Article Text

Download PDFPDF

Original article
Relationship between optical coherence tomography derived intraluminal and intramural criteria and haemodynamic relevance as determined by fractional flow reserve in intermediate coronary stenoses of patients with type 2 diabetes
  1. Sebastian Reith1,
  2. Simone Battermann1,
  3. Agnes Jaskolka2,
  4. Walter Lehmacher3,
  5. Rainer Hoffmann1,
  6. Nikolaus Marx1,
  7. Mathias Burgmaier1,2
  1. 1Department of Cardiology, University Hospital of the RWTH Aachen, Aachen, Germany
  2. 2Interdisciplinary Centre for Clinical Research, University Hospital of the RWTH Aachen, Aachen, Germany
  3. 3Institute of Medical Statistics, Informatics and Epidemiology, University of Cologne, Cologne, Germany
  1. Correspondence to Dr Sebastian Reith, Department of Cardiology/Medical Clinic I, University Hospital of the RWTH Aachen, Pauwelsstrasse. 30, Aachen D-52074, Germany; sreith{at}ukaachen.de

Abstract

Background The relationship between functional relevance and optical coherence tomography (OCT)-derived measurements of coronary lesions is incompletely understood and of critical importance, particularly in cardiovascular high-risk patients with type 2 diabetes.

Objective To investigate the association between functional relevance of intermediate grade coronary stenoses as assessed by fractional flow reserve (FFR) and OCT-derived lesion parameters in patients with diabetes.

Methods In 46 diabetic patients with stable coronary artery disease, FFR and OCT were performed in 62 coronary lesions with intermediate severity as determined by quantitative coronary angiography. Among lesions haemodynamic relevance was defined as FFR≤0.8.

Results There was a significant association between FFR and OCT-derived minimal lumen area (r2=0.379) and minimal lumen diameter (r2=0.268), all p<0.001. Receiver operating curve (ROC)-analysis demonstrated an OCT-derived minimal lumen area <1.59 mm2 and minimal lumen diameter <1.31 mm to be optimal cut-off values to predict FFR≤0.8. Furthermore, in lipid-rich plaques FFR was significantly associated with minimal fibrous cap thickness (FCT, r2=0.399). Minimal FCT in lesions with FFR≤0.8 was significantly smaller (60.7±15.0 µm) compared with those lesions with FFR>0.8 (106.0±13.0 µm, p<0.001). ROC-analysis revealed that 0.81 is the ideal FFR cut-off to identify lesions with a minimal FCT≤80 µm (accuracy 97.3%, sensitivity 100%, specificity 93.8%, area under the curve 0.943 (95% CI 0.836 to 1.000)).

Conclusions Haemodynamic relevance of intermediate grade lesions in patients with type 2 diabetes is closely related to (1) intraluminal measurements, which are smaller than previously described in non-diabetic cohorts and to (2) minimal FCT. Furthermore, FFR may be useful to identify vulnerable (minimal FCT≤80 µm) lesions among those with intermediate severity in lipid-rich plaques.

  • IMAGING AND DIAGNOSTICS
View Full Text

Statistics from Altmetric.com

Introduction

Whereas the decision for coronary revascularisation in severe coronary stenoses (>70% diameter in quantitative coronary angiography (QCA)) is mainly based on the interpretation of the coronary angiogram, the assessment of intermediate grade coronary lesions, defined as a diameter stenosis between 40% and 70% in QCA, continues to be challenging for the interventional cardiologist1 and thus requires additional diagnostic work-up.

The Fractional Flow Reserve (FFR) versus Angiography for Multivessel Evaluation (FAME) study demonstrated that FFR-guided percutaneous coronary intervention (PCI) decreases the rate of death, non-fatal myocardial infarction and repeat revascularisation in patients with multivessel coronary artery disease.2

Recently, studies using intravascular ultrasound (IVUS)3–5 and optical coherence tomography (OCT)6 have demonstrated that FFR measurements are significantly related to intraluminal lesion parameters such as minimal lumen area (MLA), minimal lumen diameter (MLD) and per cent area stenosis. However, to date this relationship has not been investigated in patients with type 2 diabetes. Furthermore, nothing is known about the association between FFR-derived haemodynamic severity of coronary stenoses and OCT-derived determinants of vulnerable plaques, such as the presence of thin-capped fibroatheromas (TCFAs) and a minimal fibrous cap thickness (FCT) below 65 µm. Both have commonly been used as predictors of plaque rupture with consecutive acute coronary events.7 ,8 Diabetic patients are particularly at a high risk for cardiovascular events and have been suggested to have an increased prevalence of TCFAs.9 Therefore, the relationship between the lesion composition, including FCT and the prevalence of TCFAs, as well as lesion dimensions on the one hand and the haemodynamic severity of coronary stenoses as determined by FFR on the other hand may be relevant in this population.

