Association between kinetic of anti-neutrophil cytoplasmic antibody (ANCA), renal survival and relapse risk in ANCA glomerulonephritis

ABSTRACT

Background

Anti-neutrophil cytoplasmic antibody (ANCA) kinetic in ANCA-associated vasculitis with glomerulonephritis (AAV-GN) has been suggested to be associated with AAV relapse. Few studies have focused on its association with renal prognosis. Thus we aimed to investigate the relationship between ANCA specificity and the evolutive profile and renal outcomes.

Methods

This multicentric retrospective study included patients diagnosed with ANCA-GN since 1 January 2000. Patients without ANCA at diagnosis and with fewer than three ANCA determinations during follow-up were excluded. We analysed estimated glomerular filtration rate (eGFR) variation, renal-free survival and relapse-free survival according to three ANCA profiles (negative, recurrent and persistent) and to ANCA specificity [myeloperoxidase (MPO) or proteinase 3 (PR3)].

Results

Over a follow-up of 56 months [interquartile range (IQR) 34–101], a median of 19 (IQR 13–25) ANCA determinations were performed for the 134 included patients. Patients with a recurrent/persistent ANCA profile had a lower relapse-free survival (P = .019) and tended to have a lower renal survival (P = .053) compared with those with a negative ANCA profile. Patients with a recurrent/persistent MPO-ANCA profile had the shortest renal survival (P = .015) and those with a recurrent/persistent PR3-ANCA profile had the worst relapse-free survival (P = .013) compared with other profiles. The negative ANCA profile was associated with a greater eGFR recovery. In multivariate regression analysis, it was an independent predictor of a 2-fold increase in eGFR at 2 years [odds ratio 6.79 (95% confidence interval 1.78–31.4), P = .008]).

Conclusion

ANCA kinetic after an ANCA-GN diagnosis is associated with outcomes. MPO-ANCA recurrence/persistence identifies patients with a lower potential of renal recovery and a higher risk of kidney failure, while PR3-ANCA recurrence/persistence identifies patients with a greater relapse risk. Thus ANCA kinetics may help identify patients with a smouldering disease.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

Anti-neutrophilic cytoplasmic antibody (ANCA)-associated vasculitis (AAV), mainly represented by granulomatosis with polyangiitis (GPA) and microscopic polyangiitis (MPA) [1], are rare autoimmune diseases characterized by necrotizing inflammation of small vessels. Consequently, and since overlaps are frequent [2], patients with GPA or MPA are often pooled in clinical trials. ANCAs directed against myeloperoxidase (MPO-ANCA) or proteinase-3 (PR3-ANCA) are diagnostic biomarkers present at AAV onset in most patients [3]. Recent clinical and experimental data have demonstrated that AAV genetic background, clinical features and prognosis are more closely related to ANCA specificity (MPO- or PR3-ANCA) than clinical classifications (GPA or MPA) [4–9]. Antibody subtypes now have a preponderant place in newly updated classifications [10, 11]. Kidney involvement in AAV, characterized by pauci-immune necrotizing and crescentic glomerulonephritis (AAV-GN), is frequent, sometimes life-threatening at presentation and adversely affects long-term renal and patient survival [8, 12–14].

The combination of high-dose steroids with cyclophosphamide or rituximab has improved prognosis from a nearly always fatal disease to a survival rate of ∼80% at 5 years [15, 16]. Thus AAV is now considered to be a chronic disease with a risk of relapse and of progression to kidney failure (KF) of up to 50% and 30% at 5 years, respectively [8, 17, 18]. As a consequence of greater survival, patients tend to experience treatment-related adverse events (cardiovascular, infectious and neoplastic complications), adding mortality and morbidity to that of the vasculitis itself [19]. While remission can now be achieved in >90% of patients [20], usually sustained over a prolonged period, there is a need to identify patients at higher risk of relapse, as they would benefit from standard or even extended maintenance immunosuppressive treatment, and who achieve rapid remission without disease-related complications or those who reach KF that would theoretically take advantage of immunosuppressive drug minimization.

To date, there is no reliable biomarker for predicting relapse and/or guiding the duration of immunosuppression to decrease long-term drug toxicity and related morbidity. The relevance of longitudinal monitoring of ANCA in assessing disease activity remains a matter of debate [21, 22]. In some studies, reappearance, increase or persistence of ANCA was associated with future relapse [23–33], especially in patients with PR3-ANCA [32–34] or with ANCA-GN [35]. It was also suggested that ANCA persistence could reflect a smouldering disease responsible for insidious deterioration of renal function [36–38]. However, these studies were often limited by their size or their monocentric design.

