Factors Associated with Post-Transplant Active Epstein-Barr Virus Infection and Lymphoproliferative Disease in Hematopoietic Stem Cell Transplant Recipients: A Systematic Review and Meta-Analysis

This systematic review was undertaken to identify risk factors associated with post-transplant Epstein–Barr virus (EBV) active infection and post-transplant lymphoproliferative disease (PTLD) in pediatric and adult recipients of hematopoietic stem cell transplants (HSCT). A literature search was conducted in PubMed and EMBASE to identify studies published until 30 June 2020. Descriptive information was extracted for each individual study, and data were compiled for individual risk factors, including, when possible, relative risks with 95% confidence intervals and/or p-values. Meta-analyses were planned when possible. The methodological quality and potential for bias of included studies were also evaluated. Of the 3362 titles retrieved, 77 were included (62 for EBV infection and 22 for PTLD). The overall quality of the studies was strong. Several risk factors were explored in these studies, but few statistically significant associations were identified. The use of anti-thymocyte globulin (ATG) was identified as the most important risk factor positively associated with post-transplant active EBV infection and with PTLD. The pooled relative risks obtained using the random-effect model were 5.26 (95% CI: 2.92–9.45) and 4.17 (95% CI: 2.61–6.68) for the association between ATG and post-transplant EBV infection and PTLD, respectively. Other risk factors for EBV and PTLD were found in the included studies, such as graft-versus-host disease, type of conditioning regimen or type of donor, but results are conflicting. In conclusion, the results of this systematic review indicate that ATG increases the risk of EBV infection and PTLD, but the link with all other factors is either nonexistent or much less convincing.


Introduction
Hematopoietic stem cell transplant (HSCT) recipients are at risk of developing posttransplant lymphoproliferative disease (PTLD) following primary or reactivated infection by the Epstein-Barr virus (EBV) [1][2][3][4][5][6][7]. EBV is a ubiquitous human herpesvirus with a seroprevalence approximating 50-55% of the pediatric population living in countries with high hygienic standards and reaching 90-99% by mid-adulthood [8][9][10]. EBV is the etiologic agent of infectious mononucleosis and is also associated with the development of some cancers, most notably Hodgkin's lymphoma, African Burkitt's lymphoma and nasopharyngeal carcinoma [11][12][13], as well as lymphoproliferative disease in immunocompromised individuals [14]. After primary infection, EBV establishes latent infection in B cells [15]. In immunocompetent individuals, primary infection is often subclinical and latent infection is usually well controlled by the immune system throughout life. However, when the cytolytic T-lymphocyte arm of the immune system is suppressed, primary infection can be more consequential, and latent EBV can reactivate to cause a spectrum of EBV-associated diseases ranging from fever, EBV end-organ diseases, such as pneumonia, hepatitis and encephalitis, to PTLD [16]. PTLD is a complex disorder whereby an interplay of factors is involved in facilitating tumorigenesis [17].
The occurrence of PTLD in patients receiving an allogeneic HSCT can reach 24%, depending on the presence of risk factors [16,18,19]. The highest incidence of PTLD is seen in the first six months post-transplant, with most cases occurring during the first year post-transplant [4,6,20]. Infants are generally at higher risk because they are most often EBV-naïve before transplant [21]. Several clinical risk factors have been associated with EBV infection and PTLD in HSCT, including T-cell depletion of the graft, use of unrelated donors or of two or more HLA-mismatches in related donors, use of anti-lymphocyte serum for prevention or treatment of acute graft-versus-host disease (GvHD) and use of anti-CD3 monoclonal antibodies for acute GvHD [19].
The rapid increase of EBV viral load (EBV-VL) in the blood is a well-documented predictive biomarker of EBV-associated diseases. Following transplantation, regular monitoring of EBV-VL is usually performed for better management of patients who show large spikes in VL [22,23]. A reduction in the intensity of immunosuppression or treatment with the anti-CD20 monoclonal (rituximab) is effective in decreasing EBV-VL to prevent PTLD [24]. In patients receiving HSCT, rituximab use is the more common option [7]. However, both methods for lowering EBV-VL have important disadvantages. Reduction in the intensity of immunosuppression can increase the risk of GvHD [6], while rituximab use in patients who are already immunosuppressed can incur the development of other fatal infections [6]. Rituximab targets CD20-expressing malignant B cells as well as all mature B cells, thus impeding the production of antimicrobial immunoglobulins [6]. There is clearly an important clinical advantage in preventing EBV disease rather than attempting to cure it. Numerous studies have sought to better understand the determinants of EBV infection following allogeneic HSCT. The literature contains numerous important studies that consider one or more risk factors in small to large sample sizes of patients with different characteristics; however, no systematic review is available summarizing the determinants of EBV infection in HSCT. Therefore, the aim of this work was to synthesize, through a systematic review and meta-analysis, the risk factors associated with active EBV infection and with PTLD in HSCT recipients.