OCT, a novel intravascular imaging modality with a 10-fold higher resolution than IVUS, uses the reflection of light and allows visualisation and quantification of intraluminal dimensions as well as in vivo assessment of microstructures of the atheromatous plaque and quantification of the FCT.10

The primary aim of this study was to assess the diagnostic efficiency of OCT-derived intraluminal lesion parameters in identifying haemodynamically relevant coronary stenoses among those of angiographic intermediate severity in a cardiovascular high-risk population with type 2 diabetes. Second, we intended to determine the relationship between OCT-derived plaque characteristics, including minimal FCT, and the FFR-derived functional stenosis severity.

Methods

Study population

Sixty-two de novo coronary lesions were investigated in 46 patients with type 2 diabetes mellitus planned for elective coronary angiography due to stable angina and/or documented ischaemia on stress test at the Department of Internal Medicine I, University Hospital of the RWTH Aachen, Germany between June 2011 and September 2012. FFR measurement, OCT and QCA analyses (figure 1A–C), laboratory analysis and clinical history taking were performed in all patients.

Figure 1

Representative angiographic and optical coherence tomography (OCT) images. (A) Angiographic view of an intermediate grade stenosis of the right coronary artery (fractional flow reserve measurement 0.76). (B) Corresponding OCT-image within the stenosis showing the lipid rich plaque (homogeneous, poorly delineated) with an overlying fibrous cap. (C) High power view from (B) showing the fibrous cap measuring 70 µm (see small arrows). Below longitudinal OCT image, arrow indicating the stenosis.

Inclusion criteria were stable angina pectoris, the presence of at least one intermediate grade coronary stenosis (40–70% on QCA), type 2 diabetes, age >30 years and written informant consent to the study protocol.

Exclusion criteria were left main coronary artery stenosis, bypass graft lesions, congestive heart failure, contraindications to adenosine administration, acute coronary syndromes (ACS), haemodynamic or rhythmic instability, acute or chronic renal insufficiency (serum creatinine level >1.5 mmol/l) and pregnancy. We further excluded serial stenoses in the same epicardial coronary artery and chronic totally occluded, severely tortuous or calcified vessels, not allowing the safe advancement of the OCT catheter.

The study was approved by the local ethics committee and is in accordance with the Declaration of Helsinki on ethical principles for medical research involving human subjects.

Quantitative coronary angiography

Standardised QCA was performed after intracoronary administration of nitrates (0.2 mg). Two orthogonal views of every major coronary vessel were recorded. Analysis of QCA data included reference lumen diameter, MLD, per cent diameter stenosis and stenosis length. These data were determined by an experienced reader, blinded to patient, FFR and OCT data. Offline imaging analysis was performed on a validated QCA software (Philips Inturis Cardio View, QCA V3.3, Pie Medical imaging, Eindhoven, The Netherlands).

FFR measurements

FFR measurements were performed using a 0.014in coronary pressure sensor-tip wire (Certus, St Jude Medical Systems, AB, Uppsala, Sweden). For this purpose maximal hyperaemia was induced by continuous intravenous administration of adenosine at a dosage of 140 µg/kg/min. FFR was calculated as the ratio between intracoronary and aortic pressure during maximally induced hyperaemia. Stenoses were considered to be haemodynamically relevant if FFR≤0.8.2

OCT image acquisition and analysis

For OCT image acquisition we used the Frequency-Domain-OCT C7XR system and the DragonFly catheter (St Jude Medical Systems, Lightlab Imaging Inc, Westford, Massachusetts, USA) with an automated pull-back device at a rate of 20 mm/s. Blood removal was achieved with the non-occlusion OCT-technique by injection of iso-osmolar contrast (Iodixanol, GE Healthcare, USA).

