Objective Histological remission is being increasingly acknowledged as a therapeutic endpoint in patients with UC. The work hereafter described aimed to evaluate the concordance between three histological classification systems—Geboes Score (GS), Nancy Index (NI) and RobartsHistopathologyIndex (RHI), as well as to evaluate their association with the endoscopic outcomes and the faecal calprotectin (FC) levels.
Design Biopsy samples from 377 patients with UC were blindly evaluated using GS, NI and RHI. The results were compared with the patients’ Mayo Endoscopic Score and FC levels.
Result GS, NI and RHI have a good concordance concerning the distinction between patients in histological remission or activity. RHI was particularly close to NI, with 100% of all patients classified as being in remission with NI being identified as such with RHI and 100% of all patients classified as having activity with RHI being identified as such with NI. These scores could also predict the Mayo Endoscopic Score and the FC levels, with their sensitivity and specificity levels depending on the chosen cut-offs. Moreover, higher FC levels were statistically associated with the presence of neutrophils in the epithelium, as well as with ulceration or erosion of the intestinal mucosa.
Conclusions GS, NI and RHI histopathological scoring systems are comparable in what concerns patients’ stratification into histological remission/activity. Additionally, FC levels are increased when neutrophils are present in the epithelium and the intestinal mucosa has erosions or ulcers. The presence of neutrophils in the epithelium is, indeed, the main marker of histological activity.
- inflammatory bowel disease
- stool markers
- ulcerative colitis
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Histological remission is being increasingly acknowledged as the ultimate therapeutic goal in patients with ulcerative colitis (UC).
There are several different histological indexes for UC assessment.
Histological activity is associated with the faecal calprotectin (FC) levels in UC.
What are the new findings?
The classical Geboes Score (GS) and the newly developed and validated Nancy Index (NI) and Robarts Histopathology Index (RHI) are well correlated in terms of remission/activity.
These scores can be predictive of the Mayo Endoscopic Score and the FC levels, with sensitivity and specificity levels depending on the cut-offs used.
The FC elevation seen in histologically active patients is mainly due to the presence of neutrophils in the epithelium, and the mucosa ulceration and erosion.
How might it impact on clinical practice in the foreseeable future?
The similarities between the different histological indexes should be taken into account when assessing disease activity in patients with UC in clinical trials or during daily clinical practice.
FC levels may be used to predict specific histological features.
Neutrophils in the epithelium are the main marker of histological activity.
UC is an idiopathic autoimmune inflammatory disease characterised by periods of remission intertwined with disease exacerbations or flares. Traditionally, UC activity is monitored through clinical and endoscopic parameters. Mucosal healing, that is, the absence (or near absence) of macroscopic lesions in the GI mucosa, was for many years considered to be the treatment target and has been associated with improved long-term outcomes (including fewer hospitalisations and less surgeries). However, the presence of an apparently healthy mucosa at colonoscopy does not exclude the presence of long-standing microscopic inflammation.1 Importantly, this persistent disease activity at the histological level has been related to clinical relapses, future surgery and progression to colorectal neoplasia.1–4 As so, histopathological analyses have become an important part of UC assessment both in clinical practice and in research trials. To make these analyses comparable, numerous scoring systems have been developed and used in the last decades. Still, most of them were not formally validated, which limits their use.1 5 This manuscript is focused on three histological indices that have been at least partially validated: Geboes Score (GS), Nancy Index (NI) and Robarts Histopathology Index (RHI).