Methods
We conducted, using Medline and EMBASE, a systematic search of all articles on risk factors for active EBV infection (including EBV primary infection as well as EBV reactivation) and PTLD in pediatric and adult recipients of HSCT published in peerreviewed journals between 1946 and 30 June 2020. A non-exhaustive list of concepts and keywords was obtained by referring to articles related to active EBV infection and PTLD; the list was broadened using medical subject heading (MeSH) descriptors in Medline and Emtree in EMBASE. The ovidSP interface was used to search in both databases. The search equations are presented in Table S1. The selection of the articles was done in four steps: (1) title exploration, (2) review of abstracts, (3) review of the articles' contents, and (4) review of the references of selected articles. All selection steps were performed independently by two authors (PE, MZ); in cases of disagreement, a third author (HT) was solicited for a consensual decision.
For the systematic review, three inclusion conditions were applied: (1) the study population had to be composed of pediatric and/or adult HSCT recipients; (2) risk factors for EBV infection or for PTLD had to be analyzed using univariate and/or multivariate statistical methods; and (3) the paper had to be in English or French. Abstracts, conference papers, congress papers, editorials, guidelines, reviews and case reports were excluded from the systematic review.
Two independent authors (PE and MZ) extracted the following information from the selected articles: authors, publication year, location, study type, post-transplant follow-up duration, transplant type, sample size, population (child or adult, and median or mean age, range or interquartile range), the definition of PTLD or definition of EBV infection, frequency of EBV-VL testing, blood compartment used to measure EBV-VL, and statistical methods used. In addition, for all potential risk factors explored in the studies, point estimates, such as odds ratio (OR), risk ratio (RR), the hazard ratio (HR) and subhazard ratio (SHR), confidence intervals (CI) and p-values were extracted when reported. In some cases, the corresponding author was contacted to clarify ambiguities.
The quality of each individual study was independently evaluated by two authors (PE & MZ) using a modified version of the Effective Public Health Practice Project (EPHPP) quality assessment tool for quantitative studies [25,26]. The quality assessment was based on the following components: selection bias, study design, confounders and data collection methods; it was rated as strong, moderate or weak (from high-quality to low-quality) according to the definition presented in Table S2a.
Finally, risk factors explored in these studies were described by providing the total number of studies showing a statistically significant association contrasted to the total number of studies investigating the risk factor. The data reported made it possible to perform a meta-analysis solely to measure the association between the use of anti-thymocyte globulin (ATG) and two outcomes: posttransplant EBV infection and PTLD. Studies using multivariate analysis were considered for the meta-analysis except for the study by Liu et al. [27] because only the p-value was reported (not the measure of association). Since post-transplant EBV infection is not a rare event in this population, to obtain pooled estimates, results from studies that reported adjusted HR or SHR were combined separately from those that reported adjusted OR. This distinction was not made for PTLD, which is a relatively less frequent event. Adjusted estimates were combined using the inverse variance method with the fixed-effect model or random-effect model. The choice between these two models was guided by the value of statistic I, which revealed the proportion of the total variance observed due to a real difference in the measures of effects between studies. The fixed-effect model was used when I 2 < 25% and the random-effect model when I 2 ≥ 25% [28,29]. We also performed a sensitivity analysis to assess the contribution of each study to the pooled estimate. To this end, the pooled estimate was recalculated, each time excluding only one of the studies considered [30]. The analysis was performed with software R version 3.6.1.

