Objective Initiation of DMARD-therapy in the ‘window of opportunity’ is thought to result in a more effective modification of the processes underlying rheumatoid arthritis (RA). We questioned whether this effect is true or hyped and performed a systematic literature review.
Methods Medical literature databases up to June 2012 were systematically reviewed for cohort studies and randomised controlled trials reporting outcome data of early RA in relation with symptom duration at treatment initiation. The quality of these studies was assessed by two independent reviewers using a criteria scoring system of 15 items. Studies were dichotomised with the median score (79%) as cut-off. Best-evidence synthesis was applied to determine the level of evidence per outcome category. A meta-analysis was performed on the studies reporting on achieving DMARD-free sustained remission (the reverse of disease persistency).
Results Out of 836 screened articles, 18 fulfilled the selection criteria and were not duplicates. Ten were scored as high quality. Remission (various definitions) and radiographic progression were frequently studied outcomes. There was strong evidence for an association between symptom duration and radiographic progression. A meta-analysis on datasets evaluating DMARD-free sustained remission showed that symptom duration was independently associated with such remission; HR 0.989 (95% CI 0.983 to 0.995) per week increase in symptom duration. A moderate level of evidence was observed for other remission outcomes.
Conclusions Even when heterogeneity of patients is taken into account, prolonged symptom duration is associated with radiographic progression and a lower chance on DMARD-free sustained remission. These data may support the presence of a ‘window of opportunity’.
- Early Rheumatoid Arthritis
- Outcomes research
- Qualitative research
Statistics from Altmetric.com
Delay in initiating treatment after the diagnosis rheumatoid arthritis (RA) has been established is associated with progression of joint damage.1 Recently it has been suggested that it is important to initiate treatment even earlier, in the so-called therapeutic ‘window of opportunity’.2 Although not yet fully understood, this window is said to represent a very early phase in the disease in which therapeutic disease modification is more successful, presumably because of not fully matured underlying disease processes. The thought is that intervention in this period may hamper disease progression in such a way that chronicity is reduced. In case such a period is present, this has important consequences for the care of RA, as using this period will reduce the burden of the disease and may, for instance, reduce the number of patients who ultimately require biologics. If a period in which the disease is more susceptible to disease modifying drugs truly exists, efforts resulting in identification of a very early stage of RA may also be cost effective.
However, the question arises; does it really exist? One reason for advocating an absence of this particular window is citation bias, assuming that positive studies are preferentially cited and those with negative results neglected. Since many studies have included symptom duration as covariate in their analysis, it is relevant to systematically review all such studies in order to assess whether short symptom duration at treatment initiation is associated with less progression of the disease.
It has also been suggested that the window of opportunity encompasses the first 12 weeks after symptom onset.3 ,4 A difficulty here is that the definition of symptom onset is highly variable between studies5; it can relate to the start of symptoms or the start of swelling, and be self-reported by patients or recorded by physicians. Such differences yield incompatibility between studies with regard to the timeframe that was assessed, and preclude performing an extensive meta-analysis. We therefore aimed to perform a qualitative systematic literature review, focusing on two outcomes of early RA; achieving remission and the severity of radiographic joint damage progression.
Identification of studies
To identify studies investigating the relationship between symptom duration and the outcome of RA, we searched with the assistance of a medical librarian (JS) in medical literature databases (MEDLINE (OVID-version), PubMed, Embase (OVID_version), Web of Science, CochraneLibrary, CINAHL, Academic Search Premier and ScienceDirect) up to June 2012. Central terms in our analysis were ‘RA’ and ‘symptom duration’, for a detailed overview of the search per database, please see online supplementary file 2. Additional articles were searched in the reference lists of identified articles or via expert opinion.
Inclusion and exclusion criteria
Selection of titles, abstract and articles was performed independently by two reviewers (JABvN and AHMvdH-vM). In case of disagreement, consensus was reached after discussion. First, all retrieved titles were screened, subsequently, abstracts were retrieved for detailed review and, finally, full text articles were read and screened on our inclusion/exclusion criteria.
Included studies fulfilled the following criteria: (1) patients with RA were studied, (2) patients were prescribed DMARDs, allowing to evaluate the time period before treatment initiation, (3) patients had early disease; symptom duration (period between symptom onset and start of treatment) was recorded and <2 years, (4) patients were followed prospectively and follow-up was ≥1 year and (5) symptom duration was part of the analyses on disease outcome.
Animal studies, studies with patients <18 years, reviews, (conference) abstracts, letters to the editor, case reports, case series and studies in languages other than English were excluded.
A standardised form was used to extract information about the following data: (1) study population (population size, setting and time period of the study, symptom duration, age and gender), (2) follow-up period, (3) outcomes (joint damage progression, remission or other outcomes), (4) effect estimates.
