Effectiveness of implementing a decentralized delivery of hepatitis C virus treatment with direct-acting antivirals: A systematic review with meta-analysis

Direct-acting agents (DAAs) for hepatitis C virus (HCV) treatment are safe and highly effective. Few studies described the sustained virologic response rates of treatment conducted by non-specialists. We performed a systematic review and meta‐analysis to evaluate the effectiveness of decentralized strategies of HCV treatment with DAAs. PubMed, Embase, Scopus and LILACS were searched until March-2019. Studies were screened by two researchers according to the following inclusion criteria: HCV treatment using DAAs on real-life cohort studies or clinical trials conducted by non-specialized health personnel. The primary endpoint was the sustained virologic response rate at week 12 after the end-of-treatment (SVR12), which is binary at the patient level. Data were extracted in duplicate using electronic-forms and quality appraisal was performed with the NIH Quality Assessment Tool. Heterogeneity was assessed by I2 statistics. Random-effects meta-analysis models were used for pooling SVR12 rates. Publication bias was assessed using funnel plots. Among the 130 selected studies, nine papers were included for quantitative synthesis. The quality-appraisal was good for two, fair for three and poor for four studies. The pooled relative risk (RR) of SVR12 was not statistically different between decentralized strategy and treatment by specialists [RR = 1.05; 95% confidence interval (95% CI): 0.98–1.1; I2 = 45% (95% CI: 0–84%), p = 0.145]. SVR12 rate for decentralized HCV treatment was 81% [SVR12 95% CI: 72–89%; I2 = 93% (95% CI: 88–96%)] and 95% [SVR12 95%CI: 92–98%; I2 = 77% (95% CI: 52–89%)] with intention to treat analysis and per-protocol analysis, respectively. SVR12 rates using DAAs managed by non-specialized health personnel were satisfactory and similar to those obtained by specialists. This new delivery strategy can improve access to HCV treatment, especially in resource-limited settings. PROSPERO #: CRD42019122609.

Direct-acting agents (DAAs) for hepatitis C virus (HCV) treatment are safe and highly effective. Few studies described the sustained virologic response rates of treatment conducted by non-specialists. We performed a systematic review and meta-analysis to evaluate the effectiveness of decentralized strategies of HCV treatment with DAAs. PubMed, Embase, Scopus and LILACS were searched until March-2019. Studies were screened by two researchers according to the following inclusion criteria: HCV treatment using DAAs on reallife cohort studies or clinical trials conducted by non-specialized health personnel. The primary endpoint was the sustained virologic response rate at week 12 after the end-of-treatment (SVR12), which is binary at the patient level. Data were extracted in duplicate using electronic-forms and quality appraisal was performed with the NIH Quality Assessment Tool. Heterogeneity was assessed by I 2 statistics. Random-effects meta-analysis models were used for pooling SVR12 rates. Publication bias was assessed using funnel plots. Among the 130 selected studies, nine papers were included for quantitative synthesis. The quality-appraisal was good for two, fair for three and poor for four studies. The pooled relative risk (RR) of SVR12 was not statistically different between decentralized strategy and treatment by specialists [RR = 1.05; 95% confidence interval (95% CI): 0.98-1.1; I 2 = 45% (95% CI: 0-84%), p = 0.145]. SVR12 rate for decentralized HCV treatment was 81% [SVR12 95% CI: 72-89%; I 2 = 93% (95% CI: 88-96%)] and 95% [SVR12 95%CI: 92-98%; I 2 = 77% (95% CI: 52-89%)] with intention to treat analysis and per-protocol analysis, respectively. SVR12 rates using DAAs managed by non-specialized health personnel were satisfactory and similar to those obtained by specialists. This new delivery strategy can improve access to HCV treatment, especially in resource-limited settings. PROSPERO #: CRD42019122609. PLOS

