Article Text

Download PDFPDF
Pre-operative surgical algorithm: sentinel lymph node biopsy as predictor of parametrial involvement in early-stage cervical cancer
  1. Vincent Balaya1,2,
  2. Arnaud Bresset3,
  3. Benedetta Guani1,2,
  4. Louise Benoit4,
  5. Laurent Magaud5,
  6. Hélène Bonsang-Kitzis6,
  7. Charlotte Ngo6,
  8. Patrice Mathevet1,2 and
  9. Fabrice Lécuru7 SENTICOL group
  1. 1 Gynecology Department, University Hospital of Lausanne Department of Gynaecology and Paediatrics, Lausanne, Vaud, Switzerland
  2. 2 University of Lausanne, Lausanne, Switzerland
  3. 3 Obstetrics and Gynecology department, Beaujon Hospital, Clichy, France
  4. 4 Surgical Oncology Department, Curie Institute, St-Cloud, France
  5. 5 Public Health Department, Civil Hospices of Lyon, Lyon, Auvergne-Rhône-Alpes, France
  6. 6 Gynecologic and Breast Surgery Center, Groupe Ramsay hôpital des Peupliers, Paris, France
  7. 7 Breast, Gynecology and Reconstructive Surgery Unit, Institute Curie, Paris, France
  1. Correspondence to Dr Vincent Balaya, Gynecology Department, University Hospital of Lausanne Department of Gynaecology and Paediatrics, Lausanne, Switzerland; vbalaya{at}hotmail.com

Abstract

Introduction Radical hysterectomy is the gold standard in the management of early-stage cervical cancer. Parametrectomy aims to remove occult disease but is associated with significant surgical morbidity. Avoiding unnecessary parametrectomy in a subset of patients at low risk of parametrial involvement may decrease the incidence of such morbidity. The purpose of this study was to identify patients at low risk of parametrial involvement in early-stage cervical cancer potentially eligible for less radical surgery based on pre-operative criteria and sentinel lymph node (SLN) status.

Methods We performed an ancillary analysis of data from two prospective trials on sentinel node biopsy for cervical cancer (SENTICOL I and II). Patients with International Federation of Gynecology and Obstetrics (FIGO) IA–IIA cervical cancer who underwent primary radical surgery and bilateral SLN mapping were identified between 2005 and 2012 from 25 French oncologic centers. Patients who underwent pre-operative brachytherapy or did not undergo radical surgery (simple trachelectomy, simple hysterectomy, or lymph node staging only) were excluded.

Results Of 174 patients who fullfiled the inclusion criteria, 9 patients (5.2%) had parametrial involvement and 24 patients (13.8%) had positive SLN. Most patients had 2018 FIGO stage IB1 disease (86.1%) and squamous cell carcinomas (68.9%). Parametrial involvement was significantly associated with tumor size ≥20 mm on pelvic magnetic resonance imaging (MRI) (adjusted odds ratio (ORa) 9.30, 95% CI 1.71 to 50.57, p=0.01) and micrometastic or macrometastatic SLN (ORa 8.98, 95% CI 1.59 to 50.84, p=0.01). Of 114 patients with tumors <20 mm on pre-operative MRI and negative SLN after ultrastaging, only one patient had parametrial involvement (0.9%). By triaging patients with both of these criteria in a two-step surgical procedure, unjustified and contra-indicated radical hysterectomy would have been avoided in 65.5% and 8.6% of cases, respectively.

Conclusions Less radical surgery may be an option for patients with bilateral negative SLN after ultrastaging and tumors <20 mm. SLN status should be integrated into the decision-making process for tailored surgery in early-stage cervical cancer.

  • cervical cancer
  • gynecologic surgical procedures
  • sentinel lymph node
  • surgical oncology
View Full Text

Statistics from Altmetric.com

HIGHLIGHTS

  • Patients with cervical cancer and a low risk of parametrial involvement should be eligible for less radical surgery

  • Patients with negative sentinel lymph nodes and tumor size <20 mm have a low risk of parametrial involvement

  • Sentinel lymph node status should be integrated into the decision-making process for tailored surgery in early-stage cervical cancer

Introduction

Radical hysterectomy is the gold standard for early-stage cervical cancer.1 Considering that the parametrium is a key point of lymphatic drainage of the cervix and cervical cancer has a lymphatic spread pattern,2 3 the rationale for parametrectomy is to remove occult disease at the same time as resection of the primary cervical lesion. However, patients may experience lower urinary tract dysfunction, sexual dysfunction, and colorectal motility disorders.4 This post-operative morbidity is mainly due to the removal of the parametrium which contains autonomic nerve fibers.5 Although nerve-sparing techniques have been developed to decrease nerve damage,6 avoiding unnecessary parametrectomy in a subset of patients at low risk of parametrial spread may decrease the incidence of such morbidities and improve the quality of life without compromising oncologic outcomes.