Methodological quality assessment
The quality of each included paper was reviewed by two reviewers independently (JABvN and AHMvdHvM) using a maximum of 15 criteria (see online supplementary table S1), which were based on previous systematic reviews in the field of musculoskeletal disorders.6 ,7 The criteria were adapted for our research question of symptom duration. When a criterion was met in the article, a ‘1' was given, otherwise a ‘0’. A ‘0’ was also given when no information was given about the specific criterion mentioned in the article. In case of differences in ranking, a consensus was agreed after discussion. The maximum score (100%) for each study was based only on the items that were applicable for that study design (randomised controlled trial with radiographic outcome 15, and 14 without, observational cohort with radiographic outcome 14, and 13 without). Total scores per study were calculated as the percentage of maximum obtainable scores.
Rating level of evidence
Because the studies obtained were heterogeneous with regard to the reported effect estimates, a pooled-effect estimate could not be calculated (except for a meta-analysis on the outcome achieving DMARD-free sustained remission). Therefore, we performed a best-evidence synthesis based on the guidelines on systematic review of the Cochrane Collaboration Back Review Group.8 This is a method to summarise evidence in observational studies if the population, outcomes and data analyses are heterogenic. It consists of five levels of evidence (table 1). A study was considered to be of high quality (HQ) if the total quality score was ≥79% (which is the median of the quality scores obtained in this study).
Three studies reported on the same outcome (achieving a DMARD-free sustained remission) and were homogeneous in the definition of symptom onset (by the patient reported start of symptoms).9–11 The data of these three datasets were evaluated with univariable and multivariable Cox regression analyses. An inverse weighted meta-analysis on these datasets was performed; the effect estimate was the HR. A fixed-effect model was used. The follow-up durations of these datasets, as reported in the articles, were 10 years9 ,10 and 5 years.11 STATA V.12 was used.
Selection and inclusion of articles
In total, 1625 titles were identified; after removing duplicate references, 836 unique references were left for screening (see figure 1). After detailed review, 22 articles satisfied the inclusion and exclusion criteria.9–30 Four of these were excluded, since they concerned duplicate patient populations (two were based on the Leiden Early Arthritis Clinic,20 ,29 one publication on the BeSt trial19 and one publication on the FINRACo trial26). Consequently, in total, 18 articles were used for further analyses (see table 2).
Methodological quality assessment
The two reviewers scored 249 items in total and agreed on 234 items (94%, table 2), the 16 disagreements were resolved in consensus (see online supplementary table S1). The median quality score was 79% (mean 72.7%, range 43–100%). Consequently, articles with scores ≥79% were ranked as HQ. Low-quality (LQ) studies frequently missed points on the following items: description of the source population and information on the accurateness of determined outcomes. Additionally, few studies assessed the presence and consequences of lost to follow-up.9 ,10 ,16 ,21 ,22 ,30 Finally, of all included studies, information on the definition of symptom onset was provided in only 28%.9 ,10 ,16 ,21 ,30
The characteristics of the 18 included studies are shown in table 3. The majority of the patients were female and, generally, patients were aged >50 years. In all studies, RA was classified according to the 1987 American College of Rheumatology (ACR) classification criteria, except for one study that used the diagnosis of the rheumatologist.22 The number of patients included varied between 40 and 895. Frequencies of symptom duration were reported in different ways; mean±SD, median (IQR) or percentages (table 3). Various outcome measured were assessed; the majority related to radiographic joint damage or remission, though actual measures were variable. Progression of radiographic joint destruction was measured using the Larsen score (n=4), the Sharp–van der Heijde score (SHS) (n=4), the Sharp score (n=1) or the presence/absence of erosiveness (n=1). Remission was defined as a state after treatment (eg, Disease activity score (DAS)-28<1.6, DAS-28<2.6, DAS-44<1.6, ACR-remission (n=1, n=2, n=1 and n=2, respectively)), or as the resolution of disease persistency, which is DMARD-free sustained remission (n=3). This latter outcome was defined as sustained absence of synovitis for at least 1 year after cessation of DMARD-therapy, and of all outcomes studies the closest approximation of ‘cure’. Three studies had included analyses on changes in functional ability using the health assessment questionnaire (HAQ).