Introduction
Hepatitis C virus (HCV) therapy was revolutionized by use of safe and highly effective directacting agents (DAAs) [1,2]. Treatment with DAAs is associated with reduced risk for mortality and hepatocellular carcinoma and should be considered in all HCV-infected patients [3]. The World Health Organization (WHO) defined strategies to eliminate HCV and to reduce viral hepatitis related deaths by 2030 [4]. However, one of the main barriers to improving HCV care is the lack of an effective linkage-to-care policy for HCV infection involving treatment by non-specialists [5]. In the pegylated interferon (Peg-IFN) era, specialists in gastroenterology or hepatology were the most frequent prescribers of HCV treatment in the United States of America (USA). However, in the last years, HCV treatment managed by infectious disease specialists, internists, general practitioners (GPs) has been increased due to the safety of the new regimens [4,6]. In order to reach WHO goals in the next decade, non-specialists can play a major role in HCV treatment and long-term follow-up [7]. Several strategies to scale up HCV treatment have been proposed, such as the universal access to highly effective, well-tolerated and affordable regimens. Delivery of treatment using DAAs can involve less trained professionals at lower level health facilities [8]. Implementing new delivery strategies to improve patient access to DAAs in resource-limited settings is crucial to expand testing and treatment to eliminate HCV by 2030 as proposed by the WHO [9]. However, few studies reported relatively low sustained virologic response (SVR) rates (47-66%) of HCV treatment with DAAs conducted by non-specialists in real-world primary care settings [10,11].
To scale up the decentralization of HCV treatment with DAAs prescribed by non-specialized personnel at primary care settings, evidence synthesis studies need to be conducted to show the effectiveness and safety of these strategies. The aim of this study was to perform a systematic review and meta-analysis to assess the effectiveness of decentralized strategies of HCV treatment using DAAs.

Registration, search strategy and eligibility criteria
The protocol of this review is at the international prospective register of systematic reviews (PROSPERO #: CRD42019122609) in the following web address: http://www.crd.york.ac.uk/ PROSPERO/display_record.php?ID=CRD42019122609 Literature search strategies were conducted in PubMed, Embase, Scopus, LILACS until March-2019 by experienced researchers, with no language or publication period restrictions (S1 Table). References checking, hand searching and contact with authors were strategies used to identify additional studies. After the removal of duplicate studies, titles and abstracts were independently screened by two trained researchers using the Rayyan QRCI web application (https:// rayyan.qcri.org/) [12]. Discordances were solved in a panel with the participation of two experienced reviewers. The decision of an experienced hepatologist solved persistent disagreements. The following pre-specified eligibility criteria was adopted: HCV treatment using DAAs on real-life cohort studies or clinical trials at primary care settings, conducted by non-specialized health personnel (GP, family doctor, or any professional without specialization) with description of SVR rates. Cure of HCV was defined by sustained virologic response 12 weeks after the end-of-treatment (SVR12), which is a binary outcome at the patient-level. A detectable HCV viral load 12 weeks after treatment characterized failure of treatment (no SVR12). Studies were included regardless of presenting a comparison group that received the standard specialized care, managed by hepatologists or infectious diseases specialists. Studies were excluded if the primary endpoint, SVR12, was not showed or disaggregated to allow the assessment of the specific effectiveness of the decentralized treatment delivery strategies (S2 Table).

Data extraction
Study data were extracted in duplicate and managed using electronic data capture tools. Electronic extraction forms were created using REDCap (Research Electronic Data Capture), a secure web-based application designed to support data capture for research studies [13]. SVR12 rates of included studies considering the intention to treat (ITT) and/or the per-protocol analysis were extracted.

Risk of bias assessment
Quality appraisal was performed using the "Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies Personnel" from the National Institutes of Health (https://www. nhlbi.nih.gov/health-topics/study-quality-assessment-tools). The 14-item checklist on this form was designed to focus on the key concepts for evaluating the internal validity of a study. The quality of studies was rated as good, fair or poor.

Data analysis
Heterogeneity between the included studies was measured by I 2 statistics with the following cut-off points were used to classify heterogeneity: 25-50%, 50-75% and >75% considered as mild, moderate and severe, respectively. The "heterogi" [14] program was used to obtain 95% confidence intervals (95% CI) for I 2 in relative risk meta-analysis and "metaan" [15] for metaanalysis of proportions of SVR12. Random effects models using the method of DerSimonian & Laird, with the estimates of heterogeneity being taken from the Mantel-Haenszel models were used to pool the risk ratio (comparison between decentralized versus specialized strategies and among patients with and without cirrhosis) using "metan" command [16]. Maximum likelihood random effects models (ml) were used for pooling SVR12 rate of decentralized HCV treatment strategies and to estimate heterogeneity using "metaan" [15]. We performed subgroup analyses to explore the heterogeneity and to assess how the pre-identified variables affected the pooled estimates. In addition, fixed effects models were planned to be used only after the identification of an absence of heterogeneity using random-effects models. Results with the same data used in "metaan" for meta-analysis of proportion of SVR12 were additionally presented using random effects models (method of DerSimonian and Laird, with the estimate of heterogeneity being taken from the inverse-variance fixed-effect model) with pooled estimate after Freeman-Tukey Double Arcsine Transformation to stabilize the variances, conducted using "metaprop" [17] Stata program. Funnel plots and Egger's test assessed publication bias and small-study effects, respectively for analysis of relative risks of SVR12 with two groups' comparison using "metafunnel" [18]. Additional funnel plots were presented considering publication year using "metabias" [19]. The significance level adopted was 5% and statistical analyses were conducted using the "metan" [16] and "metaprop" [17], "metaan" [15], "metafunnel" [18] and "metabias" [19] packages from Stata-SE (2017; StataCorp LP, College Station, TX, USA).

