Is symptom duration before DMARD therapy a determinant of direct and indirect costs in DMARD-naïve RA patients? A systematic review

Abstract Objective Early treatment of RA improves clinical outcomes; however, the impact on health economic outcomes is unclear. This review sought to investigate the relationship between symptom/disease duration and resource utilization/costs and the responsiveness of costs following RA diagnosis. Methods A systematic search was performed on Pubmed, EMBASE, CINAHL and Medline. Studies were eligible if patients were DMARD-naïve and fulfilled 1987 ACR or 2010 ACR/EULAR RA classification criteria. Studies had to report symptom/disease duration and resource utilization or direct/indirect costs as health economic outcomes. The relationships between symptom/disease duration and costs were explored. Results Three hundred and fifty-seven records were identified in a systematic search; nine were eligible for analysis. The mean/median of symptom/disease duration in studies ranged between 25 days and 6 years. Annual direct costs of RA following diagnosis showed a U-shaped distribution in two studies. Longer symptom duration before starting a DMARD (>180 days) was associated with lower health-care utilization in the first year of RA diagnosis in one study. Annual direct and indirect costs 6 months before RA diagnosis were higher in patients with shorter symptom duration (<6 months) in one study. Given the clinical and methodological heterogeneities, the association between symptom/disease duration and costs after diagnosis was not computed. Conclusion The association between symptom/disease duration at the time of DMARD initiation and resource utilization/cost in patients with RA remains unclear. Health economic modelling with clearly defined symptom duration, resource utilization and long-term productivity is vital to address this evidence gap.


Introduction
The impact of early treatment on clinical outcomes in RA is well reported [1]. However, the impact of early treatment on health economic outcomes is less clear. Patients with RA treated with intensive DMARD were more likely to stay in the workforce long term [2,3]. This might result long term in overall lower indirect costs (i.e. lower loss of productivity). However, diagnostic decisions are vulnerable to false-positive and false-negative results. The consequence of over-diagnosis and over-treatment might lead to overall higher direct costs (i.e. higher medical costs) in the longer run, which might offset the cost savings made from improved productivity. Therefore, long-term economic diagnostic and treatment decision models are required to inform the optimal threshold for diagnostic/treatment decisions from an economic perspective. This will facilitate the estimation of long-term RA-related costs.
Therefore, as a first step, the relationship between symptom/diagnosis duration at the time of DMARD initiation and subsequent resource utilization/costs needs to be identified. We sought to investigate this through a systematic review of cost-of-illness and cost-effectiveness studies of DMARD-naïve RA patients.

Methods
The full Methods section is detailed in Supplementary Data S1, available at Rheumatology Advances in Practice online.

Study identification/search strategy
PubMed, EMBASE, CINAHL and Medline electronic databases were searched up to 25 January 2023. All systematic searches were conducted using the same search terms and strategy (Supplementary Data S2, available at Rheumatology Advances in Practice online). Additional records were identified through independent manual database searching, external sources and reference scanning of relevant retrieved full-text articles. Study selection, data extraction and quality assessment were done independently by two authors (I.S. and R.S.); discrepancies were resolved by consensus or through a third reviewer (A.Bo.). Table 1 shows the PICOT framework.

Study selection
Study inclusion criteria were as follows: aged 18 years and fulfilling the 1987 ACR or 2010 ACR/EULAR RA classification criteria; DMARD-naïve; symptom/disease duration reported; cross-sectional and longitudinal study; and health economic outcomes reported as costs or resource utilization. Studies excluded were studies of non-RA inflammatory arthritides and conference abstracts, systematic reviews and review articles.

Data extraction
The following data were extracted: study characteristics; potential determinants of RA costs; sources of resource utilization and costs; and health economic outcomes.

Quality assessment
The Strengthening The Reporting of Observational Studies in Epidemiology (STROBE) checklist [4] and a modified checklist by Drummond and Jefferson [5] were used for quality assessment.

Data synthesis and statistical analysis
A meta-analysis/regression on the association between disease/ symptom duration and costs could not be performed owing to the number of studies and methodological heterogeneity, especially in reporting of health economic outcomes. Cost data per patient per year for the reported duration in studies were recorded and summarized in a unifying currency of US Dollars 2021 after adjusting for the Purchasing Power Parity (PPP) and Consumer Price Index (CPI) 2021 [6,7].