We hypothesized that the ANCA evolutive profile over time (ANCA kinetic) may be useful for identifying patients at higher risk of KF or relapse. The aim of this multicentre study was to investigate the link between ANCA kinetic and outcome in AAV-GN patients, with a focus on renal function and survival.

MATERIALS AND METHODS

Selection of patients

This multicentre retrospective study included adult patients with AAV-GN from the Maine-Anjou AAV registry and Tours and Poitiers University Hospitals. The Maine-Anjou AAV registry [39–42] includes all successive AAV-GN patients diagnosed between 2000 and 2019 in four nephrology departments from central western France (Angers University Hospital and Le Mans, Cholet and Laval Hospitals). AAV-GN patients from Tours and Poitiers University Hospitals were consecutively diagnosed between 2003 and 2020 and also retrospectively included.

The AAV diagnosis was based on the revised 2012 Chapel Hill Consensus Conference [1]. The AAV-GN diagnosis was assessed on active renal involvement [active urinary sediment with haematuria, protein:creatinine ratio >0.3 g/g and/or impaired renal function with an estimated glomerular filtration rate (eGFR) <60 ml/min] and, in most cases, confirmed by kidney biopsy showing pauci-immune glomerulonephritis. Included patients were newly diagnosed with AAV [130/134 (97%)] or were experiencing an AAV relapse and renal involvement [4/134 (3%)]. Patients with negative ANCA at diagnosis, undergoing kidney replacement therapy prior to the time of diagnosis or without a complete ANCA profile (see definition below) or with fewer than three ANCA determinations after diagnosis were excluded from the study.

Clinical, biological and histological data collection

Collected data included patients’ baseline characteristics, significant past medical history and organ involvement at AAV diagnosis. Therapeutic management was also retrieved. The AAV activity score was determined using the Birmingham Vasculitis Activity Score (BVAS) 2003 [43]. Baseline biological data included complete blood count, ANCA and renal parameters (serum creatinine, proteinuria, haematuria). Histopathological classification [13] was assessed for every kidney biopsy when available. In a subgroup of patients, pro-inflammatory molecules [interleukin-6 (IL-6), complement 5a (C5a), tumour necrosis factor α (TNF-α) and pentraxin-3 (PTX3)], assessed by a bead-based multiplex assay according to the manufacturer's instructions (Bio-Techne, Minneapolis, MN, USA), and C-reactive protein (CRP) were measured at diagnosis and 6 months later.

ANCA was determined according to local practices by indirect immunofluorescence (IIF) with ethanol and formol fixed neutrophils and by at least one quantitative enzyme immunoassay (EIA) for PR3 and MPO specificity. The main EIA assays used in each centre are detailed in Supplementary Table S1.

Definitions

KF was defined as the need for kidney replacement therapy (KRT) for >3 months. The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [44]. The eGFR percentage variation over time was calculated as follows: (eGFR at a given time − eGFR at diagnosis)/eGFR at diagnosis. Patients undergoing KRT were considered to have an eGFR of 5 ml/min. Relapse was defined as the recurrence or renewed appearance of organ involvement attributable to AAV activity with a BVAS increase (>1) and requiring an increase in the steroid dose or introduction of a novel immunosuppressive drug [45]. Only the first relapse was considered. ANCA positivity was not mandatory to define a relapse.

ANCA profiles

ANCA was determined by IIF or EIA as positive or negative according to the positive threshold set by the manufacturer (or ≥1/20 for IIF). Uninterpretable or atypical data were not considered. A complete ANCA profile was defined if the ANCA test was positive at diagnosis, then performed at least once at 6–12 months and once at 12–24 months. Thus the minimal follow-up time for a patient included was at least 12.1 months.

Three longitudinal ANCA profiles were then defined: a negative profile, whereby all ANCA values were negative after 6 months since diagnosis; a persistent profile, where ANCA remained positive after 6 months and was never followed by a negative value within 24 months or where ANCA was positive after 6 months and then negative later on (late-onset negative patients); and recurrent profile, where there was at least one ANCA negative value after 6 months followed by a positive value within 24 months. In case of ANCA profile discordance by IIF or EIA, priority was assigned as follows: positive, recurrent and negative ANCA profile. The ANCA profile flowchart attribution is illustrated in Supplementary Fig. S1. ANCA status after 24 months was not considered.