Several series have described risk factors for parametrial involvement such as larger tumor size, presence of lymphovascular space invasion, deeper stromal invasion, and positive lymph nodes.7 8 However, most of these data are derived from retrospective analyses of post-operative surgical specimens. To assess the impact of less radical surgery in cervical cancer, two prospective clinical trials are currently ongoing, SHAPE (NCT01658930) and GOG278 (NCT01649089), and preliminary results of a third one have recently been presented (Concerv9). In the era of step-down in the radicality of surgical management for cervical cancer, few data are available on the predictive value of sentinel lymph node (SLN) status for parametrial involvement.

The purpose of this study was to identify patients at low risk of parametrial involvement who might be eligible for less radical surgery in early-stage cervical cancer based on pre-operative criteria and SLN status.

Methods

Population study

We performed a retrospective analysis of data from two prospective trials on sentinel node biopsy for cervical cancer (SENTICOL I and II).10 11 Patients with early cervical cancer (stage IA1 with lymphovascular emboli to stage IB1) were included consecutively from seven French gynecological oncology centers between 2005 and 2007 for SENTICOL I and 23 French gynecological oncology centers between 2009 and 2012 for SENTICOL II.

In this study, patients having a radical hysterectomy or radical trachelectomy with bilateral SLN mapping were identified. Patients who did not undergo radical surgery (simple trachelectomy, simple hysterectomy, or lymph node staging only) were excluded because parametrial status was not assessable. Patients who underwent pre-operative brachytherapy were also excluded. This study was approved by the Lyon’s Institutional Review Board. Patients included in the two studies gave written consent stating the use of data for secondary analyses.

Data analysis

From these two prospective multicentric databases, demographic characteristics, surgical history, clinical data including International Federation of Gynecology and Obstetrics (FIGO) stage and pre-operative pelvic magnetic resonance imaging (MRI) were extracted. At pre-operative imaging, no patients had suspicion of metastatic nodes or parametrial involvement. Operative records were reviewed and data about conization and the type of radical surgery performed (hysterectomy or trachelectomy) were collected. If a pre-operative conization was performed, tumor size, histology type, the presence or absence of lymphovascular space invasion and margin status were described in the conization specimens. All pathology slides were analyzed at the center where they were performed by experienced gynecologic pathologists. Parametrial involvement was defined as any evidence of disease in the parametrial tissue: direct microscopic spread, positive parametrial nodes, and lymphovascular space invasion, defined as tumor cells present within lymphovascular channels in the parametrium.

Detection of SLN was performed using a combined labeling technique (radioactive tracer (99mTc) and patent blue). Frozen section analysis was performed either routinely or only on suspected metastasis nodes at the surgeon's discretion. SLNs were analyzed after hematoxylin and eosin staining of 200 µm sections. Negative SLNs were then examined by immunohistochemistry with anti-cytokeratin AE1–AE3 antibodies. Isolated tumor cells were defined as <0.2 mm, micrometastases between 0.2 and 2 mm, and macrometastases >2 mm. In accordance with the journal guidelines, we will provide our data for the reproducibility of this study in other centers if such is requested.

Statistical analysis

Patients were divided into two groups according to their parametrial status on the surgical specimens at final pathologic examination: positive or negative. Qualitative variables were described as n (%) and quantitative data as mean (range). The chi-square test was used to compare qualitative variables and the Student’s t-test was used to compare quantitative variables. P values <0.05 were considered significant. Receiver operating characteristic (ROC) analysis of the significant quantitative factors was made to define threshold values. To determine the agreement of tumor size between pelvic MRI and conization measurement to pathologic findings, a correlation Spearman test was performed. Significant variables in univariate analysis were entered into a multivariate logistic regression model to determine pre-operative variables independently associated with parametrial involvement. Based on these independent variables, an asymptotic exact logistic regression test was applied to determine the prediction risk of parametrial involvement. All statistical analyses were run using XLStat Biomed software (AddInsoft V19.4).