Subsequent offline and pull-back analysis was performed by two independent observers throughout the entire target lesion frame by frame in 0.2 mm intervals, using the standard software provided by Lightlab Imaging and in adaptation to the recently published consensus for quantitative and qualitative assessment.8 ,10 ,11 In case of divergent results, they reached a consensus measurement. Moreover, the observers repeated a blinded analysis after 14 days.

The following quantitative and qualitative assessments were taken:

  1. MLA, MLD and maximum lumen diameter at the frame with the smallest intraluminal area. The eccentricity lumen index was calculated as (maximum lumen diameter—MLD)/maximum lumen diameter.

  2. Reference lumen area at the reference cross section with the largest lumen within 10 mm proximal or distal to the MLA and before any side branch.10 Per cent area stenosis was calculated as (reference lumen area—MLA)/reference lumen area×100.

  3. Stenosis length as the segment around the MLA with the cross-sectional area of at least 50% compared with the predefined reference segment lumen area.

  4. The presence of calcified (signal-poor region with defined borders), fibrous (homogeneous, signal-rich region) and lipid (signal-poor region with diffuse borders) plaques was recorded. A TCFA was defined as a lipid-rich plaque with an overlying FCT≤65 µm extending at least two quadrants of the vessel circumference. A non-TCFA was present if the minimal FCT measured more than 65 µm.10 The minimal FCT was defined as the minimal distance between the arterial lumen and the inner border of the lipid pool, the mean FCT as the average over the entire length of the lipid plaque. The minimal FCT cut-off to define plaque instability was adjusted to 80 µm based on currently published data.12

Statistical analysis

Statistical analyses were performed with SPSS V.20 (SPSS Inc., Chicago, Illinois, USA). Categorical variables were expressed as per cent, continuous variables as mean±SD. The data were analysed on a per-patient basis for clinical characteristics and on a per-stenosis basis for lesion morphology. Continuous variables were tested for normal distribution with the Shapiro-Wilk test, groups were tested for equal variances with the Levene's test and then compared with Wilcoxon-Mann-Whitney test, Student t test or Welch's t test where appropriate. Pearson's χ2 test was used to compare nominal variables. To assess the interobserver and intraobserver variability the results were compared with the κ-test of concordance.

Linear and non-linear regression analyses were performed to determine the association between FFR and OCT as well as QCA parameters. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC) and optimal cut-off values were calculated from the receiver-operating characteristic (ROC) curve to predict FFR≤0.8 and a minimal FCT≤80 µm. Values with the highest Youden-index (sensitivity+specificity−1) were identified as optimal cut-off-values. A classification of the diagnostic efficiency of OCT and QCA parameters according to the values of the area under the curve (AUC) was used as described elsewhere.13 The areas under correlated ROC curves were compared according to the DeLong-method. A p value<0.05 was regarded as statistically significant.

Results

Clinical and lesion characteristics

Forty-six diabetic patients with 62 lesions of intermediate degree and stable coronary artery disease were enrolled in this study. Single-vessel disease was present in 12(26%) patients, multivessel disease in the remaining 34(74%). There were no periprocedural or postprocedural complications associated with the intracoronary use of FFR and OCT. For clinical patient characteristics, QCA and OCT data please refer to tables 1 and 2.

Table 1

Patient characteristics

Table 2

QCA and OCT characteristics

No significant difference was documented in HbA1C values between lesions from patients with (HbA1C 7.52±1.71%) and without haemodynamic relevance (HbA1C 7.20±1.13%).

FFR measurements

Mean FFR of all stenoses was 0.790±0.125. A similar distribution of lesions with FFR≤0.8 (33(53.2%)) and FFR>0.8 (29(46.8%)) was observed. As displayed in table 3, stenoses with an FFR≤0.8 compared with those with FFR>0.8 were significantly longer and more advanced as determined by stenosis parameters including MLA, MLD and per cent area stenosis.

Table 3

OCT-derived quantitative and qualitative differences in lesions with FFR≤0.8 and >0.8

OCT lesion characteristics

In linear and non-linear regression analysis FFR was significantly related to intraluminal and intramural parameters as determined by OCT such as MLA (r2=0.379, p<0.001), per cent area stenosis (r2=0.362, p<0.001), MLD (r2=0.268, p<0.001) and minimal FCT (r2=0.399, p<0.001; figure 2A–D), and to a lesser extent to QCA-derived diameter stenosis (r2=0.172, p<0.001) and MLD (r2=0.112, p=0.008; figure 2E,F).