Statistical analysis

Patients’ characteristics are reported as numbers and percentages for qualitative variables and median [interquartile range (IQR)] for continuous variables. Continuous data are shown as boxplots (median ± IQR). Vertical bars represent minimum and maximum values. Data were compared using the chi-squared test (or Fisher’s exact test if necessary) for categorical variables and the Mann–Whitney test for continuous variables.

The Kaplan–Meier method was used to estimate the time to events according to ANCA specificity and/or profile. A logrank test was used to compare the survival curves between groups. Due to the relatively small number of events beyond 5 years, survival analysis was censored at 5 years.

A logistic regression analysis was performed to examine factors associated with the chance of eGFR doubling at 2 years. Multivariable logistic regression analysis included all parameters with P < .1 in the univariate analysis or parameters judged as confounding or clinically relevant. To optimize the multivariable model, variables were selected using manual step-by-step backward selection with a removal criterion of P > .1 [46]. Odds ratios (ORs) with 95% confidence intervals (Cis) are reported.

No imputation of missing data was performed. Statistical analysis was performed using R software version 4.0 (R Foundation for Statistical Computing, Vienna Austria).

Ethical issues

The Maine-Anjou Registry was authorized by the Commission National Informatique et Liberté (CNIL; agreement 2018-MR03-02). Participants gave their written informed consent for the use of their biological material through our Biological Resources Center (authorization AC-2017-2993). This study was approved by the local ethics committee of Angers University Hospital (CE 2021-208).

RESULTS

Baseline characteristics of the cohort

Among the 265 patients from the Maine-Anjou Registry, Tours and Poitiers University Hospitals, 134 patients with ANCA-GN were included in the present study (Supplementary Fig. S2). The median age was 67 years (IQR 55–74)] with a predominance of males (61%). The median BVAS was 15 (IQR 12–20). After kidney involvement, lung and ear–nose–throat (ENT) were most frequently involved (46% and 34%, respectively). At diagnosis, the median serum creatinine was 230 μmol/l (IQR 139–392) and the median eGFR was 23 ml/min (IQR 11–39). Most patients underwent a kidney biopsy (95%). According to Berden's histopathological classification, 30% of patients had a focal class, 35% a crescentic class, 17% a mixed class and 17% a sclerotic class. Detailed baseline patient characteristics are presented in Table 1 and Supplementary Table S2.

Cyclophosphamide (79%) was mainly used as an induction regimen, while rituximab was the main drug given as maintenance therapy (62%). After a median follow-up of 56 months (IQR 34–101), 39 patients (29%) experienced a relapse, most commonly involving the kidneys (62% of patients). Twenty patients (15%) required early KRT (within 30 days of diagnosis), 28 (21%) reached KF and 18 (14%) died. Detailed therapeutic management information and outcomes are presented in Table 2 and Supplementary Table S2.

Outcomes according to ANCA specificity at diagnosis

At diagnosis, 100 patients (75%) were MPO-ANCA positive and 34 (25%) were PR3-ANCA positive. Time to KF was significantly lower in patients with MPO-ANCA at diagnosis compared with PR3-ANCA patients (P = .046; Fig. 1A) and there was a trend towards a longer time to relapse for MPO-ANCA patients compared with PR3-ANCA patients (P = .071; Fig. 1B).

ANCA profiles

Overall, when considering both IIF and EIA methods, there was a median of 19 (IQR 13–25) ANCA determinations per patient (Supplementary Fig. S3A). A total of 132 patients (98.5%) underwent at least six ANCA determinations during follow-up (Supplementary Fig. S3B). Twenty-eight patients were classified with a negative ANCA profile, 14 with a recurrent ANCA profile and 92 with a persistent ANCA profile (including 8 late-onset negative patients). Since patients with recurrent and persistent ANCA profiles showed similar event-free survival, considering either relapse-free or renal survival (Supplementary Fig. S4), these subgroups were merged into a single group, hereafter named ‘recurrent/persistent’. The frequency of ANCA determination in the negative or recurrent/persistent groups was similar, regardless of the follow-up period studied (Supplementary Fig. S3A). Of note, eight patients were considered as late-onset ANCA negative (i.e. ANCA positive at 6 months and negative between 6 and 24 months). Given the length of time these patients remained ANCA positive [median 12.5 months (IQR 11.7–15.3)], and event-free survival was similar to the patients with a ‘recurrent/persistent’ profile (data not shown), we considered that it was more clinically and pathophysiologically relevant to regroup these patients together. According to the ANCA profile, clinical features, biological characteristics, renal histology and therapeutic management at diagnosis were similar between the two groups (Table 1 and Table 2).