Results

Clinicopathological characteristics

In SENTICOL I and II, 326 patients had radical surgery (radical trachelectomy or hysterectomy). Of these, 152 patients were excluded: 78 patients had pre-operative brachytherapy, 6 patients had no SLNs detected, 29 patients had SLNs detected only in one side, 37 patients did not have any pre-operative pelvic MRI or conization, and 2 patients had missing data on parametrial status at final pathologic examination. Finally, a total of 174 patients fulfilled the inclusion criteria and were analyzed. Overall, nine patients (5.2%) had parametrial involvement on the surgical specimens: six had direct microscopic invasion, two had lymphovascular space invasion in the parametrium, and one had a parametrial macrometastic node. Unilateral spread was seen in six patients and three patients had bilateral spread. Clinicopathological factors of the patients are detailed in Table 1.

Table 1

Univariate analysis of predictive factors associated with parametrial involvement

The mean age of the entire cohort was 43.3 years (range 22–81) and the mean BMI was 23.5 kg/m2 (range 14.6–41.4). Most patients had stage IB1 and IB2 clinical 2018 FIGO stage (54.0% and 32.8%, respectively). Squamous cell carcinoma was most frequent (63.6%). Radical hysterectomy was performed in 132 patients (75.9%) whereas 42 patients had a radical trachelectomy (24.1%). No patients had pelvic MRI findings indicating parametrial involvement or suspicious metastatic SLNs. The tumor diameter measured by MRI ranged from zero (no demonstrable lesion on the cervix) to 41 mm, with a mean diameter of 10.4 mm. In total, 47 patients (27%) had a tumor size that measured more than 20 mm. Prior to the radical surgery, 105 patients (60.3%) had a conization. Patients with clinical FIGO stage IA were more likely to undergo pre-operative conization than patients with clinical FIGO stage IB1 and IB2 (80.0% vs 59.6% and 57.9%, respectively; p=0.047).

All patients underwent successful bilateral SLN mapping and 119 patients (68.4%) underwent additional pelvic lymphadenectomy. In 143 patients (82.2%), frozen section analysis identified three patients with macrometastasis and two patients with micrometastasis but missed one patient with macrometastasis, four patients with micrometastasis, and three patients with isolated tumor cells. The sensitivity and the negative predictive value were 26.3% and 89.8%, respectively, for all types of metastasis and 45.4% and 95.6% if isolated tumor cells were excluded. Overall, 24 patients (13.8%) had positive SLN after ultrastaging: six patients (3.4%) with macrometastasis, nine (5.2%) with micrometastasis, and nine (5.2%) with isolated tumor cells.

Predictive factors

By univariate analysis, BMI tended to be higher, but not significantly, in patients with parametrial involvement (26.5±6.0 kg/m2 vs 23.3±4.9 kg/m2, p=0.07) (Table 1). On pre-operative MRI, mean tumor size was significantly higher in patients with lymphovascular space invasion (18.9±12.9 mm vs 10.0±10.5 mm, p=0.02). ROC analysis showed that a threshold of 19 mm in tumor size assessed by MRI predicted parametrial involvement with a 62.5% sensitivity, 75.8% specificity, and an area under the curve of 0.714. Among the 105 patients who had a prior conization, tumor size on the conization specimens was significantly higher in the group of patients with parametrial involvement (20.8±4.3 mm vs 13.3±7.3 mm, p=0.02). After ultrastaging, patients with positive SLN more frequently had parametrial involvement (20.8% vs 2.7%, p=0.0002), especially those with macrometastasis (33.3% vs 3.4%, p=0.001). At final pathologic examination, patients with tumor size >20 mm and patients with deep stromal invasion (deeper than 10 mm) had more frequent parametrial involvement (in 17.1% and 15.4% of cases, respectively). Patients with vaginal invasion and those with positive surgical margins had associated parametrial involvement in 54.5% and 44.4% of cases, respectively. Although the presence of lymphovascular space invasion in surgical specimens was significantly associated with parametrial involvement (11.5% vs 1.8% p=0.006), this association did not reach statistical significance in the biopsy or conization specimens (8.1% vs 4.4%, p=0.39).