GS remains up to today the most commonly used histological index in UC: although not formally validated, this scoring system has been found to have good reproducibility after interobserver agreement analyses.6 GS is a progressive classification system based on the existence of six major grades: 0—structural changes; 1—chronic inflammation; 2—lamina propria neutrophils; 3—neutrophils in the epithelium; 4—crypt destruction and 5—erosion and ulcers. Each of these grades is subdivided into four categories.6 Recently, a simplified version of this score has been developed, aiming to facilitate its application to the daily clinical practice.7
The NI is one of the two histological scoring systems that has been validated so far. It was developed by multiple linear regression and a bootstrap process (to match the Global Visual Evaluation (GVE)) that yielded a five-level classification: 0—absence of significant histological disease; 1—chronic inflammatory infiltrate with no acute inflammatory infiltrate; 2—mildly active disease; 3—moderately active disease and 4—severely active disease.8 9 The classification in one of these five grades depends on the presence/absence of three items: ulceration, acute inflammatory cells infiltrate and chronic inflammatory cells infiltrate.8 9 The main advantages of this index are its practicality, high intraobserver and interobserver reliability and responsiveness to change.8
The RHI has also been fully validated: this scoring system was developed after a model-building process followed by a step-down procedure resampled by bootstrap.10 RHI score ranges from 0 (no disease activity) to 33 (severe disease activity) based on the evaluation of four main parameters: chronic inflammatory infiltrate, lamina propria neutrophils and eosinophils, neutrophils in the epithelium and erosion and ulceration.10 Overall, this histological index has been reported to be reproducible, responsive and valid.10
Besides being an important feature in all UC-associated histopathological indexes, the presence of neutrophils in the GI mucosa leads to the release of these cells’ cytosolic proteins into the lumen. One of these proteins—calprotectin—is a calcium-binding and zinc-binding protein that accounts for approximately 60% of the neutrophils’ cytosolic proteins. Its resistance to bacterial degradation, evenly distribution and stability in stools (up to 1 week at room temperature) makes this protein an accurate surrogate marker of the influx of neutrophils into the intestinal mucosa.11–14 In fact, several different studies have shown that faecal calprotectin (FC) levels can be used to distinguish organic from non-organic intestinal disorders (eg, irritable bowel syndrome) to monitor UC activity and response to therapy and to predict relapses, need for colectomy and postoperative recurrences.11–22
In this work, we have classified a cohort of patients with UC (mostly asymptomatic) according to the three partially or fully validated histological indices: GS, NI and RHI. Our aims were: (1) to compare these different scoring systems as a mean to assess microscopic inflammation; (2) to relate these scoring systems with the patients’ endoscopic outcomes and inflammation burden and (3) to unveil which (if any) histological features could be predicted from FC levels.
Material and methods
Patients and study design
Patients with UC previously recruited for other studies (Accuracy of Calprotectine in Evaluating sub-clinical inflammation in Ulcerative colitis (ACERTIVE) and Correlation between IFX Serum levels and anti-TNF antibody therapy (ATIs) in different pharmacokinetics times and Endoscopic healing (CISAE)) were included in this study whenever they had available all the pertinent information (namely Mayo Endoscopic Score, FC levels and histological information).23 24 These patients had been recruited from 14 universities and community hospitals. Inclusion and exclusion criteria have been previously described.23 24 Topical treatment was not allowed in the 15 days prior to recruitment.
This study was approved by the ethics committee of all hospitals involved and by the Portuguese Data Protection Authority. All patients enrolled have signed an informed written consent.
All patients underwent a sigmoidoscopy. Endoscopic activity was evaluated with the Mayo endoscopic subscore: mucosal healing was defined as a Mayo endoscopic subscore equal to 0 or ≤1.25
Rectum and sigmoid biopsies (two samples of each) were taken during the sigmoidoscopies and subjected to central reading by two independent pathologists who were blinded regarding patients' disease status and endoscopic results. Patients’ scores were attributed based on the sample with the most severe involvement. Readings concerning the different scores were made at least 15 days apart, and the pathologists were unaware of the previous classifications given to each sample. Disagreements between the pathologists were resolved in the presence of a third pathologist (KG) using a multiheaded microscope to make the decision about the final score. Cut-offs for disease remission/activity concerning each score were based on the literature.6 8–10
Stools were collected within 3 days of the sigmoidoscopy procedure and kept at room temperature for up to 7 days. Extraction was performed using the ‘Faecal sample preparation kit’ (Roche Diagnostics), according to the manufacturers’ instructions, and afterwards samples were stored at −80°C. FC levels were quantified using two different assays: ‘Quantum Blue’, hereafter referred to as QB (Buhlmann, Switzerland) and automated fluoroimmunoassay-EliA test, hereafter referred to as EliA (Phadia, ThermoFisher, Sweden).
Categorical variables were described through absolute (n) and relative (%) frequencies and continuous variables were described as mean and SD, median, IQR and minimum/maximum values, whenever appropriate. When testing hypothesis concerning continuous variables, the non-parametric Mann-Whitney test was used as appropriate, taking into account normality assumptions and the number of groups compared.