Results
In total, 3362 titles were identified in the research bases, 1883 in EMBASE and 1479 in Medline. Once duplicates and papers with exclusion criteria were removed, 77 articles [4,16,23,24,27, fulfilled the inclusion criteria for our systematic review (62 for EBV and 22 for PTLD). Among the 22 articles selected for PTLD, seven were also retained for EBV. Detailed information on the selection procedure is provided in the flow diagram ( Figure 1). Among the 62 articles included to analyze risk factors for post-transplant EBV infection, two relate exactly to the same patient cohort (Bogunia-Kubik et al. [36] and Bogunia-Kubik et al. [35]) and 11 relate to non-disjoint samples (include some of the same patients) (Cesaro et al. [41] and Cesaro et al. [42]; Liu et al. [74] and Liu et al. [27]; Xuan et al. [99] and Liu et al. [74]; Liu et al. [75] and Liu et al. [73]; Wang et al. [97] and Ru et al. [86]; Zhou et al. [102] and Zhou et al. [101]). However, none of these studies but one [101] were excluded from the qualitative synthesis for duplication because the risk factors explored were different. The study by Zhou et al. [102] was excluded because all variables in this paper were explored using univariate analysis and were considered in the study by Zhou et al. [101] using multivariate analysis. With respect to the analysis of PTLD risk factors, the studies by Sundin et al. [103] and Omar et al. [81] were discarded because Uhlin et al. [94] explored the same factors and sample population as these two studies. Hoegh-Petersen et al. [60] and Kalra et al. [67] used non-disjoint samples. However, these two studies were retained in the review because the risk factors explored were not completely identical. For the same risk factors explored in both studies, those from Kalra et al. [67] were retained as the analyses were done on a larger sample. In addition, the Ali et al. [31] and Althubaiti et al. [32] studies were carried out with non-disjoint samples, but both were retained because different variables were explored. Characteristics of the selected studies are described in Table 1; more details are provided in Table S3. Briefly, among the 77 studies, seven were conducted in France, seven in Italy, six in Poland, one in Belgium, three in Spain, six in the United Kingdom, one in Finland, five in the United States, three in Japan, 16 in China, two in Korea, one in Russia, five in Canada, three in Sweden, one in Turkey, three in Germany, two in the Netherlands, one in Greece, one in Portugal and three were multi-national. Twenty-three studies were prospective, 51 were retrospective, two were case-control studies and one was a randomized control trial. The sample size ranged from 26 to 64,539 HSCT recipients ( Figure 2). It is noteworthy that most studies were performed with pooled pediatric and adult populations (n = 41), while 19 included only children and 17 only adults. to the type of statistical analysis carried out, the type of population and the outcome; (F) Number of studies by type of outcome and by quality level. EBV and PTLD: The two outcomes were studied separately in the same article; EBV/PTLD: The two outcomes were combined into one. * Studies with the outcome PTLD/EBV and the studies with outcome PTLD were considered together. The definition of post-transplant EBV infection and the diagnostic criteria for PTLD differed among studies. Active post-transplant EBV infection was diagnosed when the EBV-VL in blood, determined using a PCR test, was above a given threshold. In some cases, thresholds were not readily comparable because there was no direct conversion between the units of measurement used. In two studies [45,91], active EBV infection was defined as a reactivation event because all patients showed positive EBV serology when the follow-up period started. In other studies, no distinction was made between primary infection and reactivation: both were considered active EBV infection. There was also some variability between studies with respect to the frequency of PCR testing, but it was performed weekly in most studies during the early post-transplant period. The type of specimen tested by PCR varied, with peripheral blood in 19 studies, plasma in 11 studies, serum in five studies, serum or plasma in one study, whole blood in 13 studies, peripheral blood or whole blood in two studies, and whole blood and plasma in one study; 10 studies provided no information on specimen type. The method used to diagnose PTLD was not detailed in one study. The length of follow-up was an important source of variation between studies; in some cases the follow-up period was not reported [31,32,34,40,46,48,52,55,59,61,63,73,81,83,86,90,94,98,100].
Various statistical methods were used. Logistic regression was used in 13 studies, Cox model in 22, survival analysis using the log-rank test in one, multiple linear regression in two, Fine and Gray competitive risk model in 10 and Poisson regression for grouped survival data in one. Among the 28 studies employing univariate analysis, the statistical method was not explicitly reported in one study; one study used univariate logistic regression, another used the univariate Cox model, another used time-dependent landmark, while other studies used at least one of the following tests: Log-rank test, Gray's test, Chi 2 test, Wilcoxon nonparametric test, Kruskal-Wallis test, Fisher's exact test, Mann-Whitney test or Wald test. Among the 49 studies in which a multivariate analysis was performed, the criteria for selecting variables for the multivariate model were explicitly indicated in 20.