To determine predictive factors of parametrial involvement, a multivariate analysis including only pre-operative factors was performed. To predict tumor size at definitive pathologic examination, tumor size assessed by pre-operative imaging had a better sensitivity and specificity than that assessed on conization specimens (68.3% vs 57.1% and 86.2% vs 74.7%, respectively). Moreover, correlation with post-operative measurement in tumor size was higher for pelvic MRI (r[S]=0.537) than that for conization specimen (r[S]=0.144). Multivariate analysis retained tumor size >20 mm at pre-operative imaging (adjusted odds ratio (ORa) 9.30, 95% CI 1.71 to 50.57; p=0.01) and SLNs positive for micrometastases or macrometastases (ORa 8.98, 95% CI 1.59 to 50.84; p=0.01) as independent pre-operative risk factors of parametrial involvement (Table 2).

Table 2

Multivariate analysis of pre-operative factors associated with parametrial involvement

Risk stratification of parametrial involvement

Due to statistical significance, tumor size at preoperative imaging and SLN status were included to perform an asymptotic exact logistic regression test to determine the prediction risk of parametrial involvement (Table 3). Based on these results, we suggested the following risk stratification of parametrial involvement: patients with no risk factors may define a low-risk group (114 patients, 65.5%), patients with one risk factor (positive SLNs or tumor size ≥20 mm) an intermediate-risk group (49 patients, 28.2%), and patients with two risk factors (positive SLNs and tumor size ≥20 mm) a high-risk group (11 patients, 6.3%). The parametrial involvement rates for low-risk, intermediate-risk, and high-risk groups were 0.9%, 8.2%, and 36.4%, respectively (p<0.0001). Figure 1 shows the ROC curve using tumor size on MRI and SLN status for discriminating between low-risk, intermediate-risk, and high-risk patients. The area under the curve was 0.838 (95% CI 0.704 to 0.973, p<0.0001).

Figure 1

ROC curve using tumor size on MRI and sentinel lymph node status for risk stratification of parametrial involvement. AUC, area under the curve.

Table 3

Probability of parametrial involvement according to pre-operative factors

Surgical algorithm

Tumor size and SLN status were applied to our cohort as a two-step flow chart algorithm (Figure 2). Among the 114 patients classified as low-risk, only one patient had parametrial involvement (0.9%). In this low-risk subgroup, radical hysterectomy might have been unjustified and these patients would be eligible for simple hysterectomy or trachelectomy. If the same algorithm was applied in a one-step strategy by considering results of frozen section analysis, 100 patients would be classified as low-risk and, among them, two would have parametrial involvement (2%).

Figure 2

Surgical algorithm. ITC, isolated tumor cells; MIC, micrometastases; MAC, macrometastases; CCR, concomitant chemoradiotherapy; PI, parametrial involvement.

Among the 15 patients who had micrometastatic or macrometastatic SLNs, three patients also had parametrial involvement. In these cases with positive nodes, radical hysterectomy would have been contra-indicated and patients would have been referred to radiochemotherapy instead of undergoing surgery. By contrast, surgery would remain indicated if isolated tumor cells are present given that they do not change the FIGO stage.12 Even if no patients with isolated tumor cells on SLNs and tumor <20 mm had parametrial involvement, radical hysterectomy should be performed due to a significant risk of parametrial involvement of 4.6% in this subgroup of patients. Pre-operative brachytherapy could be an option in cases with tumor size >20 mm.

By triaging these 174 patients with tumor size at pre-operative imaging and SLN status after ultrastaging, unjustified and contra-indicated radical hysterectomy would have been avoided in 114 patients (65.5%) and in 15 patients (8.6%), respectively.

Overall, 32 patients received adjuvant radiation therapy with or without vaginal brachytherapy: 28 patients (including nine patients with parametrial involvement) were considered as high-risk and four as intermediate-risk according to Sedlis criteria. By taking into account our parametrial involvement risk stratification, all of the 11 high-risk patients (100%), 18 of 49 intermediate-risk patients (36.7%), and only 3 of 114 low-risk patients (2.6%) required adjuvant radiation therapy.

