach will generate clinically more relevant information
because it will predict the long-term response characteristics of patients better and will reduce the risk of falsepositive and -negative findings.
Rheumatology key message
.
Longitudinal designs and repeated measures
statistics vs cross-sectional analysis prevent falsepositive findings in MTX pharmacogenetics.
Acknowledgements
Funding: This study was financially supported by the
Dutch Arthritis Association (nr. 06-02-402 to R. de J.
and nr. 07-01-402 to N.M.W.).
Disclosure statement: The authors have declared no
conflicts of interest.
Maurits C. F. J. de Rotte1, Jolanda J. Luime2,
Maja Bulatovic3, Johanna M. W. Hazes2,
Nico M. Wulffraat3 and Robert de Jonge1
1
Department of Clinical Chemistry, 2Department of
Rheumatology, Erasmus University Medical Center,
Rotterdam and 3Department of Pediatric Immunology,
University Medical Center Utrecht, Wilhelmina Children’s
Hospital, Utrecht, The Netherlands
Accepted 26 November 2009
Correspondence to: Robert de Jonge, Erasmus MC – Clinical
www.rheumatology.oxfordjournals.org
Chemistry, Dr. Molewaterplein 40, Rotterdam 3015 GD,
The Netherlands. E-mail: [email protected]
References
1 Dervieux T. Methotrexate pharmacogenomics in rheumatoid arthritis: introducing false-positive report probability.
Rheumatology 2009;48:597–8.
2 Lee YC, Cui J, Costenbader KH, Shadick NA,
Weinblatt ME, Karlson EW. Investigation of candidate
polymorphisms and disease activity in rheumatoid
arthritis patients on methotrexate. Rheumatology 2009;48:
613–7.
3 Kooloos WM, Guchelaar HJ, Huizinga TW, Wessels JA.
Comment on: investigation of candidate polymorphisms
and disease activity in rheumatoid arthritis patients on
methotrexate. Rheumatology 2009;48:1176–7, author
reply 7.
4 Wessels JA, Kooloos WM, De Jonge R et al. Relationship
between genetic variants in the adenosine pathway and
outcome of methotrexate treatment in patients with recentonset rheumatoid arthritis. Arthritis Rheum 2006;54:2830–9.
5 Wessels JA, van der Kooij SM, le Cessie S et al. A clinical
pharmacogenetic model to predict the efficacy of methotrexate monotherapy in recent-onset rheumatoid arthritis.
Arthritis Rheum 2007;56:1765–75.
6 Dervieux T, Furst D, Lein DO et al. Pharmacogenetic and
metabolite measurements are associated with clinical
status in patients with rheumatoid arthritis treated with
methotrexate: results of a multicentred cross sectional
observational study. Ann Rheum Dis 2005;64:1180–5.
7 Fisher MC, Cronstein BN. Metaanalysis of methylenetetrahydrofolate reductase (MTHFR) polymorphisms affecting methotrexate toxicity. J Rheumatol 2009;36:539–45.
Rheumatology 2010;49:1201–1203
doi:10.1093/rheumatology/keq036
Advance Access publication 24 February 2010
Is a 12-week trial sufficient to evaluate clinical
responses to etanercept or MTX treatment in
early RA?
SIR, Treatment of RA during the early phase of disease
appears to produce better control than treatment that is
delayed until disease is more advanced [1–4]. Although
clinical response to TNF inhibitors or MTX often occurs
within the first days or weeks of treatment [5], some RA
patients have a more delayed response to these agents
[6]. Because decisions to continue or change therapies
may be made at 12 weeks (as described in the TICORA
[7] and BeSt [8] trials), it is unknown whether, and to what
extent, some patients would respond if allowed to continue for a longer period of treatment. This question is
particularly relevant since high-level responses, such as
ACR50 and ACR70, may take up to 24–30 weeks to peak.
To determine whether more patients would respond
during a treatment trial longer than 12 weeks, we evaluated ACR20, ACR50 and ACR70 responses and disease
activity scores using 28 joints (DAS-28) at Week 26 in
patients with early RA (ERA) who did not respond to
1201