Results
Nine articles were included in this systematic review. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart shows the literature search results (Fig. 1). Table 2 summarizes study characteristics, cost categories and annual costs in international USD 2021. Six papers were cost-of-illness studies [8][9][10][11][12][13] and the remainder cost-utility Key messages • The association between symptom/disease duration before DMARD initiation and health economic outcomes in RA is unclear.
f Subtotal of medical cost includes costs owing to contacts with health-care workers, days spent in care facilities, medication, monitoring for side effects and alternative medicine. Subtotal of non-medical direct cost includes costs of adaptations in the home, devices and other costs. g HCA¼ mean productivity per day over a 5-year follow-up was calculated for each patient and multiplied by the cumulative number of their days off work to yield the patients' loss of productivity by the HCA. FCA¼ estimation of loss of productivity, with the assumption that someone replaces the disabled worker after the friction period, that the initial production level is restored, and that production losses are confined to the friction period. RA-related work disability days were obtained from the official register, divided by the duration (in years) of follow-up during which the patient had not retired owing to other diseases or because of age. All final cost column states the cost per person per year in USD 2021 after adjusting for purchasing power parity and Consumer Price Index 2021. GP: general practitioner; IQR: interquartile range; RCT: randomized controlled trial; TCZ: tocilizumab; USD: US dollars.

Systematic review of direct and indirect costs in DMARD-naive RA patients
source of cost reference and results in local currency are summarized in Supplementary Table S2, available at Rheumatology Advances in Practice online.
Resource utilization and cost data across studies were heterogeneous (Table 1). Three studies reported costs (i.e. monetary value) but not resource utilization [13,15,16]. One study reported resource utilization without monetary values [8]. Three studies reported resource utilization and costs [10,12,14]. Two studies reported costs data as loss of productivity costs [9,11].
Loss of productivity (indirect cost) was recorded in four studies [9,11,14,15]. Two studies calculated productivity loss using the human capital and friction cost approach [9,15]. One study used only the human capital approach [11], and one study used only the friction cost approach [14].
Study perspective refers to the point of view adopted in the economic evaluations [17], i.e. who pays for the cost. Common study perspectives are the patient, health-care system or society. Three studies reported societal perspectives (i.e. health-care and productivity loss costs) [13][14][15]. Two studies reported a partial societal perspective (productivity loss costs) [9,11], and two studies reported costs from the health-care perspective [8,16]. In addition, two studies reported both health-care (direct medical costs) and patient perspectives [10,12].
Quality assessment has been included in Supplementary Data S3 and Table S3, available at Rheumatology Advances in Practice online.