Discussion

In this cohort of 174 patients with prospective data we found that tumor size >20 mm on pre-operative imaging and positive SLNs were significantly associated with parametrial involvement in early-stage cervical cancer. Patients at low risk of parametrial involvement may be safely selected with these criteria and only one patient in the low-risk group had parametrial involvement (0.9%). Although the majority of data have been derived from retrospective studies or post-operative surgical specimen analysis, several surgical triage algorithms have been described including different factors such as age,13 14 lymphovascular space invasion,15 16 pelvic lymph nodes,13 16 17 deep stromal invasion,18 and conization specimen.19 20 Selection criteria for less radical surgery described in the literature are detailed in the online supplementary table.

Supplemental material

Tumor size >20 mm is the most described risk factor for parametrial involvement.8 Pre-operatively, tumor size could be assessed at the physical examination, by pelvic MRI, or on the conization specimen.1 12 According to Lee et al, pelvic examination was more accurate than pelvic MRI in cases of small tumor volume whereas radiologic measurement was superior for bulky tumors of cervical cancer.21 For Zhang et al, the accuracy of MRI in measuring tumor size was significantly higher than that of pelvic examination, both in exophytic and endocervical tumor.22 Although several previous studies have assessed the tumor size at MRI as a risk factor of parametrial involvement, no consensual cut-off has been determined. In our study the ROC curve analysis determined a threshold of 19 mm in tumor size assessed by MRI to predict parametrial involvement. This cut-off corresponds to the limit between stage IB1 and IB2 of the revised 2018 FIGO classification12 and is similar to that suggested by Kamimori et al.23 However, some authors have reported a lower cut-off of 10 mm24 or a higher cut-off of 25 mm25 or 30 mm.26 27 To improve the accuracy of MRI, other imaging criteria have been added such as tumor volume and disruption of the cervical stromal ring,27 but these might be less reproducible in current practice.

Tumor size could be otherwise determined in conization specimens. We found that patients with parametrial involvement had a larger tumor size on the conization specimen than patients without parametrial involvement, but this variable was less reliable for tumor size at the final pathological examination than in pelvic MRI. We speculate that patients who underwent pre-operative conization had a decreased tumor size found at the final pathological examination of hysterectomy specimens. Smith et al aimed to determine risk factors for parametrial involvement based on conization specimens and found that tumor size on conization was not associated with parametrial involvement.20 The authors explained this result by the limited sample of patients having tumor size >20 mm and pre-operative conization. We agree that, in our cohort, patients with clinical FIGO stage IA were more likely to undergo pre-operative conization than patients with clinical FIGO stages IB1 and IB2. Moreover, patients with a small tumor are likely to have a diagnostic conization whereas, for patients with a large tumor, a simple biopsy is enough to diagnose cervical cancer.

The presence of lymphovascular space invasion is a well-known risk factor of parametrial involvement.13 15–17 19 26 28–31 Even if the presence of lymphovascular space invasion on the surgical specimen was significantly associated with parametrial involvement, we did not see this association in the biopsy or conization specimens. Detection of lymphovascular space invasion in biopsy or conization specimens lacks reliability and had a low negative predictive value for detection of lymphovascular space invasion at the final pathologic examination.13 20 32 However, the single low-risk patient who had parametrial involvement had lymphovascular space invasion on the final pathologic specimen and we believe that the presence of lymphovascular space invasion still has to be considered a risk factor for parametrial involvement.

Several investigators have reported positive lymph nodes as a risk factor for parametrial involvement.13 15–17 19 28 31 However, knowledge of node status requires pelvic lymphadenectomy and is not available pre-operatively. Moreover, patients with positive nodes diagnosed on pre-operative imaging should be referred for concomitant radiochemotherapy. A few studies have investigated the contribution of SLN biopsy to selecting patients eligible for less radical surgery. Coutant et al did not find any significant association between the SLN status and parametrial involvement but, among patients with tumor size <20 mm and negative SLNs, none had parametrial involvement.33 In a prospective study, Strnad et al reported no parametrial involvement among the 133 patients with negative SLNs whereas 7 of 25 patients (28%) with positive SLNs had parametrial involvement.34 Klat et al did not find any parametrial involvement among 60 patients with tumors <20 mm in diameter and negative SLNs.35 These results are consistent with ours and highlight the impact of SLN status as a predictive factor of parametrial involvement.