Figure 2

Association between fractional flow reserve (FFR) and measurements by optical coherence tomography (OCT) and quantitative coronary angiography (QCA). Linear and non-linear regression analysis is displayed for FFR and OCT-derived (A) minimal fibrous cap thickness, (B) minimal lumen area, (C) per cent area stenosis (D) minimal lumen diameter and QCA-derived (E) diameter stenosis and (F) minimal lumen diameter.

The frequency of fibrous (29/33(87.8%) vs 19/29(65.5%), p=0.036) plaques as well as the number of TCFAs (13/33(39.4%) vs 1/29(3.4%), p=0.003) was higher with FFR≤0.8 compared with FFR>0.8 (table 3).

Differences in lesions’ dimensions and FFR-values between lipid plaques and non-lipid plaques did not reach statistical significance (table 4). On the contrary, vulnerable lipid-rich lesions (FCT≤80 µm) were more advanced as determined by OCT-derived lesion parameters, including MLA (1.33±0.39 mm2 vs 2.10±0.94 mm2), MLD (1.07±0.19 vs 1.38±0.32 mm) and per cent area stenosis (75.28±5.62% vs 67.68±7.82%) and had a lower FFR-value (0.683±0.096 vs 0.837±0.104) compared with non-vulnerable plaques, defined as either a lipid-rich lesion with an FCT>80 µm or non-lipid plaques (all p<0.001; table 5).

Table 4

OCT/QCA parameters and FFR value in lipid plaques compared with non-lipid plaques

Table 5

OCT/QCA parameters and FFR value in vulnerable plaques (FCT≤80 µm) compared with non-vulnerable plaques (lipid-plaque with FCT>80 µm or non-lipid plaques)

Predictors for FFR≤0.8

ROC-analysis demonstrated a moderate diagnostic efficiency for OCT-derived parameters including MLA (AUC 0.813, 95% CI 0.728 to 0.930), MLD (AUC 0.816, 95% CI 0.702 to 0.929) and per cent area stenosis (AUC 0.807, 95% CI 0.699 to 0.914) to predict FFR≤0.8 (figure 3A–C). Best cut-off-values were 1.59 mm2 for MLA (sensitivity 75.8%, specificity 79.3%, PPV 80.6%, NPV 74.2%, ACC 77.4%), 1.31 mm for MLD (sensitivity 87.9%, specificity 72.4%, PPV 78.4%, NPV 84%, ACC 80.6%) and 70.6% for per cent area stenosis (sensitivity 75.8%, specificity 72.4%, PPV 75.8%, NPV 72.4%, ACC 72.4%).

Figure 3

Optical coherence tomography (OCT)-derived and quantitative coronary angiography (QCA)-derived parameters predict functional stenosis severity. Receiver-operating characteristic (ROC) analysis for OCT-derived (A) minimal lumen area, and OCT-derived and QCA-derived (B) minimal lumen diameter and (C) per cent area (OCT) or diameter (QCA) stenosis to predict fractional flow reserve≤0.8. The optimal cut-off value is indicated as ‘X’. The differences between OCT and QCA-derived ROC analysis were not statistically significant (p=0.177 for MLD (B) and p=0.141 for per cent stenosis (C)).

In contrast, QCA displayed a lower diagnostic efficiency to predict FFR≤0.8 for MLD (sensitivity 72.7%, specificity 69.0%, PPV 69.4%, NPV 50%, ACC 69.4%, best cut-off-value 1.21 mm) and for per cent diameter stenosis (sensitivity 75.8%, specificity 58.6%, PPV 67.6%, NPV 68.0%, ACC 67.7%, best cut-off-value 47.5%). However, the differences between OCT and QCA were not statistically significant for MLD (p=0.177) and per cent stenosis (p=0.141; figure 3B,C).

FFR predicts minimal FCT≤80 µm

As regression analysis had demonstrated a good association between minimal FCT and FFR (figure 2A), we determined the power of FFR to predict plaque vulnerability.

In 37 lesions with lipid-rich plaque we found a significant difference in minimal FCT between functionally relevant lesions (60.7±15.0 µm) compared with non-relevant ones (106.0±13.0 µm, p<0.001; figure 4A/table 3). FFR had an excellent diagnostic ACC (97.3%) to predict a minimal FCT≤80 µm (sensitivity 100%, specificity 93.8%, PPV 95.5%, NPV 100%, AUC 0.943, 95% CI 0.836 to 1.000) with an optimal FFR cut-off value of 0.81 (figure 4B).