Narrative synthesis
Luurssen-Masurel et al. [14] performed a cost-utility study in seronegative RA patients in the Rotterdam Early Arthritis Cohort (tREACH) trial. The median symptom duration was 134 days [interquartile range (IQR) 95-205 days]; follow-up duration was 1 year. Initial treatment strategies were MTX (iMTX) 25 mg once weekly, HCQ (iHCQ) 400 mg daily or a tapering course of oral glucocorticoids (iGC). There was no significant difference in the mean cumulative health-care costs over 1 year for treatment with iMTX, iHCQ and iGCs ( Table 2). The difference in productivity costs over 1 year between the three groups was mainly attributed to different levels of presenteeism (Table 1). After adjusting for PPP and CPI 2021, mean total costs (health-care and productivity costs) by treatment strategy groups in USD 2021 were $14 485, $14 988 and $14 044 for the iMTX, iHCQ and iGC groups, respectively. The association between symptom duration and health-care/productivity costs in the overall cohort or by treatment groups was not assessed.
Verhoeven et al. [15] reported a 5-year cost-utility analysis of an RCT comparing tocilizumab (TCZ) plus MTX or TCZ monotherapy with MTX monotherapy in DMARD-naïve early RA patients. The median (IQR) symptom duration by treatment groups was 25  days, 26 (18-45) days and 27  days for the TCZ plus MTX, TCZ and MTX groups, respectively. Cumulative 5-year productivity cost loss [by human capital approach (HCA)] was highest in the TCZ plus MTX group (e51 700; n ¼ 106) compared with the TCZ monotherapy and MTX monotherapy groups [e39 900; n ¼ 103 and e46 500, n ¼ 108 respectively]. Cumulative 5year productivity cost loss (HCA) was highest in the TCZ plus MTX group (e51 700) compared with the TCZ monotherapy and MTX monotherapy groups (e39 900 and e46 500, respectively). After adjusting for PPP and CPI 2021, total direct health-care-related costs (mean) in USD 2021 at the end of year 1 were $15 546, $8350 and $17 840 per patient for the TCZ plus MTX, TCZ and MTX groups, respectively. The association between symptom duration and health-care or productivity costs in the overall cohort or by treatment groups was not assessed.
Syngle et al. [16] reported RA-related health-care costs in a single-centre prospective observational study of 3 months in India. The study assessed the cost-effectiveness of synthetic DMARDs in DMARD-naïve RA patients [16]. The mean disease duration was 5.78 years (S.D. 4.84 years). Costs reported were the average total direct medical cost per prescription per month over the 3-month study period. This figure equates to 997.05 Indian Rupees per patient. After adjusting for PPP and CPI 2021, the average (extrapolated) annual direct medical costs at the end of year 1 in USD 2021 was $1008 per patient. The association between disease duration and direct medical costs was not assessed.
Kuijper et al. [8] compared health-care utilization between arthralgia and DMARD-naïve early RA patients at baseline, 6 and 12 months in a Dutch inception observational cohort study [8]. The median symptom duration for RA patients was 103 days (range 7-373 days). Use of DMARDs was not reported. A longer (>180 days) vs short symptom duration (90-180 days) at baseline was associated with lower levels of health-care utilization over 12 months [Incidence Ratio Rate of 0.65 (95% CI 0.50, 0.85, P ¼ 0.002)]. The mean number of visits to medical specialists peaked at 6 months in the RA group (Table 2). However, a decrease in overall health-care visits (i.e. general practitioner, medical specialist, physiotherapist and alternative health practitioner visits) was observed following diagnosis (Table 2). No monetary value was reported in this study. In summary, longer symptom duration (>180 days) was associated with lower health-care utilization over the first year of diagnosis.
Puolakka et al. [9] assessed the impact of the Stanford Health Questionnaire (HAQ) index on loss of productivity in early DMARD-naïve RA patients in the Finnish RA Combination Therapy (FIN-RACo) open-label extension clinical trial in Finland. Patients were randomized to either a combination of three DMARDs (SSZ, MTX and HCQ) and prednisolone, or a single DMARD with or without prednisolone [9] for 2 years and were followed up for 5 years. The mean disease duration across the four HAQ groups was between 8 and 11 months. In the overall cohort and over 5 years, the annual mean loss of productivity per patient was e8344 (95% CI 6516, 10 480) by the HCA and e1928 (95% CI 1567, 2298) by the friction cost approach (FCA). Functional capacity was assessed by HAQ at baseline and 6 months. The HAQ score after 6 months of treatment, but not the level of HAQ at baseline, predicted productivity costs in the overall cohort. Over 5 years, the top HAQ quartile had the highest work disability days per year [mean 273 days (95% CI 194, 328)], compared with the lowest HAQ quartile [mean 34 days (95% CI 5, 145)]. After adjusting for PPP and CPI 2021, the annual mean loss of productivity in USD 2021 in the top quartile group was $40 116 by the HCA method and $6125 by the FCA method. No analysis was performed to assess the impact of disease duration on costs in the overall cohort or by HAQ groups.