Receiver operating characteristics (ROC) curves were generated and the areas under the curve (AUC) were calculated to summarise the predictive ability of the RHI score regarding the outcomes under assessment (Mayo endoscopic subscore and FC quantification). The rate of agreement between histological scores and the other outcomes was obtained through the sum of true positives and true negatives in the overall sample. The performance of the histological scores was evaluated for the estimation of sensitivity, specificity, accuracy, predictive values and Kappa values.
The level of statistical significance was set at 0.05. Statistical Package for Social Sciences V.20.0 (IBM SPSS Statistics for Windows) and GraphPad Prism V.7.00 for Mac OS X (GraphPad Software, La Jolla, California USA, www.graphpad.com) were used for the statistical analysis and plots’ design.
Cohort characterisation and disease outcomes
The cohort enrolled in this study was composed of 377 patients with UC, 94% of whom were in clinical remission according to the Montreal classification (table 1). The majority of patients were female (52.3%) and had never smoked (65.5%). Left-sided colitis was the disease location in 57.3% of the patients, whereas 42.7% of them had extensive colitis. The majority of patients had no extraintestinal manifestations (80.6%), no steroid dependence (69.3%) and no steroid resistance (93.2%). Regarding patients’ treatment, 84.9% were on oral 5-aminosalicylic acid, 37.8% were on azathioprine and 25.1% were on antitumour necrosis factor by the time the study started.
Patients’ analysed outcomes are described in table 2. Depending on whether the Mayo endoscopic cut-off was set at 0 or 1, 72.4% and 92.9% of all patients were considered to have mucosal healing, respectively. FC levels had a median (IQR) of 127.0 (45.0–342.0) µg/g and 46.7 (8.4–210.3) µg/g when the samples were assessed with the QB or the EliA kit, respectively. Three different FC cut-offs were considered: 150, 200 and 250 µg/g. When the QB kit was used, 53.7%, 60.9% and 69.3% of the samples were below these cut-offs, respectively; when the EliA kit was used, 71.5%, 74.2% and 78.5% of the samples were below the same cut-offs, respectively. Finally, and in what concerns histopathological analysis, 73.6%, 46.6%, 68.6% and 80.9% were considered to be in remission according to the GS using a cut-off of 3.1, the GS using a cut-off of 2 (deep remission), the NI using a cut-off of 1 and the RHI using a cut-off of 6, respectively. The distribution of samples through all levels of these scoring systems is depicted in the online supplementary table 1.
Supplementary file 1
Accuracy between the histological indices’ cut-off values
To unveil whether the cut-offs defined for each histological index were concordant, patients considered to be in histological remission/activity were compared across the different scoring systems. Table 3 depicts the sensitivity (proportion of patients with active disease according to the reference that are identified as such by the score considered to be the test), specificity (proportion of patients in remission according to the reference that are identified as such by the score considered to be test), positive predictive value (PPV; proportion of patients with active disease according to the test that actually have it according to the reference), negative predictive value (NPV; proportion of patients in remission according to the test that actually are in remission according to the reference), accuracy and kappa between histological remission defined according to the different scoring systems. Overall, there was a moderate to good accuracy (ranging from 65% to 92%). Considering GS as the reference and using 3.1 as the cut-off for defining remission, 87% and 71% of the patients considered to have histological activity are identified as such by the NI and RHI systems (sensitivity), whereas 89% and 100% of the patients considered to be in remission are identified as such by the NI and RHI systems, respectively (specificity). The specificity raises to almost perfect (99%) with both NI and RHI when a stricter GS cut-off is considered (2.0), but the sensitivity values decrease to 58% and 35% with NI and RHI, respectively. Comparing NI and RHI, all patients considered to be in remission according to NI are identified as such by the RHI, whereas only 85% of the patients considered to be in remission according to RHI are identified as such by the NI. On the other hand, all the patients identified as having histological activity with the RHI are also classified as such by the NI, whereas only 61% of the patients identified as having histological activity with the NI are also classified as such by the RHI.
The above-mentioned parameters were further computed for patients stratified according to their disease location (see online supplementary table 2). The results show an overlapping of the 95% CIs between patients with left-sided and extensive colitis.