Study characteristics
The literature search resulted in 130 unique studies (S1 Table). After the screening of titles and abstracts by two blinded investigators, 12 studies [7, 10, 11, 20-28] were eligible for full-text analysis. The percent agreement and Cohen's kappa between the two independent reviewers were 95% and 0.69, respectively. Nine papers [10, 11, 20, 22-25, 27, 28] fulfilled our eligibility criteria and were finally included after full-text reading. The flow-chart describes details on the selection process (Fig 1).
Overall, 2,099 individuals were treated in the included studies [n = 1,479 by decentralized strategies / n = 620 by specialized delivery of care]. The number of the included patients in the meta-analysis of the proportion of SVR12 with the decentralized strategy was 1,173 for ITT Effectiveness of decentralized HCV treatment: A systematic review with meta-analysis analysis and 917 for per-protocol analysis, considering that some studies showed results of both types of analyses.
Considering the three manuscripts [10, 23, 24] that compared both strategies, HCV treatment with decentralized strategies was delivered for 711 patients, while 620 received specialized care. Among these three studies, only one manuscript had the reporting of SVR12 rates for both ITT and per-protocol analysis [10].
Overall, studies included patients from five countries: USA (n = 3), Australia (n = 3), Pakistan (n = 1), Canada (n = 1), and Rwanda (n = 1). Sofosbuvir (SOF) was the most used DAA, which was present in all included studies regimens. The duration of HCV treatment was for 12 or 24 weeks but was not available for five studies. HCV patients with genotype 1 patients were the most frequently included. Several types of health personnel were used in the decentralization strategies (e.g. mid-level practitioners, social workers, nurses, general practitioners, sexual health physicians, family doctors, and internists) and different types of specialized support (e.g. telephone, electronic messaging, e-mail, education session, remote consultation, pre/posttreatment assessment, and referral to visit by clinical need). Table 1 summarizes the characteristics of included studies.  Table). Low participation rate, limited sample and absence of sample size calculation were issues for four studies.

Risk of SVR12 of decentralized versus specialized strategies
The pooled relative risk (RR) of SVR12 was not statistically different between decentralized strategy and treatment by specialists [RR = 1.05 (95% CI: 0.98-1.1); I 2 = 45% (95% CI: 0-84%); p = 0.145]. This synthesis included the three studies [10, 23, 24] that provide data on the comparison of SVR12 between decentralized and specialized strategies. Overall, 1,331 patients were treated in these studies. The pooled results considered all the types of analysis (ITT and per-protocol). Of these, 85.9% (611/711) had SVR12 with decentralized and 85.8% (532/620) with specialized HCV treatment (Fig 2). The funnel plots, even with presence of some asymmetry did not show significant evidence of publication bias, but the small number of studies limits the interpretation. The Egger's test obtained a p-value of 0.07 and an intercept of 1.67 representing no small-study effects (S1
For per-protocol analysis, the heterogeneity was not resolved by subgroup analysis using the levels of quality (S6 Fig). One study from Rwanda [28] had good quality and a lower SVR12 rate in comparison with the others. The heterogeneity decreased when the studies where grouped by levels of country income. For per-protocol analysis, the final pooled SVR12 rate for decentralized HCV treatment was 96% [SVR12 95% CI: 95-98%; I 2 = 48% (95% CI: 0-75%)] (Fig 4), including studies with 617 patients in the higher income subgroup (S7 Fig). In addition, we have found similar results with the same data used for Figs 3 and 4 using random effects models, method of DerSimonian and Laird, with the estimate of heterogeneity being taken from the inverse-variance fixed-effect model with pooled estimates after Freeman-Tukey Double Arcsine Transformation in "metaprop" Stata program (S5 Fig and S7 Fig).

SVR results of patients with and without cirrhosis treated with decentralization
Data extraction on SVR12 rates stratified by absence or presence of cirrhosis was feasible only in three studies using ITT analysis [22, 23, 25]. The SVR12 rates of patients with and without