Verstappen et al. [10] assessed the total annual direct costs over different follow-up periods after first DMARD in Dutch patients with RA and identified sociodemographic, clinical and psychological predictors of high costs in two RCTs. Patients in the first RCT were randomized into one of four treatment arms [pyramid (NSAID followed by a DMARD for treatment failure), i.m. gold, MTX or HCQ]. Patients from the second RCT were randomized into intensive vs conventional MTX regimes. In this study, costs data were classified into three groups with increasing follow-up duration after diagnosis (0 to 2 years, 2 to 6 years and 6 to 10 years). In addition, RA patients with disease duration 10 years from the Utrecht RA Cohort study group were included to capture costs data for patients with longstanding RA. There was a significant difference in annual direct costs between the four groups. The median annual direct costs per patient showed a U-shaped distribution, i.e. costs were high for patients with follow-up duration 0 to 2 years (e2923) and reduced after 2-6 years (e1967), but increased again for 10 years followup duration (e3778). Data from the group with the shortest follow-up duration were extracted for Table 1. Functional disability (HAQ) was the most important variable associated with high costs after adjusting for sociodemographic, clinical and psychological variables. After adjusting for PPP and CPI 2021, the annual mean (median) of total direct costs per patient in USD 2021 was $14 613 ($8159). The annual direct costs of early RA follow a U-shaped distribution over 10 years following the start of DMARDs. No analysis was performed to assess the impact of disease duration at baseline on costs in the overall cohort.
Merkesdal et al. [11] reported the magnitude of indirect costs, changes within cost components and the correlation between changes in cost and social, clinical and occupational variables within first 3 years for DMARD-naïve RA patients in a multicentre observational study in Germany. The average indirect cost in early RA at the 24-month follow-up was high; $11 750 per person-year (US dollars for the period 1994-1996), which related to 126 days of loss of productivity. Loss of productivity owing to sick leave accounted for 84% of overall loss of productivity (sick leave, work disability and other work loss) between the onset of disease and the end of the first year after study enrolment, compared with only 25% at the end of the second year of the study enrolment [11]. After adjusting for PPP and CPI 2021, the mean costs associated with total sick leave, work disability and other work losses in USD 2021 were $20 180 after 12 months of followup and $18 848 per person per year at the 24-month followup time point. The relationship between disease duration and loss of productivity was not reported.
Newhall-Perry et al. [13] assessed the direct and indirect costs of seropositive RA patients 6 months before diagnosis in a longitudinal observational study at rheumatology centres in the western USA and Mexico. All patients were DMARDnaïve and had clinically active disease, with at least nine tender and six swollen joints and a positive RF. Patients were classified as disease duration of <6 months (n ¼ 87) and 6 months (n ¼ 63). At baseline, the mean total direct costs and indirect costs of RA 6 months before diagnosis were $200 per month and $281 per month in 1994 USD, respectively. The total direct costs of RA [mean (S.D. )] 6 months before diagnosis in patients with disease duration <6 months compared with 6 months were $240/month 6 $285 and $144/ month 6 $149, P < 0.001, respectively. Likewise, indirect costs were higher in patients with a disease duration <6 months as opposed to 6 months ($348/month 6 $567 vs $188/month 6 $506; P < 0.005) at baseline. After adjusting for PPP and CPI 2021, the annual mean total direct and indirect costs 6 months before diagnosis per person in USD 2021 were $12 663 for <6 months and $7174 for 6 months groups. Overall, annual direct and indirect costs 6 months before RA diagnosis were higher in patients with shorter symptom duration (<6 months).
van Jaarsveld et al. [12] assessed the annual direct cost related to RA during the first 6 years and identified socioeconomic and clinical determinants of these costs in an RCT conducted in the Netherlands. Patients were recruited between 1990 and 1996, and cost questionnaires were sent to those not lost to follow-up in April 1996. Mean annual direct costs by follow-up duration (year 1-6) followed a U-shaped distribution, as follows: Dutch florin (Dfl.) 14 455/patient in year 1; Dfl.13 800/patient in year 2; Dfl. 9457/patient in year 3; Dfl. 6233/patient in year 4; Dfl. 13 005/patient in year 5; and Dfl. 11 158/patient in year 6. After adjusting for PPP and CPI 2021, total direct costs per patient (mean) in USD 2021 were $24 094 after 1 year follow-up duration. The annual direct costs of early RA showed a U-shaped distribution over 6 years following the start of DMARDs. No analysis was performed to assess the impact of disease duration at baseline on costs in the overall cohort.
A number of studies were excluded because study participants could receive at least one DMARD before study enrolment [18][19][20][21]. Tables 3 and 4 summarize the direct and indirect costs in USD 2021, respectively, and outcomes by increasing symptom or disease duration.