Additionally, the nature of the RHI scoring system allowed the construction of ROC curves (see online supplementary figure 1). The AUCs in the assessment of the GS using a cut-off of 3.1 and NI using a cut-off of 1 are similar (0.934 (95% CI 0.896 to 0.973) and 0.932 (95% CI 0.903 to 0.962), respectively).
Supplementary file 2
Accuracy between the histological indices and the endoscopic outcomes and FC levels
The sensitivity, specificity, PPV, NVP, accuracy and kappa between histological remission (defined according to the different scoring systems) and endoscopic outcomes and FC levels are depicted in table 4. The results show that the specificity is rather high when the Mayo endoscopic cut-off is set at 0: 80, 93, 87% and 56% of the patients in endoscopic remission are classified as being in histological remission according to NI, RHI, GS (using a threshold of 3.1) and GS (using a threshold of 2.0). On the other hand, the sensitivity is higher when the Mayo endoscopic cut-off is set at 1: 100, 96, 100% and 100% of the patients that have endoscopic activity are classified as having histological inflammation according to NI, RHI, GS (using a threshold of 3.1) and GS (using a threshold of 2.0).
FC levels have a clear association with histological remission: irrespective of the scoring system and FC quantification kit used, FC levels are always significantly higher in patients who present histological activity (figure 1 and online supplementary table 3). The relationship between the different FC cut-offs measured using different methods and histological remission is further depicted in table 4. When FC is quantified using the QB assay, the accuracy (the sum of true positives and true negatives) is higher using a 250 µg/g threshold: 78%, 89%, 84% and 54% of patients who have FC levels below 250 µg/g are classified as being in histological remission according to NI, RHI, GS (using a threshold of 3.1) and GS (using a threshold of 2.0), whereas 52%, 38%, 49% and 72% of patients who have FC levels above 250 µg/g are classified as having histological inflammation according to NI, RHI, GS (using a threshold of 3.1) and GS (using a threshold of 2.0), respectively. On the other hand, when FC is measured with the EliA assay, the accuracy is higher using a threshold of 200 µg/g, although almost equally high with the other two thresholds. Overall, values were not very different from those obtained with QB method, reaching a maximum specificity of 90% (when compared with the RHI and using a cut-off of either 150 or 200 µg/g) and a maximum sensitivity of 75% (when compared with the GS using a threshold of 2 and using an FC cut-off of 250 µg/g).
The performance parameters of both endoscopic outcomes and FC levels in the prediction of the histological classification were further computed for patients stratified according to their disease location (see online supplementary table 4). The results show that most of the time, there is an overlapping between the 95% CIs of the patients with left-sided and extensive colitis.
Additionally, the RHI ROC curves had a high AUC in the assessment of endoscopic lesions (using a Mayo Endoscopic Score threshold ≤1): 0.923 (95% CI 0.862 to 0.984). As for the assessment of FC levels above 150, 200 and 250 µg/g the AUC is lower, ranging from 0.666 to 0.696 (see online supplementary figure 1).
Association between specific histological features and FC levels
As shown above, FC levels are related with histological remission irrespective of the scoring system used (figure 1 and online supplementary table 3). To investigate this issue further, we compared the FC levels between groups of patients presenting specific histological features. The results show that FC levels are not statistically associated with the presence of a chronic inflammatory infiltrate (figure 2 and online supplementary table 5) nor with the presence of neutrophils in the lamina propria (figure 3A,B and online supplementary table 6). However, FC levels are significantly elevated when neutrophils are present in the epithelium according to RHI and GS (figure 3C,D and online supplementary table 7). Concerning NI, the presence of neutrophils in the lamina propria and epithelium is evaluated in the same item (acute inflammatory infiltrate), and the FC levels are significantly higher in the presence of neutrophils when quantified with QB (figure 3E and online supplementary table 8). Finally, irrespective of the FC quantification assay used or the histological scoring system applied, FC levels are significantly higher when tissues present erosion and/or ulceration (figure 4 and online supplementary table 9).