Figure 4

Fractional flow reserve (FFR) predicts plaque instability in intermediate grade coronary lesions. (A) Box plot displays minimal fibrous cap thickness in lesions with FFR>0.80 (n=15) and FFR≤0.80 (n=22), *p<0.001, (B) receiver-operating characteristic (ROC) curve for FFR to predict minimal fibrous cap thickness (FCT)≤80 µm.

Interobserver and intraobserver variability

The intraclass correlation coefficient for interobserver agreements of FCT measurements (κ=0.84) and detection of lipid-rich plaques (κ=0.86) were smaller than the intraobserver variability of FCT measurements (κ=0.89) and detection of lipid-rich plaques (κ=0.92).

Discussion

The main findings of this investigation in type 2 diabetic patients with intermediate grade coronary stenoses are:

  1. OCT-derived intraluminal parameters demonstrate a moderate diagnostic efficiency to predict functional stenosis severity. The best cut-off-values for MLA (1.59 mm2) and MLD (1.31 mm) in this diabetic cohort are smaller than described in previous OCT and IVUS studies in mixed diabetic and non-diabetic populations.

  2. FFR predicts minimal FCT≤80 µm in coronary stenoses and may identify vulnerable lesions among lipid-rich plaques with angiographically intermediate severity.

Whereas PCI is considered to be beneficial in high-grade coronary lesions, the prognostic relevance of an intervention in an intermediate-grade lesion is not proven.14 FFR-guided PCI studies demonstrated that patients with functionally relevant intermediate grade stenoses gain prognostic and symptomatic benefit from coronary interventions.2 ,15 In the FAME study a reduction of the cardiac event rate after 2 years from 20.1% to 16.4% was demonstrated if FFR≤0.8 was used to guide coronary interventions instead of angiography alone.2 ,16 Hence, FFR is presently considered to be the gold standard for characterising haemodynamic relevance among intermediate grade stenoses.2 ,15

In the past, several studies with small patient numbers were conducted to examine the interaction of intraluminal coronary vessel dimensions assessed by IVUS compared with FFR-measurements and described an ideal cut-off-value for MLA between 3–4 mm2 to predict haemodynamic relevance with FFR<0.75.4 ,5 Recently the optimal IVUS-derived cut-off value for MLA was redefined to be 2.4 mm2 to predict FFR<0.8 in a larger patient population.3 According to this investigation and according to consensus the present study also defined the FFR-threshold for haemodynamic relevant coronary stenoses to be 0.8.2 Gonzalo et al6 were the first to evaluate the diagnostic efficiency of OCT-derived intraluminal dimensions in identifying haemodynamically severe coronary stenoses (FFR≤0.8) in 56 patients with 61 lesions and furthermore demonstrated a better diagnostic efficiency with OCT-derived rather than IVUS-derived intraluminal parameters, particularly in small vessels with a diameter<3 mm. They reported an ideal threshold for MLA to be 1.95 mm2 to predict FFR≤0.8 with a moderate diagnostic ACC.

These data were recently confirmed by Shiono et al,17 who demonstrated a similar OCT-derived threshold for MLA (1.91 mm2) in 59 patients and 62 lesions. However, as they used a lower cut-off for functional relevance (FFR<0.75) and given that OCT measurements were performed by the occlusion technique (time-domain), comparison of this study to the one by Gonzalo et al6 and our own, both using the non-occlusion OCT-technique (frequency-domain), is limited, as the occlusive OCT technique has been proven to underestimate true intraluminal dimensions due to the reduction of intracoronary perfusion pressure during balloon inflation.18 The present study found a lower ideal cut-off value for MLA (1.59 mm2) to predict FFR≤0.8. However, in contrast to our study Gonzalo et al6 did not perform a detailed analysis for high-risk patients with type 2 diabetes, as they included only 33.9% diabetic patients. Thus, the discrepancy in MLA between their and our data may reflect the exclusively diabetic population in the present study with 85.5% of target vessels having a reference diameter <3 mm on QCA.