Renal survival according to ANCA profiles and specificity

Renal survival was lower in patients with a recurrent/persistent profile when compared with the negative profile group (P = .053; Fig. 2A). This remained true after the exclusion of patients with renal relapses (Supplementary Fig. S5). The proportion of patients who reached KF within 5 years from diagnosis was significantly higher in those with a recurrent/persistent profile as compared with those with a negative profile (20% versus 3.6%; P = .044; Table 1).

After stratification of patients according to ANCA specificity, patients with recurrent/persistent MPO-ANCA had a lower renal survival when compared with negative MPO-ANCA patients (P = .026; Fig. 2B). Renal survival did not seem influenced by ANCA profile in PR3-ANCA patients (P = .46; Fig. 2C). The proportion of patients who reached KF within 5 years was higher in those with a recurrent/persistent MPO-ANCA profile compared with patients with a negative MPO-ANCA profile (25% versus 0%; P = .011). Among PR3-ANCA patients, the prevalence of KF between recurrent/persistent and negative profiles did not reach statistical significance (4% versus 11%; P = .50) (Supplementary Table S2).

When considering both ANCA profiles and ANCA specificity, a recurrent/persistent MPO-ANCA profile had the worst renal survival and negative MPO-ANCA and PR3-ANCA patients had similar renal survivals (P = .015; Fig. 2D). However, this last result is to be interpreted with caution, as MPO-ANCA patients had worse renal function and more chronic/sclerotic lesions on kidney biopsy (Supplementary Table S3).

eGFR improvement over time according to ANCA profiles and specificity

Patients with MPO-ANCA had significantly lower eGFR values at diagnosis and during follow-up when compared with PR3-ANCA patients (Supplementary Fig. S6A). During the first 2 years following diagnosis, renal function improved in a similar manner in MPO- and PR3-ANCA patients (Supplementary Fig. S6B).

The eGFR level at diagnosis was comparable between the negative and recurrent/persistent profile groups (Table 1 and Supplementary Fig. S6C). Patients were more likely to undergo a 2-fold increase in eGFR at 6 months, 1 year and 2 years if they had a negative ANCA profile as compared with a recurrent/persistent ANCA profile (P = .02, .06 and .006, respectively; Fig. 3).

In univariable logistic regression analysis, factors associated with eGFR doubling at 2 years were male gender, heart involvement, eGFR at diagnosis, percentage of crescentic glomeruli and sclerotic glomeruli on kidney biopsy and negative ANCA profile (Supplementary Table S4). In multivariable linear regression analysis, a negative ANCA profile was an independent predictor of a 2-fold increase in eGFR at 2 years [OR 6.79 (95% CI 1.78–31.4), P = .008; Supplementary Table S5 and Fig. 4).

Patients with a recurrent/persistent profile had a lower increase in eGFR when compared with those with a negative profile at 6 months [+25% (IQR 0–94) versus +88% (IQR 11–231), P = .019], 1 year [+38% (IQR 0–101) versus +71% (IQR 17–283), P = .067] and 2 years [+36% (IQR 0–105) versus +110 (IQR 17–267), P = .039] (Supplementary Fig. S6D). According to the ANCA specificity and profile, there was a trend for a lower recovery in renal function in patients with a recurrent/persistent MPO-ANCA profile at 6 (P = .037), 12 (P = .079) and 24 months (P = .095) (Supplementary Fig. S6E), while the ANCA profile in PR3-ANCA patients did not seem to have a significant effect on renal function recovery (P > .20; Supplementary Fig. S6F).

Relapse-free survival according to ANCA profiles and specificity

Relapse-free survival was lower in patients with a recurrent/persistent profile compared with patients with a negative profile (P = .019; Fig. 5A). The proportion of patients who experienced a relapse during follow-up was significantly higher in those with a recurrent/persistent profile compared with those with a negative profile (34% versus 11%, P = .016; Table 1).

