HLA-B27 detection test for individuals with suspected axial spondyloarthritis to Brazilian public health system: accuracy, cost-effectiveness, and budget impact analysis

Vinicius Lins Ferreira; Layssa Andrade Oliveira; Haliton Oliveira Junior; Rosa Camila Lucchetta

doi:10.3205/hta000141

Research Article

HLA-B27 detection test for individuals with suspected axial spondyloarthritis to Brazilian public health system: accuracy, cost-effectiveness, and budget impact analysis

Vinicius Lins Ferreira ¹
Layssa Andrade Oliveira ¹
Haliton Oliveira Junior ¹
Rosa Camila Lucchetta ¹

¹ Health Technology Assessment Unit, Hospital Alemão Oswaldo Cruz, São Paulo, São Paulo, Brazil

Abstract

Objective: the aim of this study is to assess the accuracy, cost-effectiveness, and budget impact of the HLA-B27 test in diagnosis of axial spondyloarthritis to Brazilian Public Health System (SUS).

Methods: A systematic review was conducted to assess the sensitivity and specificity of HLA-B27 compared to test ASAS or New York diagnostic criteria. Additionally, a cost-effectiveness analysis (a decision tree model coupled with a Markov model and effectiveness outcome of quality-adjusted life year (QALY)) and budget impact analysis were also developed. Costs were obtained in Brazilian currency and converted to dollars considering the PPP conversion factor.

Results: Twenty-nine studies have shown that the sensitivity of the HLA-B27 test was 68% (95% CI: 67-69%) and the specificity was 88% (95% CI: 87-88%) (very low certainty of evidence to both outcomes) when compared with any comparator. The analysis of two studies showed that the combination of HLA-B27+clinical parameters still achieved sensitivity and specificity comparable to or greater than the combination of imaging and clinical parameters, or clinical parameters alone. When compared with clinical evaluation and clinical evaluation±imaging, the HLA-B27 test+clinical evaluation was close to the cost-effectiveness threshold ($17,628 per QALY) and dominated the comparator, respectively. Furthermore, incorporating HLA-B27 into the SUS would result in an increase in costs of $530,393 in 5-years of analysis (mean of six thousands/year eligible patients to HLA-B27).

Conclusion: The findings supported the coverage of the HLA-B27 test in SUS, considering that test increases accuracy when associated with other diagnostic strategies and has the potential to be cost-effective or dominant.

Keywords

spondyloarthritis, Brazil, diagnosis, HLA-B27 antigen

Introduction

Axial spondyloarthritis (SpA) is a chronic inflammatory autoimmune disease that primarily affects the spine, but also peripheral joints, and extra-articular sites such as the eye and intestine. It is characterized by intense pain, joint stiffeness due to enthesitis (inflammation at tendon insertions), and progressive functional impairment, leading to irreversible structural damage to the sacroiliac and spinal joints, accompanied by radiographic changes and excessive bone formation. Progression of the arthropathy can lead to spinal fusion and cause extreme disability, with loss of mobility in the spine, pelvis, and lower back [1], [2], [3], [4], [5].

SpA affects approximately 0.1 to 1.5% of the population, with significant geographic variation in prevalence and main clinical and phenotypic manifestations [6], [7], [8], [9]. A study reported the prevalence of the disease in South America as 0.14% (95% CI 0.02%–0.34%) [10].

The HLA-B27 antigen is strongly correlated with the onset of the disease, and a positive test for this marker is found in 80% to 98% of cases [11], [12]. In Brazil, several studies suggest smaller proportions, where HLA-B27 antigen carriers represent approximately 60–70% of patients [13], [14], [15], [16], [17], [18], [19], [20], [21].

The ASAS (Assessment of SpondyloArthritis International Society) criteria allow the diagnosis of patients without structural damage, and the modified New York classification criteria allow the diagnosis of patients with radiographic changes, in a more advanced stage of the disease. ASAS classification criteria for SpA (patients with back pain ≥3 months and age of onset <45 years) should be one of the following:

Sacroiliitis on imaging + ≥1 SpA feature, or
HLA-B27 + ≥2 others SpA features (note: SpA features includes inflammatory back pain, arthritis, enthesitis, uveitis, dactylitis, Crohn disease/ulcerative colitis, psoriasis, family history of spondyloarthritis, good response to nonsteroidal anti-inflammatory drugs, HLA-B27 positivity, elevated C-reactive protein). However, the ASAS diagnostic criteria may include the HLA-B27 test, which is not available in the Brazilian public health system [22].