Variables with a p-value < 0.1 in univariate analysis were retained for multivariate analysis in eight studies, a p-value < 0.2 in three, a p-value < 0.3 in one, a p-value < 0.05 in one; a p-value < 0.05 was used for the multivariate analysis in three studies. In four studies, the investigators used a p-value < 0.1 in the univariate analysis combined with a p-value < 0.05 in multivariate analysis. Altogether, 52 studies were considered as properly adjusted for confounding bias. Table S2b reports the quality assessment of the 74 articles included in the review according to the outcome. Regarding post-transplant EBV infection, 27 (42.9%) were rated "strong", 15 (23.8%) "moderate" and 21 (33.3%) "weak". For PTLD, 12 (57.1%) were rated "strong", three (14.3%) "moderate" and six (28.6%) "weak". The lack of information on retention and potential for selection bias, as well as the absence of control for potential confounding bias, were the main contributors to the lower overall rating of articles. It is important to note that the absence of control for confusion in several articles could be justified by the fact that their main focus did not involve analysis of factors associated with either post-transplant EBV infection or PTLD. Table S4 provides a detailed list of the risk factors for post-transplant EBV infection and for PTLD explored in the 77 included studies with a description, when possible, of the measures of association and CI and/or p-value. Figure 3 presents, for every individual risk factor, the total number of studies that investigated the risk factor contrasted to the number that showed a statistically significant association. Table 2 presents a summary of the risk factors (with measures of association) for post-transplant EBV infection and for PTLD that were analyzed in studies using multivariate analysis. The presence of GvHD, use of ATG and type of conditioning regimen were the three risk factors most frequently associated with EBV infection and PTLD.
Vaccines 2021, 9, x FOR PEER REVIEW 10 of 30 Table S4 provides a detailed list of the risk factors for post-transplant EBV infection and for PTLD explored in the 77 included studies with a description, when possible, of the measures of association and CI and/or p-value. Figure 3 presents, for every individual risk factor, the total number of studies that investigated the risk factor contrasted to the number that showed a statistically significant association. Table 2 presents a summary of the risk factors (with measures of association) for post-transplant EBV infection and for PTLD that were analyzed in studies using multivariate analysis. The presence of GvHD, use of ATG and type of conditioning regimen were the three risk factors most frequently associated with EBV infection and PTLD.
The meta-analyses that we performed are presented in Figures 4 and 5. The pooled HR for the association between ATG use and post-transplant EBV infection obtained using the random-effect model was 5.26 (95% CI: 2.92-9.45) with an I 2 = 63.2% (Figure 4). We performed sensitivity analyses by recalculating the pooled estimate after excluding only one study at a time: the results vary between 4.13 and 6.49, and the I 2 heterogeneity statistic varies between 22% and 69%. The studies by Laberko et al. [69] and Liu et al. [74] had the greatest influence on the pooled estimate and on the level of heterogeneity. However, regardless of the study excluded, the overall result remains statistically significant. With respect to studies that estimated an adjusted OR to report the association between ATG and post-transplant EBV infection, the pooled estimate was 2.74 [1.03-7.31] and I 2 = 40.3% ( Figure 5). The sensitivity analyses highlighted a variation of the pooled estimate from 2.07 to 4.00 and of I 2 from 28% to 58%. The studies by Christopeit et al. [45] and Cohen et al. [46] had the greatest influence on the pooled estimate and heterogeneity. The pooled estimate was no longer significant if a single study was removed from the analysis, except for the study by Christopeit et al. [45], which was carried out with the smallest sample. The pooled RR for the association between ATG and PTLD obtained using the random-effect model was 4.17 (95% CI: 2.61-6.68) with an I 2 = 56.7%. The sensitivity analysis revealed that the pooled estimate ranged from 3.34 to 5.02 and the I 2 from 9% to 67%. The studies by Fujimoto et al. [55] and Van der Velden et al. [95] had the biggest influence on the pooled estimate and the I 2 . The sensitivity analysis did not question the statistically significant association between ATG and PTLD.
The results of these meta-analyses should be understood cautiously given the highlevel of heterogeneity observed between studies. Due to the small number of articles, we did not explore the sources of heterogeneity further by performing a subgroup analysis or a meta-regression.