This observation study is, to our best knowledge, the first comparison between one of the most widely used histological scoring systems in UC—the GS—and NI or RHI. The latter are the only histological scoring systems that have been formally validated so far. Moreover, this is also the first assessment of how the histological status evaluated by GS, NI and RHI is related to the endoscopic outcomes and the biomarker FC level. Bressenot et al made a somewhat similar study but without integrating the validated scores: these authors compared GS with Riley, Gupta and Gramlich scores and with GVE, and their results show that the five classification systems are strongly correlated, and that intraobserver reproducibility and interobserver agreement are very good for all indexes.26
Despite the fact that 94% of all patients involved in this study were in remission, 27.6% and 7.1% of them were considered to have endoscopic activity (depending on whether the Mayo Endoscopic Score cut-off was set at 0 or 1, respectively). Moreover, 19.1% to 53.4% of all patients were considered to have histological activity, depending on the score system used: the GS using a cut-off of 2 was the stricter approach (53.4% of all patients had histological activity), followed by the NI (31.4% of all patients had histological activity), the GS using a cut-off of 3.1 (26.4% of all patients had histological activity) and the RHI (19.1% of all patients had histological activity). This discrepancy between clinical assessment, endoscopic evaluation and microscopic inflammation has been noticed before: several studies have previously shown that neither are the clinical symptoms a good indicator of macroscopic lesions nor is the endoscopic assessment a good indicator of histological activity.23 27–29 At this point, it is important to highlight that there is a number of patients with no endoscopic activity and low FC levels that nevertheless have histological activity: 13.8%, 4.6% and 9.4% according to NI, RHI and GS (threshold of 3.1), respectively, have histological activity and yet have Mayo Score=0 and FC <250 µg/g (according to the QB method). As so, biopsies should be taken and evaluated during any endoscopic procedure, even if the other outcomes indicate disease remission.
The concordance between the histological indexes analysed, evaluated in terms of histological remission/activity, was considerably high. In fact, using the GS with a 3.1 cut-off as the reference, 87% and 71% of the patients with histological activity and 89% and 100% of the patients in histological remission were correctly identified by the NI and RHI, respectively. The differences encountered are probably related to the fact that the scoring systems considered are quite different in what comes to their dimension—the number of levels (and therefore the detail of the histological characterisation) is much higher in the RHI score when compared with the other two. Therefore, patients considered to have different classifications according to RHI might fall into the same category using NI or GS, which can cause discrepancies on the remission/activity classification. Furthermore, and in qualitative terms, the main difference between the three considered scores relies on the classification of inflammatory cells infiltrate, whereas GS and RHI differentiate the presence of acute inflammatory infiltrate between lamina propria and epithelium, NI does not include this differentiation and the presence of acute inflammatory infiltrate is considered despite its localisation. On the other hand, RHI does not take into account the presence of eosinophils in lamina propria, a parameter that is valued by the GS. This information is missing from the NI, which does not specify whether the acute inflammatory infiltrate includes neutrophils and eosinophils.
As for the endoscopic outcomes, the sensitivity is higher when one is used as the Mayo Endoscopic Score cut-off (96% to 100% of all patients with macroscopic lesions have histological inflammation), and the specificity is higher when 0 is used as the Mayo Endoscopic Score cut-off (56% to 93% of all patients considered to be in mucosal remission have no histological activity). Concerning FC, the results are clear in what comes to the differences observed between patients with or without histological activity: patients with histological activity have significantly higher FC levels irrespective of the histological grading system and the assay used to quantify FC. However, the establishment of an FC cut-off is not as straightforward, with sensitivity and specificity values varying within the same range for thresholds between 150 and 250 µg/g. The issue of establishing a cut-off for the FC biomarker has been attempted before, and although most values are between 50 and 250 µg/g, one can see some variation within and beyond this interval.21 23 30–34 This shows that FC cut-offs are not universal but rather depend on the patient overall status, the outcome being assessed and the FC quantification methodology used.
Interestingly, one can look at table 4 in an inverted fashion, that is, considering the histological status as the reference and the FC levels (or endoscopic outcomes) as the test: in that case, sensitivity values become the PPV, whereas specificity values become the NPV. NPV is particularly important in this setting, as it represents the proportion of patients that are in histological remission among those that have a negative test result (ie, FC levels below the threshold). For instance, 81%, 93% and 88% of the patients that have an FC level below 150 µg/g (using the QB method) are in histological remission according to NI, RHI and GS (using 3.1 as the cut-off). In other words, this NPV is indicative of the confidence with which one can spare a patient from an endoscopic examination based solely on the non-invasive and cheaper FC test.