Radiographic joint damage in RA
Five HQ studies evaluated joint damage as outcome. Four of these (75%) demonstrated that shorter symptom duration was associated with less radiographic progression over time.16 ,21 ,22 ,24 This implies that the evidence on the association between symptom duration and joint damage progression is strong. One HQ study showed the opposite, this study included a severe subset of patients with RA (>6 swollen joints, >9 tender joints, 90% anti-citrullinated protein antibody (ACPA) positivity) and compared patients with RA with symptom duration of <5 and >5 months at treatment initiation; the patients with RA who started >5 months of symptoms had a better outcome.30 Three of the four HQ studies that reported a beneficial effect of early initiation of treatment had divided patients in groups with symptoms of <3 and >3 months, and observed that patients with symptoms <3 months at start of treatment developed less severe joint damage. Lukas et al22 reported a mean SHS progression of 0.8 vs 1.7 units after 1 year, respectively (p=0.033), Nell et al24 showed a mean Larson progression of 3.6 vs 14.7 after 3 years (p<0.05) and Van der Linden et al21 reported a 1.3-fold higher rate of joint destruction during 6 years follow-up in patients with RA with a symptom duration of >3 months (p<0.001). In three of the six LQ studies,12 ,13 ,27 a significant association was also observed between shorter symptom duration and less radiographic progression. Two LQ studies showed the same trend in the data, but statistical significance was not achieved (table 3).18 ,25
DMARD-free sustained remission in RA
Three HQ studies assessed DMARD-free sustained remission in patients with RA. With the permission of the authors, the raw data were used to perform a meta-analysis.9–11 First, a meta-analyses of the univariable association between symptom duration and achieving DMARD-free sustained remission was performed, showing a significant beneficial effect of a shorter symptom duration (HR on DMARD-free sustained remission 0.990 (95% CI 0.984 to 0.996) per week increase in symptom duration figure 2A). Then analyses were adjusted for age, gender and applied treatment, as these variables possibly modify the effect. This revealed an almost unchanged association between symptom duration and DMARD-free sustained remission (figure 2B). Certain levels of inflammation markers or auto-antibodies are a reflection of disease severity, and are also associated with the chance of achieving DMARD-free sustained remission. Therefore, analyses were finally adjusted for age, gender, treatment, erythrocyte sedimentation rate (ESR) and presence of rheumatoid factor (RF) (anti-cyclic citrullinated peptide (CCP) was not available in all cohorts).
The adjustment factors in the latter analyses allowed differentiating between effects of patient characteristics and of symptom duration on DMARD-free sustained remission. Also, this analysis revealed that symptom duration was independently associated with DMARD-free sustained remission; each week increase in symptom duration decreased the chance of achieving DMARD-free sustained remission (HR 0.989, 95% CI 0.982 to 0.995, p<0.001, figure 2C). Assuming a linear correlation with time in the early disease stage, the HR for achieving DMARD-free sustained remission in case of 12 weeks symptom duration at treatment initiation is 0.88.
Other outcomes in RA
Eight studies reported other outcomes which were collected during DMARD-therapy. Five of these eight studies showed a significant association between shorter symptom duration and increased frequency of remission.12 ,14 ,17 ,24 Of the four HQ studies, one study did not observe a significant association with symptom duration,30 two studies found a signification association,17 ,24 and one study assessed two cohorts of differently treated patients, showing a significant association in one but not in the other patient group.23 Based on these findings, the evidence for an association between symptom duration and these remission outcomes in RA is moderate. Three of the eight studies had also assessed changes in HAQ scores. Two studies (1 HQ and 1 LQ) reported that shorter symptom duration was associated with a higher improvement in HAQ scores.24 ,28
At present, early identification of RA and early start of DMARD therapy have become important targets in the treatment of RA. Despite the current focus on the early phases of the disease, we questioned whether very early intervention with the disease is indeed more effective with regard to modifying the disease outcome, or that positive findings are overemphasised in current thinking. We performed a systematic literature review to address this question. We identified strong evidence for an association between shorter symptom duration and less severe radiographic joint damage in patients with early RA. Furthermore a meta-analysis on the outcome DMARD-free sustained remission demonstrated that prolonged symptom duration is independently associated with a decreased chance on this remission; in case of symptom duration of 12 weeks (the suggested period of the window of opportunity) at treatment initiation the HR on remission was 0.88. At present, a DMARD-free sustained remission is the closest available proxy of ‘cure’ of RA; this outcome is presumably most suited to evaluate whether an early period in which the disease is most susceptible tot treatment exists. Current data supports the presence of a ‘window of opportunity’ effect.
This study, however, has limitations. First, this study addressed possible citation bias but not publication bias. Another major limitation is that heterogeneity in the definition of symptom onset, and consequently differences in time periods studied, prohibited us from performing a meta-analysis on all 18 selected studies. There were also large differences in study designs and study outcomes. A qualitative review was performed instead. Strengths of the current approach are that two independent readers scored all the articles, that criteria for evaluation of the quality were in line with those of previous studies,6 ,7 and that predefined and stringent qualitative levels of evidence8 were used to summarise the data (strong evidence was defined as ≥75% of the HQ studies reporting an effect in the same direction). With this approach, we tried to optimise the accuracy of this systematic qualitative literature review. An important drawback of a systematic literature review is that by simply counting the number of positive and negative studies, the sample sizes and power of the individual studies was ignored. Hence, studies with a tendency in the data that were underpowered to obtain a significant result were considered equally negative as studies with true negative findings. With a qualitative systematic literature review, the data can also not be presented in a funnel plot because an overall effect size cannot be generated.