Several studies used multivariate analysis to examine the relationship between recipient sex and post-transplant active EBV infection [4,35,39,50,52,56,65,68,69,72,74,76,85,86,88,89,96,101] or PTLD [16,27,40,56,88,94,95,98]; none found a significant association. Three out of six studies [35,53,56,65,79,101] that analyzed the association between donor sex and post-HSCT active EBV infection showed a statistically significant association but in the opposite direction. In two studies, the risk for active EBV infection post-HSCT was higher in patients receiving a male donor transplant [53,56] while, in the other, patients receiving a female donor transplant appeared to be at greater risk [65]. The only study [56] that explored the association between donor sex and PTLD did not find a statistically significant association. Moreover, no statistically significant association was found between the donor/recipient sex combination and post-transplant EBV infection [35,41,53,57,65]. Among all studies that examined the sex of the dyad donor/recipient and PTLD [57,82,94,98], only one [82] found a statistically significant association suggesting a higher risk of PTLD in patients who received a transplant from a different sex donor.

Discussion
This systematic review includes 77 papers. It aims to characterize risk factors associated with active post-transplant EBV infection and PTLD in HSCT recipients. Active EBV infection can result in rapidly increasing EBV-VL, which is a high-risk marker for PTLD development. Proper identification of the risk factors associated with active EBV infection and PTLD is needed for effective patient management.
In this systematic review, we focused on risk factors explored in published studies; very few statistically significant associations were found. The use of ATG was identified as one of the most important risk factors for the development of active post-transplant EBV infection and PTLD. The pooled relative risks estimated from the meta-analysis that was carried out confirmed a positive and statistically significant association between ATG and EBV infection (RR = 3.98 (95% CI: 2.20-7.18) and PTLD (RR = 3.69 (95% CI: 2.24-6.08)). ATG is a potent immunosuppressive agent that obliterates the T-cell pool [104][105][106], thereby enabling reactivation of latent EBV contained in mature B cells along with the malignant expansion of infected cells [104]. In the HSCT setting, ATG is used for the prevention of aGvHD, given its ability to target and deplete T lymphocytes [107,108].
Some studies included in this review also found an association with the presence of GvHD, which is an immune-mediated complication of HSCT whereby donor T cells present in the graft initiate an alloreactive process that ultimately causes destruction of host tissues [109]. aGvHD usually occurs within the first three months post-transplant and is categorized into four grades ranging from 1 (light disease) to 4 (severe disease) [110]. cGvHD usually occurs beyond the initial three months post-transplant. The pathophysiology of GvHD, especially that of cGvHD, is complex [111]. T and B lymphocytes are probably involved in the pathophysiology of GvHD, although the mechanism linking these cells to GvHD is not well-known [108]. In short, the etiology of GvHD is complex, and it is difficult to conclude whether GvHD is an independent risk factor for EBV and PTLD or whether the relationship found in some studies is the result of confounding by indication related to the use of ATG. The analyses that we are currently running among pediatric HSCT recipients recruited in our TREASuRE cohort study [112] confirm that EBV is strongly associated with ATG but not with GvHD, following adequate control for confounding bias.
Many other variables were analyzed in the 77 included studies, but results were either inconsistent, failed to find an association, or limited in terms of the number of studies that investigated the risk factor. Some studies showed that primary diagnosis was associated with post-transplant active EBV infection [23,39,88], more specifically in the case of Hodgkin disease [23,88]. Some forms of Hodgkin's lymphoma are etiologically linked to EBV [113,114] and may occur in individuals who are not able to properly control EBV infection. These individuals may be thought to be more susceptible to other EBV diseases (such as post-transplant active EBV infection) along the continuum of care, but HSCT should have corrected any immune cell problem. Although interesting, further studies are needed to confirm the potential association between Hodgkin's disease and post-transplant active EBV infection in HSCT patients.