We suggested a two-step surgical algorithm taking into account the pre-operative tumor size and SLN status after ultrastaging. In this algorithm, positive SLNs were subdivided into two groups: isolated tumor cells on the one hand and micrometastases/macrometastases on the other. According to the last revised 2018 FIGO classification, the presence of macrometastases or micrometastases defined FIGO stage IIIC whereas isolated tumor cells might be simply recorded without changing the stage.12 36 Two large cohorts revealed that the presence of isolated tumor cells had no significant impact on disease-free survival37 38 and additional adjuvant therapy would not be required for these patients.39 In our cohort, only three patients had isolated tumor cells and tumor size <20 mm and, of these, none had parametrial involvement. However, the risk of parametrial involvement was not negligible (4.6%) in this subgroup of patients and, therefore, we support the idea that radical surgery should still be considered. In patients with tumor size >20 mm, the risk of parametrial involvement was 8.3% if SLNs were negative and 33.3% if isolated tumor cells were present. For these patients, we proposed pre-operative brachytherapy followed by radical surgery as an option since it may decrease initial tumor size and lymphovascular space invasion.40–42 According to the European Society of Gynaecological Oncology guidelines, this strategy is an acceptable alternative option for intermediate-risk and high-risk patients defined by Sedlis criteria.1 However, caution should be paid to an increased risk of surgical morbidity.4

In a pilot study, Pluta et al did not report any recurrence among 55 patients with tumor size <20 mm and negative SLNs.43 Interestingly, the authors stressed the diagnostic value of frozen section analysis of SLNs and reported a low false-negative rate of 3.6%. The accuracy of frozen section analysis of SLNs is of paramount importance to identify patients at low risk for parametrial involvement in a one-step strategy. By taking into consideration frozen section analysis, 100 of 143 patients (69.9%) would have had only a simple hysterectomy and five patients (3.5%) would have avoided combined treatment. Nonetheless, some limitations have to be noted with this one-step strategy. First, the sensitivity of frozen section analysis was poor for low-volume nodal metastases and three patients with parametrial involvement would have been missed without ultrastaging (one with micrometastasis and two with isolated tumor cells). This low sensivity for small nodal metastases has been reported by other authors.44 45 However, frozen section analysis has a high negative predictive value of 89.8%–95.6% and, among the 100 patients with negative SLNs in frozen section analysis and tumor size <20 mm, only two had parametrial involvement. The second limitation is whether to abort radical surgery in cases of positive SLNs remains controversial,46 47 but the ABRAX study will provide data on this point.48

Ultrastaging of SLNs provides a more precise risk assessment of parametrial involvement and a two-step approach (SLN biopsy and ultrastaging first, radical or simple surgery second) is required for this algorithm to be fully reliable. By applying this strategy in our cohort of 174 patients, 114 patients (65.5%) would have had only a simple hysterectomy and 15 patients (8.6%) would have avoided radical hysterectomy with additional radiochemotherapy. The cost induced by performing two surgeries might be counterbalanced with potential benefits of shorter length of hospital stay and decreased morbidity. Among the limitations of our study are the low rate of parametrial involvement in a preselected population linked to inclusion criteria of both SENTICOL I and II studies (small tumor size and previous MRI without suspicious nodes or parametrial involvement) and the retrospective analysis of two databases which were not designed for our objectives. This work deserved nonetheless to assess exclusively pre-operative risk factors of parametrial involvement with prospective data.

Conclusion

Less radical surgery may be an option for patients with bilateral negative SLNs and tumors <20 mm. SLN status should be integrated into the decision-making process for tailored surgery in early-stage cervical cancer.

References

View Abstract

Footnotes

  • Twitter @BenedettaGuani

  • Collaborators SENTICOL group.

  • Contributors Study concepts: VB, BG, PM, FL. Study design: VB, AB, PM, FL. Data acquisition: LM, PM, FL, VB, AB. Quality control of data and algorithms: LM, FL, PM, BG, VB. Data analysis and interpretation: VB, FL, LB, HB-K, CN, AB. Statistical analysis: LB, HB-K, VB, BG. Manuscript preparation: VB, AB, FL, PM. Manuscript editing: VB, FL, CN, HB-K, BG. Manuscript review: FL, PM, BG.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information. All data relevant to the study are included in the article or uploaded as supplementary information.

Request Permissions

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.