Finally, we have explored which histological features are associated with the presence of high FC levels. Our results reveal that FC levels are significantly higher whenever there are neutrophils’ infiltrations in the epithelium or when the tissue presents erosions and/or ulceration. Given that calprotectin is a cytosolic protein particularly abundant in neutrophils, these results are not unexpected. In fact, an augmentation of neutrophils in the epithelium or in the intestinal lumen (due to rupture in the epithelial tissue) should, indeed, increase the amount of FC. However, this is—to our knowledge—the first time that this association is formally demonstrated. The fundamental role of neutrophil infiltration in the mucosa in the pathophysiology of UC has long been acknowledged, and its association with endoscopic severity and systemic inflammatory indexes such as C reactive protein levels has been previously demonstrated.35 Targeting neutrophils is, indeed, a possible therapeutic approach that ought to be further explored.
It is important to notice that a sigmoidoscopy, rather than a colonoscopy, was used to assess the endoscopic status and to collect samples for histological analysis. The sigmoidoscopy was considered to be preferable as it avoids bowel preparation, which could have biased the results. However, this choice could also be seen as a limitation, as the sigmoidoscopy only detects lesions in the lower third of the colon. To address that issue, we have analysed the patients stratified by their disease location (left-sided colitis vs extensive colitis) concerning the performance of the endoscopic outcomes, FC levels and histological classification in the assessment of the histological indexes. For most cases, there was an overlap of the 95% CI of all parameters between patients with left-sided colitis and extensive colitis, suggesting that the presence of lesions above the lower third of the colon do not have a substantial impact in these results.
This study has a few limitations that we hereafter acknowledge: the fact that we based the FC levels quantification on a single sample (consecutive measurements were shown to be more specific than a single one at least in predicting relapse19); the fact that the vast majority of our patients was in clinical remission (and therefore our results can only apply to this specific group of patients) and the lack of a central reader in the endoscopic assessment.
In conclusion, our results show that, at least for patients in clinical remission, the classical histological grading system proposed by Geboes is in accordance with the newly developed and fully validated NI and RHI in what comes to patients’ stratification into histological remission/activity. Moreover, these histological indexes can be associated with the endoscopic outcomes with a high sensitivity when Mayo Endoscopic Score cut-off is set at 1. Finally, FC levels are also associated with these histological indexes and can be related to the presence of neutrophils in the epithelium and tissue erosion/ulceration. Overall, the status of histological remission is dictated by neutrophils’ infiltration: their presence in the epithelium clearly indicates active disease, whereas their presence in lamina propria may occur in patients in remission who still present a certain level of microscopic persistent activity. The total absence of neutrophils from the epithelium and lamina propria (GS <2.0) is indicative of deep remission. The relationship of FC with neutrophils in the epithelium is particularly important, as FC is considered to be a prognostic factor and therefore related with the infiltration of polymorphonuclear neutrophils (PMNs) in the epithelium and the presence of crypt abscesses.36–39 It remains to be clarified whether the presence of neutrophils in lamina propria only has prognostic value.
CCD would like to acknowledge the project NanoSTIMA (NORTE-01-0145-FEDER-000016), which is financed by the North Portugal Regional Operational Programme (NORTE 2020) under the PORTUGAL 2020 Partnership Agreement and through the European Regional Development Fund (ERDF). The authors would like to express their acknowledgement to Sandra Dias for her involvement as the GEDII coordinator and all the help during data collection, as well as to Catarina L Santos for scientific writing assistance.
Contributors FM: Study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; study supervision; critical revision of the manuscript for important intellectual content. JA: faecal calprotectin assays; analysis and interpretation of data. JL: histological analysis. CCD: statistical analysis. KG: supervisor of the histological analysis; critical revision of the manuscript for important intellectual content. FC: responsible for the histological analysis; critical revision of the manuscript for important intellectual content. All the other authors: recruitment of patients and collection of samples. All authors read and approved the final version of the manuscript.
Funding This work was supported by the Portuguese IBD Group (GEDII—Grupo de Estudo da Doença Inflamatória Intestinal).
Competing interests FM served as speaker and received honoraria from Merck Sharp & Dohme, Abbvie, Vifor, Falk, Laboratorios Vitoria, Ferring, Hospira and Biogen. All other authors have nothing to declare.
Patient consent Obtained.
Ethics approval This study was approved by the ethic committee of all hospitals involved and by the Portuguese Data Protection Authority.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.