In Brazilian public health system (i.e., SUS; Brazil’s Unified Health System), the process of decision about coverage of technologies involves requests based on evidence of efficacy, effectiveness, accuracy, safety, cost-effectiveness and budgetary impact for the system [23]. In this context, the objective of this study is to report the potential impact of inclusion of HLA-B27 test in diagnosis of individuals with suspected axial spondyloarthritis into SUS.

Methods

Systematic review

Initially, an updated of a systematic review originally conducted by the National Institute for Health and Care Excellence (NICE; published in 2016) was done [24]. For guide this update, the following research question was formulated: “in individuals with suspected axial spondyloarthritis, what is the sensitivity, specificity, and diagnostic utility of the HLA-B27 test compared to the ASAS criteria, New York criteria and diagnosis by a specialist”? The PIRO question is presented below (Table 1 [Tab. 1]).

Based on the PIRO question presented before, a search was conducted in January 2023 on the PubMed, Cochrane Library, and EMBASE platforms (see Tab. S1 in Attachment 1 [Att. 1]). This systematic review was conducted according to Cochrane handbook to diagnostic test accuracy [25], and reported according to PRISMA recommendations [26].

The retrieved records were imported into Rayyan [27], where duplicates were identified and removed. Then, the records were selected by a single reviewer, with a second reviewer consulted for screening (reading titles and abstracts) and eligibility (reading full texts) when necessary.

Data extraction was performed by a single evaluator using Microsoft Office Excel^® spreadsheets. The following data were extracted:

Study and participant characteristics: author, year, country, general characteristics of the population (age, sex, disease duration), number of participants, alternatives compared, reference test, and inclusion criteria.
Diagnostic accuracy outcomes: Sensitivity, specificity, and absolute numbers (true positives, true negatives, false positives, and false negatives) extracted directly from the text. When unavailable, sensitivity and specificity were calculated from absolute numbers.

To assess the risk of bias in the studies, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool was used [28]. The quality or confidence of the evidence was assessed considering the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach for diagnostic accuracy studies [29], [30], [31].

Whenever possible, individual results were pooled in meta-analyses conducted in Meta-DiSc (version 1.4) [32]. Sensitivity, specificity, and predictive values were calculated with a 95% confidence interval, and the sROC (receiver operator characteristic curve) and the area under the curve (AUC) were established based on these. Subgroup analyses were performed considering the different comparators:

ASAS diagnostic criteria without HLA-B27,
modified New York diagnostic criteria, and
other diagnostic criteria (rheumatologists, more than one diagnostic criterion evaluated in the same study, or other criteria). Heterogeneity test, such as I2 statistics, was not performed, as it is not recommended for systematic reviews of test accuracy [33].

Economic analysis

An economic evaluation was conducted to estimate the incremental cost-effectiveness ratio (ICER) of diagnosis strategies for patients with a suspected diagnosis of axial spondyloarthritis, adhering to the CHEERS Task Force Report – 2022 reporting checklist [34]. The age at entry into the model was 35 years old, with an equal sex distribution (50% male, 50% female), reflecting the average demographics of the studies included in the clinical evidence synthesis [35], [36], [37], [38], [39], [40]. The perspective adopted was SUS.

The intervention analyzed was the HLA-B27 testing and clinical evaluation, defined as at least two features of SpA (i.e., inflammatory back pain according to specialists; extraspinal manifestations such as arthritis, enthesitis, uveitis, psoriasis, ulcerative colitis, or Crohn’s disease; good response to nonsteroidal anti-inflammatory drugs; family history of SpA; elevation in C-reactive protein or erythrocyte sedimentation rate).

The comparators defined for this analysis were:

Clinical evaluation only: three or more SpA characteristics; or
Clinical evaluation ± imaging exam: at least 1 SpA characteristic + sacroiliitis on imaging exam (plain X-ray or MRI); for both comparators available SpA characteristics were according to ASAS classification criteria for SpA (as described in the last paragraph).

A decision tree coupled with the transitional state (Markov) model was developed (Figure 1 [Fig. 1]). The decision tree represented the performance of the tests, so that a positive or negative detection result could be obtained, and these could be true or false.