However, due to the moderate diagnostic ACC neither a single of the usually used intraluminal OCT-derived parameters such as MLA, MLD or per cent area stenosis nor a fixed cut-off-value of such a marker appears to be applicable for accurately and individually predicting coronary flow haemodynamics.3 As FFR-interpretation is discussed controversially, particularly in diabetic patients with a wide proportion of functional and structural microvascular abnormalities,19–22 OCT-derived parameters might be a valuable addition to confirm functional stenosis severity.

Previous studies in the general population postulated that major coronary events such as ACS and sudden cardiac death frequently occur in angiographically insignificant stenoses23 ,24 and usually arise from ruptured or vulnerable rupture-prone plaques, characterised by the presence of TCFAs and according to histological investigations to a critically low FCT (≤65 µm).11 ,25–27 This finding could recently be linked to an association between local coronary endothelial dysfunction in non-obstructive coronary lesions and the presence of vulnerable plaque features, namely lipid-rich plaques, in a virtual histology IVUS study.28

Specifically, the PROSPECT study indicated that lesions responsible for subsequent cardiac events had a mean angiographic diameter stenosis of only 32%.24 In contrast, the current study demonstrated that in patients with type 2 diabetes the more advanced lesions (as determined by OCT-derived lesion parameters including MLA, MLD and per cent area stenosis) with haemodynamic relevance were those with a lower minimal FCT and vulnerable plaque features. Thus it is tempting to speculate, that compared with the general population, in diabetic patients with diffuse small vessel disease the features of plaque instability coincide with an increased stenosis severity. This in turn may result in a different risk assessment for vulnerable lesion recognition.

Particularly as diabetic patients are prone to plaque vulnerability9 ,29 as well as post-PCI complications,30 it is important to select those lesions for coronary intervention that are functionally and prognostically relevant.

Our data demonstrated that FFR predicts minimal FCT≤80 µm, a value that has recently been postulated to be the ideal critical threshold of FCT for plaque instability in vivo,12 with an optimal cut-off for FFR of 0.81 and a high diagnostic efficiency (AUC 0.943).

Thus, our finding may add to the current knowledge and understanding by indicating that, besides intraluminal parameters, haemodynamic relevance is furthermore associated with plaque vulnerability. Hence, FFR may be a useful adjunct to distinguish between stable (minimal FCT>80 µm) and unstable (minimal FCT≤80 µm) lesions among lipid-rich plaques with intermediate severity in high-risk patients with type 2 diabetes.

Limitations

Although the present study is, to the best of our knowledge, currently the only investigation assessing the relationship between haemodynamic relevance and OCT-derived intraluminal parameters in an exclusively diabetic cohort, the sample size remains small. In the future it will be equally necessary to confirm these data in a non-diabetic population, to evaluate possible differences between diabetic and non-diabetic patients and furthermore to perform a large scale interventional study.

Second, though we enrolled only diabetic patients with a known higher prevalence of TCFAs,12 the number of lipid-rich plaques is still limited. Thus, we could only determine a FCT in 37/62 lesions.

Moreover, we used a minimal FCT-threshold of 80 µm for the definition of plaque vulnerability. This is above the currently accepted consensus of 65 µm, a value that was based on histomorphological studies, where shrinkage of tissue may be a major issue.10 Thus, we referred to a recently published OCT-investigation which demonstrated that 95% of ruptured lipid-rich plaques in vivo had a minimal FCT<80 µm.12

Conclusion

We conclude that OCT-derived intraluminal parameters have a moderate diagnostic efficiency to predict haemodynamic relevance of intermediate grade lesions in diabetic patients, whereas the presented cut-off values are smaller than previously described in non-diabetic cohorts. Moreover, FFR may be useful to discriminate unstable coronary artery lesions among those with angiographic intermediate severity that contain lipid plaque.

Acknowledgments

This work was supported by grants from the Interdisciplinary Centre for Clinical Research within the medical faculty at the RWTH Aachen.

References

View Abstract

Footnotes

  • Contributors All authors were involved in reporting the results of this study and all approved the final version of the submitted manuscript. SR, NM and MB contributed in the conception, design and planning of the study. SR, SB, AJ, RH, NM and MB conducted the study and were involved in the analysis and interpretation of the data. WL, SR and MB did the statistical analysis. Manuscript writing: SR and MB. SR is responsible for the overall content and serves as guarantor.

  • Funding Grant from the Interdisciplinary Centre for Clinical Research within the faculty of Medicine of The University Hospital of the RWTH Aachen.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Approval by the local ethics committee of the University of the RWTH Aachen.

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

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.