Discussion
This study highlighted several interesting findings. Firstly, two studies reported a U-shaped distribution of costs over disease duration following an RA diagnosis. Total costs were high during the initial years, slightly lower thereafter, then high again for a disease duration of 5 years [12] and >10 years [10]. This indicates that costs are not a linear function of disease duration.
Secondly, functional disability was a predictor of productivity costs in three studies [9,10,12]. In one study, patients from the highest HAQ group had the highest work disability days per year, hence the highest costs for loss of productivity [9]. This finding is highly relevant. It supports the hypothesis that aggressive early treatment can reduce costs in the longer term, because those treated earlier are less likely to have a higher level of disability, which then translates to a lower loss of productivity costs in the long term.
One study reported that the annual direct and indirect costs 6 months before diagnosis were higher in those with a symptom duration of <6 months before the start of DMARD therapy compared with those with a symptom duration   [13]. In contrast, another study reported that longer symptom duration before diagnosis (>180 days) was associated with lower health-care utilization over the first year of diagnosis [8]. The contrasting trend between the two studies can be explained by the difference in the timing of when the health economic outcomes were recorded. Health-care utilization over the first year following RA diagnosis was recorded in the latter study; however, costs before RA diagnosis were recorded in the former study.
In this review, we could not delineate the aggregated-level data related to the relationship between symptom/disease/diagnosis duration and cost categories owing to the heterogeneity of the following factors: timing and duration of data collection regarding resources and costs; type of resources/ cost-categories reported; and inconsistency in reported disease, symptom or diagnosis duration (Fig. 2). Moreover, the duration of cost data recorded (i.e. 6 months vs 6 years) also differed across studies (Fig. 2).
Before the era of early treatment, RA costs were related to established disease. Patients had more frequent hospitalization [22] and more frequent joint replacement than the general population [23], and a majority were unable to work. The early introduction of biological and targeted synthetic DMARD therapy has resulted in high costs of medications [23]. However, high drug cost can potentially be offset in the long term, at least in part, by reducing disease-related costs (e.g. loss of productivity owing to work disability, hospitalization and joint surgery). In addition, patients treated early were more likely to achieve DMARD-free remission [1]. Therefore, this would reduce the proportion of patients on long-term DMARDs [24].
Clear definitions of RA onset and duration have been proposed [25], because reporting in clinical studies is currently heterogeneous [25]. RA duration can be timed from the following points: onset of RA symptoms; onset of joint swelling; when RA classification criteria were first fulfilled; or the time of RA diagnosis. Using a clearly defined onset will allow meaningful comparison of clinical outcomes and health economic outcomes between early RA studies.
A strength of this review is the broad range of health economic outcomes and types of health economic studies that were included. Both direct and indirect costs, and cost-of- Figure 2. Timing and duration for which the respective health economic outcomes are reported and the symptom duration before DMARD initiation. The blue arrows indicate the symptom/disease duration reported in each study. The green arrows indicate the timing and duration of health economic outcomes reported in each study. Puolakka et al. [9] reported six groups of patients stratified by HAQ groups. a Verstappen et al. [10] reported four groups of patients based on disease duration (defined as the time elapsed from study recruitment). Van Jaarsveld et al. [12] reported six groups of patients based on disease duration (defined as the time elapsed from study recruitment). b Kuijper et al. [8] and Newhall-Perry et al. [13] reported disease duration at the time of study enrolment. FCA: friction cost approach; HCA: human capital approach illness and cost-utility studies were within the scope of this review. Observational and clinical trials were also included.
However, only a small number of studies fulfilled our strict inclusion criteria. In addition, studies that enrolled patients who had recently been treated with DMARDs before study recruitment were not included in this review. Furthermore, meta-analyses/regression were not possible owing to the different types of health economic outcomes reported.
This review is the first to highlight a vital evidence gap in early arthritis: what is the financial consequence of diagnosing and treating patients with RA during the early disease phase? Health economic modelling with carefully defined symptom duration, resource utilization, treatment and long-term productivity costs is vital to address this important question.

Supplementary material
Supplementary material is available at Rheumatology Advances in Practice online.

Data availability
The data underlying this article will be shared on reasonable request to the corresponding author.