Discordant results were found for other variables, and, in other cases, the number of studies investigating risk factors was limited. These variables are recipient age, recipient gender, donor type, conditioning regimen, graft source, graft history, graft content (CD34 + , CD3 + , CD8 + , CD3 + /CD8 + ), genotype (IFNGgene 3/3, CCR5), splenectomy, mesenchymal stromal cells, donor gender and transfusion (red blood cells, platelets, plasma) (Table S4). In our recent study, although no relationship was statistically found between EBV and blood product transfusion, we linked a case of EBV infection in an EBV-seronegative pediatric HSCT recipient to a blood donor through viral genotype analysis [112]. One cause of discordant results is the heterogeneity observed among the various studies, most notably with regard to the different specimen types used to perform PCR tests (Table S3). The sensitivity of PCR tests is greater when whole blood is used as opposed to plasma [115]. Other sources of discordance include variations in the statistical approach and experimental design. We also noted the absence of controls for confounding and failure to report results when associations lacked statistical significance. In addition, only 42.9% of studies included in the systematic review of factors associated with post-transplant EBV infection were classified as being of strong quality, and 23.8% were classified as moderate quality; with respect to PTLD as an outcome, the proportions were, respectively 57.1% and 14.3%. An important risk for bias includes uncontrolled confounding bias and the lack of information on retention, a potential source of selection bias in cohort studies.
This review was not able to discern whether differences exist between children and adults. While statistical power was higher in studies combining both groups, differences in terms of risk factors may exist. Immune restoration through T-cell reconstitution after transplantation is different in children and adults [116], and risk factors may differ. It should be noted that 25 of the 77 studies selected in this systematic review have a sample size of less than 100; therefore, it is possible that type II error may explain why positive associations were not statistically significant in many studies. Moreover, the included studies were limited to the identification of factors associated with the first occurrence of active EBV infection post-transplant, although during follow-up a patient may experience several episodes of active EBV infection [23,116]. This latter aspect should be considered in order to better understand the dynamics of the evolution of active EBV infection posttransplant in HSCT recipients. Risk factors for the occurrence of active EBV infection may be different from those that explain the dynamics of infection. Finally, there was insufficient information on attrition, which may be the primary source of selection bias in this type of study. While we initially intended to perform a meta-analysis of all risk factors associated with active EBV infection and PTLD, this was not possible because of the diversity of outcome definitions, the variability in the definition of risk factors and the non-systematic reporting of point estimates, confidence intervals and p-values. However, as indicated above, a meta-analysis was carried out to measure the association between ATG use and post-transplant EBV infection and PTLD, respectively. The results, however, must be considered with caution, as the definition of outcome was quite variable from one study to another. Based on all the above arguments, further studies using large cohorts of children and adults are needed to better elucidate the determinants of active EBV infection and PTLD among HSCT recipients.
In conclusion, we found ATG as the most important risk factor for the development of active post-transplant EBV infection and PTLD in HSCT patients. ATG considerably increases the risk of EBV and PTLD. Other risk factors have been linked with EBV and PTLD in studies, such as GvHD or type of donor, but the association for these other factors is less clear due to conflicting results, the potential for bias, particularly confounding, or because of the low number of studies that considered these risk factors. Further studies using large cohorts of children and adults with appropriate control for confounding are needed to better characterize other determinants of active EBV infection and PTLD among HSCT recipients.
Supplementary Materials: The following are available online at https://www.mdpi.com/2076-3 93X/9/3/288/s1, Table S1: Search equation used in Medline and EMBASE, Table S2a: Component ratings of study (a modified version of the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies, Table S2b: Results of the quality evaluation of the 77 articles included in this systematic review, Table S3: Characteristics of the 77 studies included in the systematic review and Table S4 Funding: This study was supported by a grant from the Canadian Blood Services (grant CBS #201209-293922) to HT, CB, MD, JL, CA, as well as by an operating grant from the Cancer Research Society and the C17 Council (OG #23276) to CA for review of the pediatric studies. PREB was supported by doctoral awards from the CHU Sainte-Justine and Université de Montréal (tuition fee waiver scholarships for international students and graduate scholarship for the final year of PhD studies). HT holds a salary award (Chercheur-Boursier, Junior-2) from the Fonds de la recherche du Québec en santé (FRQ-S) and from Canadian Institutes of Health Research (CIHR) (New investigator salary award).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable (we did a systematic review and meta-analysis of published studies).
Data Availability: We confirm that our data are available.

Acknowledgments:
We thank Sylvie Fontaine, librarian at the paramedical library of the Université de Montréal, for her support in the identification of descriptors and keywords in the Medline and EMBASE databases.

Conflicts of Interest:
The authors of this manuscript have no conflicts of interest to disclose.