Discussion
The evidence on the effectiveness and safety of DAAs was showed in real-world studies conducted at different settings and subsequent meta-analyses [1,[29][30][31][32][33][34][35][36]. However, to our knowledge, this is the first systematic review and meta-analysis that evaluates the effectiveness of the delivery of HCV treatment with DAAs by non-specialized health personnel at primary care. This meta-analysis highlighted that the HCV treatment using DAAs by non-specialized health personnel was effective and SVR rates were similar to treatment conducted by specialists.
The WHO estimates that up to 71 million people are chronically infected by HCV worldwide [37]. The low screening rates, lack of effective linkage-to-care policies for HCV infection and high drug costs are main barriers for HCV elimination. Many individuals who are unaware that they are HCV positive until the disease progresses into cirrhosis and its complications. HCV cascade can be improved by nation-wide HCV awareness campaign targeting high-risk groups, reduced drugs prices and establishment of multidisciplinary teams to secure linkage to care [38]. Therefore, HCV treatment by non-specialists is a key-strategy to reduce the burden of HCV infection worldwide [39].
Our study reported that SVR12 rates using DAAs by GPs and primary care health personnel are similar to rates obtained by gastroenterologists/hepatologists, reinforcing that this strategy might be implemented in resource-limited setting countries. We acknowledge that patients with cirrhosis should be managed by specialists. Patients with cirrhosis have higher rates of adverse events during treatment [36] and remain at risk of liver-related complications, such as hepatocellular carcinoma after HCV cure [40]. In our systematic review, a small number of patients (n = 134) with cirrhosis were treated by non-specialists. There was a non-significant difference on the SVR rates of patients with versus without cirrhosis treated by non-specialists [RR = 1.0 (95% CI: 0.85-1.17)]. Moreover, the SVR12 was 86.6% for individuals with cirrhosis treated by non-specialists. In our meta-analysis, the presence of cirrhosis has not impacted the primary outcome of HCV treatment. However, patients with cirrhosis should be managed by hepatologist rather than non-specialists due to the higher risk of adverse events during treatment and the need of maintenance of hepatocellular carcinoma screening during a longitudinal follow-up [41]. On the other hand, patients without cirrhosis (METAVIR F�3) can be managed by GPs or primary care doctors. People with HIV and/or HBV coinfections and other comorbidities (e.g. haemophilia, thalassemia, kidney disfunction) should be referred to specialists [42,43].
Our meta-analysis has found a considerable difference between ITT and per-protocol results for HCV treatment conducted by non-specialists. Higher lost to follow up (LTFU) is expected at real-world settings, albeit is still possible to find papers reporting similar LFTU between clinical trials and real-world data [44,45]. Previously, gaps of SVR results for realworld treatment in comparison with clinical trials were due to the LTFU [30]. Moreover, LTFU was related to treatment failure in difficult to treat patients [46]. Future studies should develop and implement new strategies to tackle the LTFU in HCV treatment.
Different decentralization strategies were adopted in the included studies. Only two authors did not mention any type of support offered by specialized personnel [20,23]. Several categories and combination of health personnel were involved, specialist support or consultancy at distance, training possibilities, and mentoring. Telemedicine support by specialists can be an effective intervention for HCV treatment decentralization to primary care [7].
The main limitations of our study were the high heterogeneity for pooled overall SVR rates and the limited number of studies that compared the decentralized HCV treatment with the standard-of-care. The high heterogeneity could be explained by methodological quality and study design; population and setting characteristics from different countries where the studies were performed, and the variability of specialized support for HCV treatment. As the included studies were investigating the outcomes of new delivery strategies for HCV treatment, with previously unknown effectiveness, the support by specialists could be over implemented due to ethical and/or safety reasons at settings where specialists are available. Apparently, we were able to manage heterogeneity using subgroup analysis considering studies' quality and income of the country where the studies occurred. Especially in small meta-analyses, it is important to avoid homogeneity assumptions [47]. The report of high heterogeneity rates in our study was an important finding itself. Moreover, the presence of large 95% CI for I 2 , even after achieving a good point estimate through subgroup analysis showed that reporting confidence intervals is very important [48] and approaching heterogeneity can be more challenging than expected. The small number of included studies is a major limitation of our systematic review. Moreover, the assessment of the publication bias that included the presentation of funnel plots is strongly limited. Funnel plots were shown to illustrate this important issue but cannot be used to conclude that publication bias was absent.
The small number of studies included and the heterogeneity results limit the recommendation of treatment decentralization for all settings, but our meta-analysis results contributes to support strategies of decentralized delivery of treatment for key populations, especially in locations where it is not possible to provide specialized care for all people living with HCV. DAAs prescribed and managed by non-specialized health personnel showed good SVR12 rates. The lack of enough specialists for HCV treatment can be tackled with DAA treatment at primary care settings by non-specialized health personnel. Specialists will be necessary to manage patients with specific clinical conditions (e.g. cirrhosis, coinfections, and comorbidities) to reduce the adverse events and improve the treatment efficacy in these cases [40,42,43]. Innovative decentralization strategies could be implemented to improve access to HCV treatment, especially in resource-limited settings, contributing to the achievement of HCV elimination targets.
Supporting information S1