The Markov model considered annual cycles, a 20-year time horizon, and included 7 health states, which considered disease progression: no diagnosis, treatment with conventional drugs, treatment with first-, second-, and third-line biological drugs, standard of care, and death (shown in the Figure 2 [Fig. 2]).

It was assumed that:

Individuals with a true or false diagnosis of axial spondyloarthritis transition to the "conventional treatment" state (i.e., sulfasalazine, methotrexate or naproxen). However, individuals with a true diagnosis have a probability >0 of migrating to the other states (i.e., treatment with biological treatment with adalimumab, secukinumab, etanercept, infliximab, golimumab, or certolizumab), while individuals with a false-positive diagnosis remain in this state or transition to death;
Individuals with a true-negative diagnosis remain in the "no diagnosis" state until the end of the model or transition to death;
Individuals with a false-negative diagnosis remain in the "no diagnosis" state until the end of the second cycle. However, in the third cycle, they receive the correct diagnosis and transition to the other states in the following cycles. This assumption was made because it was considered that the lack of adequate treatment would result in the emergence of new disease characteristics, allowing for the correct diagnosis of the condition as the disease progresses.

The parameters reported by Rudwaleit, 2009a and Rudwaleit, 2009b were considered for this economic evaluation [22], [41]. Values related to the diagnostic accuracy of the compared strategies and other parameters are presented in Tab. S7 and S8 in Attachment 1 [Att. 1].

The model considered only direct costs related to diagnostic procedures and treatments specific to each cycle. Costs were obtained in reais (R$), the Brazilian currency, and converted to dollars considering the PPP conversion factor (year 2023: R$ 2.45=$ 1 [42]). The main costs considered are summarized in Table 2 [Tab. 2] and more details are presented in Tab. S9 to S13 in Attachment 1 [Att. 1].

**Table 2: Costs considered in the main analysis**

Total direct costs and effectiveness in terms of quality-adjusted life years and correctly diagnosed individuals were considered as outcomes of this analysis (cost-effectiveness threshold $16.400 per QALY gain). A discount rate of 0.05 (5.0%) has been applied.

A sensitivity analysis was performed to assess the uncertainties related to the estimates of the adopted parameters, as well as the reliability and robustness of the cost-effectiveness analysis. The minimum and maximum values used were previously described here and in Tab. 3 in Attachment 1 [Att. 1]. A probabilistic analysis with 1,000 iterations was performed and expressed as a scatterplot (cost data were varied according to the gamma distribution, while probability data were varied according to the beta distribution).

Budget impact analysis

In the reference scenario, adopting SUS perspective with a five-year time horizon, the ASAS classification criteria without the HLA-B27 test were considered as an alternative available in. Therefore, the system only provides clinical evaluation or, clinical evaluation combined with the identification of sacroiliitis on imaging tests – plain radiography or magnetic resonance imaging. In the alternative scenario, the maintenance of the ASAS Classification criteria with a diagnostic evaluation strategy, with and without HLA-B27 testing, was considered.

The market share for the reference scenario was established based on scientific literature [43], data from DataSUS (Outpatient Information System - outpatient production, SIA-PA), and expert opinion. The market share for the proposed scenario considered this information and a simulated HLA-B27 test distribution of 20% in the first year, reaching 100% in the fifth year of analysis (20% increase per year).

To estimate the eligible population, measured demand was initially considered, with data obtained from the Open Health Intelligence Situation Room (SABEIS [44]). This analysis used the number of new users using medications for the treatment of ankylosing spondylitis per year (2017 to 2022). For the remaining years of the analysis (2023 to 2027), the number of patients was calculated based on the trend in users estimated from these available data.

Furthermore, the positivity of the disease in the diagnostic evaluation identified in the studies included in the systematic review was considered. For this purpose, it was assumed that for every two individuals evaluated for axial spondyloarthritis, one would be positive; in other words, twice as many individuals with a diagnosis (prevalence) are subject to diagnostic evaluation. Finally, the proportion of 63% that would have tested positive with only clinical evaluation and imaging was subtracted.

In budget impact analysis, as imaging exam and clinical features are still going to be available for patients, as they are part of ASAS classification criteria for SpA, and Brazilian diagnosis recommendations, only the cost was HLA-B27 was considered.