After stratification of patients according to ANCA specificity, patients with recurrent/persistent PR3-ANCA had a trend for a lower relapse-free survival when compared with those with negative PR3-ANCA profile (P = .075; Fig. 5C). Relapse-free survival was less influenced by the ANCA profile in MPO-ANCA patients (P = .091; Fig. 5B). The proportion of patients who experienced a relapse during follow-up was higher in those with a recurrent/persistent PR3-ANCA profile compared with patients with a negative MPO-ANCA profile (60% versus 11%, P = .019), but did not appear different compared with those with recurrent/persistent and negative MPO-ANCA profiles (26% versus 11%, P = .20) (Supplementary Table S2).

When considering ANCA specificity and profiles together, recurrent/persistent PR3-ANCA had the lowest relapse-free survival (P = .013; Fig. 5D).

Pro-inflammatory markers according to ANCA profiles

In a subgroup of patients with available biobank (n = 35), we assessed pro-inflammatory molecules at diagnosis and 6 months later. Compared with patients with no available biobank, there was a predominance of female, higher BVAS, more frequent use of plasma exchange in induction therapy and more frequent use of rituximab as maintenance therapy (Supplementary Table S6) in this subgroup.

After 6 months from diagnosis, compared with patients with a negative ANCA profile, patients with a recurrent/persistent ANCA profile had significantly higher CRP levels [6 (IQR 3–15) versus 4 (3–4), P = .031] and a lower CRP decrease from diagnosis [−17 (IQR −63 to −3) versus −74 (−176 to −37), P = .007] (Supplementary Fig. S7).

We also found a non-significant trend for higher baseline levels of other pro-inflammatory molecules (IL-6, C5a, TNF-α and PTX3) with a lower decrease after 6 months from diagnosis (data available for 35 patients, 10 in the negative profile group, 25 in the recurrent/persistent profile group; Supplementary Fig. S7).

DISCUSSION

In this retrospective multicentre cohort of 134 AAV-GN patients, we found that the ANCA profile correlated with renal-free and relapse-free survival, but distinctively according to ANCA specificity. A recurrent/persistent MPO-ANCA profile was associated with renal function evolution over time and KF, while a recurrent/persistent PR3-ANCA profile was associated with relapse. Regardless of specificity, patients with a negative ANCA profile displayed better outcomes.

First, we found that patients with a recurrent/persistent ANCA profile, compared with those with a negative ANCA profile, had a higher risk of KF, whereas the severity of renal involvement (eGFR and need for KRT) and histologic lesions at diagnosis, which are well-known renal prognostic factors [12, 13, 17], were similar. Importantly, we observed that patients with a negative ANCA profile had a better renal recovery with a higher probability of doubling their eGFR. Interestingly, outcomes were differently affected by ANCA specificity. Indeed, eGFR variation and renal survival were correlated with the ANCA profile only in MPO-ANCA patients. This is consistent with two recent studies. Oristrell et al. [37] reported an increased risk of worsening renal function associated with a recurrent/persistent ANCA profile in 52 MPO-ANCA patients. Aljuhani et al. [38] found that serological remission (ANCA negativity) was associated with a lower risk of KF in 202 patients with either MPO- or PR3-AAV-GN. Moreover, Trivoli et al. [47] recently described a subgroup of 41 patients (5%) among 856 MPA or GPA patients who presented with slowly progressive ANCA-GN, in contrast to the typical rapidly progressive pattern. Interestingly, all 41 patients had MPA and 38/39 with available ANCA determination had MPO-ANCA positivity, suggesting that MPO-ANCA, unlike PR3-ANCA, may be associated with a smouldering renal disease. Of course, worsening of renal function may also be attributed to factors other than AAV alone, such as treatment-related adverse effects (e.g. cotrimoxazole-induced nephritis, post-chemotherapy-induced acute kidney injury and steroid-induced atherosclerosis) or even secondary focal segmental glomerulosclerosis.

Second, the ANCA profile also correlated with relapse-free survival, but this appeared to be also influenced by ANCA specificity and mainly affected PR3-ANCA patients. This point has been documented extensively in the literature: the reappearance or persistence of ANCA is associated with a higher risk of relapse, especially in GPA/PR3-ANCA patients [32, 34, 37, 38, 48–51], even if a recent meta-analysis challenged this view [22]. Nevertheless, with a strong methodology, Kemna et al. [35] demonstrated that an increase in ANCA level was strongly associated with subsequent relapse, particularly in patients with kidney involvement and/or PR3-ANCA.