The fact that a systematic literature review and meta-analysis may result in a slightly different answer is illustrated here with the data on achieving DMARD-free sustained mission. Three HQ studies were found. Two studies9 ,10 were significant; the third11 was not significant but showed a tendency towards an effect of symptom duration on remission (HR 0.99 (95% CI 0.98 to 1.00), p=0.099 in multivariable analysis). Given the predefined levels of evidence (table 1), two out of three positive studies would result in the conclusion that there is a moderate level of evidence for an association between symptom duration and DMARD-free sustained remission. However, in the meta-analyses there was a consistent and independent association between symptom duration and DMARD-free sustained remission.
Another difficulty when interpreting the results of the present study is the fact that different studies had included different adjustment factors in the analyses (see table 3). Not adjusting for patient and disease characteristics and treatment effects may potentially result in findings that are not driven by symptom duration but by the association of these factors with disease outcome. In 12 of the total 18 studies, 11 of the 14 positive studies, and seven of the 10 HQ-studies adjustments for patient and disease characteristics were applied, though not exactly the same factors. Four of the 10 HQ studies applied adjustments for differences in treatment. Furthermore, Mottönen et al23 observed a significant association in the single therapy group but not in the combination therapy group. In the meta-analyses on DMARD-free sustained remission, the association between symptom duration and disease outcome was independent of patient characteristics (age, gender), disease characteristics (ESR as marker of inflammation and RF as marker of autoantibody positivity) and the applied treatment strategies. The drugs used in the studies that applied adjustments for treatment were conventional DMARDs10 ,17 ,21 ,23 ,28 and/or biologics.14 ,22
Some studies evaluated symptom duration and had divided patients in groups of, for example, symptoms for >5 and <5 months,17 ,18 ,30 meaning that patients in both groups had passed the period of the ‘window of opportunity’ that was previously defined as the first 12 weeks.
Our formulated inclusion and exclusion criteria resulted in a selection of studied articles. A large proportion of screened articles was not included in this review because of not fulfilling the inclusion criteria. Some studies investigated patients with early (poly)arthritis, but not specifically RA, and were, therefore, not included.31 ,32 Several of the excluded studies did show that patients with RA with shorter symptom duration had a better outcome. For example, Green et al33 reported on disease persistency after 6 months in inflammatory patients with polyarthritis treated with corticosteroids, and observed that patients with symptom duration <12 weeks more often achieved remission. This study had 6 months of follow-up and did not fulfil our inclusion criteria of ≥1 year of follow-up. The latter was also the case for the study of Saevarsdottir et al34 on the SWEFOT trial, observing that prolonged symptom duration was associated with a lower chance on achieving a good EULAR response after 3–4 months methotrexate treatment. Several papers were excluded because the period between diagnosis and treatment onset was studied but not the period between symptom duration and treatment onset,35–39 or because the patients studied had symptoms >2 years.40–42 We acknowledge that the maximum of 2 years symptom duration as definition of early RA is arbitrary, though given our study question, we decided to focus on early RA.
In conclusion, the present systematic literature review observed that in early patients with RA who were mostly treated with conventional DMARDs, there is strong evidence that prolonged symptom duration is associated with severe joint damage progression. Additionally, a meta-analysis on achieving DMARD-free sustained remission observed an independent association for symptom duration. These results, therefore, support the notion of the presence of a ‘window of opportunity’. The details of the time frame (evaluation of when the window ‘opens’ and ‘closes’) are left to be explored in the near future. Many studies using uniform definitions of symptom onset are required to this end; definitions for standardised durations were recently proposed.5
We thank Dr Renée Allaart, Dr Adam Young, Dr Diane van der Woude and Diederik de Rooy for sharing their datasets.
Handling editor Tore K Kvien
Contributors JABvN, MK and AHMvdH-vM contributed to the conception and design of the review. JWS performed the systematic literature search. AHM and JABvN evaluated all articles. JABvN and AK performed the meta-analysis. JABvN and AHMvdH-vM drafted the article. MK, AK and TWJH revised the article for important intellectual content. All authors gave final approval of the version to be published.
Funding This work is supported by a grant of the Dutch Arthritis Foundation (Reumafonds), by The European Community Seventh Framework Program FP7 Health-F2-2008-223404 (Masterswitch) and by the IMI JU funded project BeTheCure, contract no 115142-2. The work of AHMvdH-vM is supported by a Vidi-grant of the Netherlands Organisation for Scientific Research.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.