Can baseline ultrasound results help to predict failure to achieve DAS28 remission after 1 year of tight control treatment in early RA patients?

The prognosis of patients with rheumatoid arthritis (RA) has improved considerably over the past decades, with early intensive treatment with conventional synthetic and biological disease-modifying anti-rheumatic drugs (DMARDs) [1, 2]. Early intensive treatment nowadays is according to ‘treat to target’ [3] and ‘tight control’ paradigms and strategies [4‐6]. This has been shown to beneficially alter the long-term outcome of RA patients, especially if applying these strategies early and within the so-called ‘window of opportunity’ [7‐11].

Personalised medicine in a ‘treat to target’ and ‘tight control’ setting in RA would require precise and valid instruments predicting which treatment strategy is necessary to achieve the treatment target. These instruments would enable us to decide which patients need more intensive therapy at the initiation of such therapy. Some patients failing to achieve the treatment target of ‘low-disease activity’ (LDA) or remission would have benefitted from combination therapy from the moment of diagnosis, while monotherapy with methotrexate suffices in others. Not many indisputable prognostic parameters in RA patients exist [12]. The parameters most unequivocally predicting a less favourable outcome of RA are high baseline disease activity [13], rheumatoid factor (RF) positivity [14], and peri-articular bone oedema on magnetic resonance imaging (MRI) [15] Arthritis assessed at ultrasonography (US) at baseline might also have predictive abilities since US has been shown to predict progression to RA in undifferentiated arthritis [16, 17]. Furthermore, in RA patients in remission, US results predict the occurrence of flare and erosive progression [18‐21]. Based on these results, it has been suggested that we should include joint US findings in remission criteria for RA [22, 23], or to use US for classification criteria of RA [24].

A recent study investigating US predictors for clinical remission did not show US to be predictive for clinical remission. However, in this study there was a more than 50% dropout rate, with significant differences between the tender joint count (TJC) and swollen joint count (SJC) of dropouts and patients included in the final analysis [25]. Thus, in a newly diagnosed RA population starting treatment, it remains equivocal whether baseline US results can help to predict clinical outcomes, as assessed by the Disease Activity Score assessing 28 joints (DAS28) [26‐28].

The aim of this study was to see whether a more individualised therapy from the start might be possible in newly diagnosed RA patients who will be treated to target; this is achieved by investigating whether failure to achieve DAS28 remission after 1 year is associated with baseline joint US arthritis findings when added to baseline clinical, laboratory, and radiographical predictors. Secondary aims include associations of joint US arthritis findings with the absence of the EULAR good response criterion after 1 year of therapy and with progression of radiographical joint damage. Additionally, we investigated the association between US joint inflammation at baseline and after 1 year.

This was a multicentre (n = 7) Dutch study prospectively following for 1 year a cohort of consecutively recruited, newly diagnosed RA patients (1987 ACR criteria) [29], consisting of four sub-cohorts of which two were ongoing clinical trials: the tREACH (Treatment in the Rotterdam Early Arthritis Cohort), with monitoring every 3 months and treatment escalation if the target of low disease activity was not met [30]; and the U-Act-Early trial [31], with monitoring every 4 weeks and treatment escalation as long as remission was not achieved. The third sub-cohort from which patients were included was the “Reumatologie Online Monitoring Applicatie” (ROMA), with monitoring intervals varying from 4 to 12 weeks and treatment escalation according to protocol, as long as remission was not achieved [32]. The fourth sub-cohort consisted of early RA patients from the other centres with monitoring at least every 3 months and treatment escalation according to treat to target, at the discretion of the physician, but following the tREACH protocol (cohort one) as closely as possible. Thus, all patients were treated according to a step-up ‘tight control’ and ‘treat to target’ strategy; treating physicians were blinded to results of US examination. Symptom duration was less than 1 year, and all patients were naive for treatment with DMARDs, including biologicals or glucocorticoids. Treatment strategies included at least a potent conventional synthetic DMARD, such as methotrexate and/or a biological DMARD, with or without glucocorticoids. Prednisolone monotherapy was accepted in nine patients. Exclusion criteria were contraindications for conventional synthetic DMARDs, biological therapy, or glucocorticoids, and illiteracy and personality disorders limiting participation in the study. Patients underwent clinical, laboratory, radiographical, and US joint examination at study enrolment (baseline) and at 1-year follow-up. The study was approved of by the medical ethical committee of the Erasmus MC in Rotterdam. All patients signed a consent form prior to inclusion in the study.