Third, according to ANCA specificity at diagnosis, we found that MPO-ANCA patients had a worse final renal function with a lower likelihood of renal survival and that PR3-ANCA patients had a lower relapse-free survival. These data are in keeping with previous studies [52, 53] that suggested that more severe kidney scarring and sclerotic lesions at MPO-ANCA-GN diagnosis could explain these differences [54–56].

To summarize these three different points, ANCA specificity and profile affects clinical outcomes differently. Those differences could be explained by different genetic predispositions [57], mechanisms leading to ANCA development [58], variable cytokine profile [59], specific endothelial damage induced by MPO or PR3 internalization [60], differences in complement activation [61] and the chronicity of lesions observed on the first kidney biopsy at diagnosis [56] between MPO- or PR3-AAV. Accumulating evidence shows that ANCAs are pathogenic by themselves, especially MPO-ANCA [62, 63]. The persistence of ANCA, even at low levels, could therefore contribute to an ongoing smouldering disease or persistent inflammation as a continuum between full biological remission and clinical active disease. Therefore recurrent/persistent ANCA may indeed induce the chronic and sustained low-grade inflammation responsible for CKD progression (MPO-ANCA) or AAV relapse (PR3-ANCA).

This is suggested by higher levels and a lower decrease in pro-inflammatory molecules at 6 months from ANCA-GN diagnosis in a subgroup of patients with a recurrent/persistent ANCA profile as compared with a negative ANCA profile.

These observations raise the issue of the therapeutic goals to achieve in ANCA vasculitis. Whether serological remission should represent the main objective of treatment in AAV (treat-to-target strategy as in other autoimmune diseases such as systemic lupus erythematosus, anti-PLA2R membranous nephropathy or Goodpasture disease [64–66]) or only identifies a subset of patients more sensitive to immunosuppressive drugs is unknown. Consequently, the relevance of modulating immunosuppression according to the ANCA profile, i.e. intensification in patients with ANCA persistence and reduction in those with a negative profile, remains to be established.

Our study has several limitations, mostly due to its observational design. Because this study included only patients with AAV-GN, our results may not be generalized to all AAV. Therapeutic management was not standardized and evolved over time according to contemporary recommendations. The frequency of ANCA assessment was not standardized and the ANCA profile was evaluated according to a qualitative result (positive or negative) since detection methods changed over time. The ANCA profile was assigned according to ANCA determinations from the first 24 months (because very few were retrieved after 36 months) but not from the entire follow-up duration (in opposition to events). This is congruent with the standard duration of therapy, including induction and maintenance regimen. Moreover, subgroup analysis was limited given the small number of patients. Indeed, the low rate of progression to KF (null in the negative MPO-ANCA profile group) precluded the building of a strong, fully adjusted, multivariate Cox model to evaluate the predictive value of recurrent/persistent ANCA. Despite these limitations, ANCA profile attribution using different immunoassays easily identifies a particular subset of patients in clinical practice. Lastly, due to limited serum samples, assessment of the levels of pro-inflammatory molecules was available in only a small subgroup of the cohort and deserves further evaluation in larger populations.

In conclusion, our findings support and add to the literature on the relevance of ANCA monitoring during patient follow-up. Although these findings need to be confirmed in prospective studies, the persistence or reappearance of ANCA beyond 6 months should alert clinicians not only to the risk of relapse, but also to a smouldering renal disease that might progress to KF. While waiting for better performing biomarkers [67, 68], frequent assessment of ANCA during follow-up is still useful to foresee patients’ outcomes. Whether the ANCA profile may support tailoring therapeutic management remains to be investigated.

ACKNOWLEDGEMENTS

The authors thank Michael Wood for editorial input on the quality of the language.

AUTHORS’ CONTRIBUTIONS

J.F.A. and B.B. designed the study. C.S., C.M., A.D., P.J. and B.B. gathered the information included in the database. C.S., J.F.A. and B.B. analysed the data. C.S. wrote the first draft of the manuscript. G.B.P., P.G., F.B., J.F.A. and B.B. revised the manuscript. All authors participated in patient care and approved the final version of the manuscript.

DATA AVAILABILITY STATEMENT

The data underlying this article will be shared upon reasonable request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflicts of interest in relation to this article.

REFERENCES

Author notes