Results of routinely performed US at baseline show no additional prognostic information when added to clinical, laboratory, and radiographical parameters in predicting failure to achieve DAS28 remission at 1 year in early RA patients treated to target. The absence of remission was associated with high DAS28 at the time of diagnosis and RF positivity, confirming data in the literature [12‐14]. The monitoring strategy with the shortest assessment interval (4-week visits), corrected for DAS28 and RF positivity at baseline, appeared to increase the chance to achieve remission.

The absence of a predictive value of US in this study contrasts with data showing that US predicts patients with undifferentiated arthritis (UA) progressing to RA [16, 17] and RA patients in remission progressing to flare and erosion [18‐21]. Our negative study results may be explained by the earlier observation that intensive therapy diminishes the prognostic value of baseline conventional prognostic factors [8]. Our negative results are in line with earlier observations in a smaller population of RA patients not treated to target, 64% of whom received glucocorticoids prior to inclusion in the study [26]. However, in that study ‘time integrated values’ of PDUS correlated with clinical outcomes, which was not found in our study.

It has been suggested that one of the most important steps we have yet to take in US research in RA is to investigate the added value of US and clinical indices in multicentre trials for clinical and radiological outcomes of RA patients [44]. Our study has done this. Although our study has shown that the added predictive value of routinely performed US at baseline in early RA patients is absent if treated to target, US still has a place in personalised medicine for specific indications. For example, in patients for whom the widely used DAS28 is less valid, e.g. patients with concomitant fibromyalgia who will have a high tender joint count and high VAS, and also for obese patients in whom the clinical joint examination may be difficult and who may have a higher ESR without joint inflammation [45]. In these patients, US may really make a clinically relevant contribution. US might help to distinguish patients with or without joint inflammation making more personalised medicine possible. To that aim, new trials should be performed using US measures as target for treatment, or using US as a basis for treatment escalation and tapering; initiatives using US as an outcome measure have been published [46, 47]. In our opinion, the observation that more than 80% of the patients who did not have US inflammation at baseline also did not have US inflammation at follow-up deserves further investigation.

A limitation of our study was heterogeneity in the strategies and assessment intervals of the sub-cohorts from which we included patients. All patients, however, were treated according to a ‘treat to target’ and ‘tight control’ strategy’. Our patients had relative mild disease with a median six swollen joints at baseline; however, this reflects an early RA population and application of the more sensitive, newest RA classification criteria [24]. Both are aspects of normal daily clinical practice in which future US trials as suggested above should also be performed [44]. Another limitation may be the use of US equipment that now might be seen as midrange and less sensitive compared to current high-end machines. However, this also reflects routine practice. More importantly, however, the PD modality in the machines used is actually very sensitive, even better than that of professional research machines [48]. Another possible limitation of our study is that we did not scan all joints included in the DAS28; it could be argued that we missed US inflammation. However, more than 90% of the patients in our study had US inflammation at baseline, and US scoring of a limited sets of joints also well reflects RA activity [49, 50]. A last issue of critical discussion is the moderate agreement coefficients between ultrasonographers, despite organizing several trainings sessions to increase reliability. Generally, reliability in ultrasound is not a bigger problem than in other imaging modalities if performed by trained examiners [51], as was the case in our study. That the agreement coefficients are moderate is probably largely attributable to relatively low levels of detectable inflammation at US in the patients participating in the trainings sessions; ICC values are low in samples with a homogeneous distribution, i.e. with only very high or very low scores of arthritis. Since the heterogeneity of the scores in the patient population largely determines the value of reliability measures [43], we tried to select joints from the patients for the training sessions with scores ranging from 0 to 3. Apparently, we did not succeed in doing this well enough. The hypothesis that the reliability coefficients in our study underestimate reliability is corroborated by the observation that the percentage exact agreement at the patient level was 95%.

In conclusion, in an early RA population, treated according to the ‘treat to target’ paradigm, adding data from US of the joints at baseline to the work-up did not improve prediction for not achieving DAS28 remission or EULAR good response, nor the prediction of radiographical progression at 1 year.