OUP user menu

D-dimer in ruling out acute aortic dissection: a systematic review and prospective cohort study

Gottfried Sodeck, Hans Domanovits, Martin Schillinger, Marek P. Ehrlich, Georg Endler, Harald Herkner, Anton Laggner
DOI: http://dx.doi.org/10.1093/eurheartj/ehm484 3067-3075 First published online: 6 November 2007

Abstract

Aims Blood D-dimer testing has been proposed as diagnostic marker with high sensitivity for exclusion of acute aortic dissection (AAD). We performed a systematic review and validated the findings in a prospective patient cohort.

Methods and results We searched MEDLINE, EMBASE, CINAHL, and BIOSIS from inception until January 2007 using a combination of search terms for aortic dissection and D-dimer. Study type, type of assay used, predefined cut-off level, result of D-dimer testing, sensitivity, and specificity were abstracted. In 16 identified studies (437 patients), the reported cut-off values ranged from 0.1 to 0.9 µg/mL. D-dimer testing provided high sensitivity (0.97 95% CI 0.94–0.98) and negative likelihood ratio (0.06 95% CI 0.02–0.13). In our cohort of 65 patients (36 male, 55%; median age 59 years, IQR 49–67) with proven AAD, D-dimer levels scattered from 0.24 to 137.88 µg/mL (median 3.47; IQR 1.55–14.49). Mean NPV for the different cut-off levels ranged from 92 % for a cut-off level of 0.9 µg/mL to 100% for a cut-off level of 0.1 µg/mL in our study population.

Conclusion Current evidence supports a routine measurement of D-dimer in excluding AAD. A D-dimer <0.1 µg/mL will exclude AAD in all cases.

Keywords
  • Aortic dissection
  • Diagnosis
  • D-dimer

Introduction

Acute aortic dissection (AAD) represents a catastrophic vascular emergency that requires early diagnosis and treatment.1 Currently, diagnosis or exclusion of AAD by clinical parameters remains challenging, since no biomarker for AAD diagnosis is available in clinical routine. Aortic dissection may present with a variety of clinical symptoms that may delay or even misslead correct diagnosis.2 A biomarker- aided approach to diagnose or exclude aortic dissection therefore would be desirable. In this context, recent studies proposed the measurement of normal D-dimer levels which may exclude AAD with 100% sensitivity.35 A validated and uniform cut-off level of normal in AAD is, however, still missing.6

In the present study, we performed a systematic review and meta-analysis of published literature to determine sensitivity and optimal cut-off level to exclude the presence of AAD by determination of D-dimer levels. We then validated the findings of the systematic review in a study population of patients with proven AAD.

Methods

Systematic review

We identified clinical trials and case reports assessing the clinical value of D-dimer levels in diagnosis of AAD. We searched MEDLINE, EMBASE, CINAHL, and BIOSIS from inception until January 2007 using an Ovid-based web interface (Ovid technologies, NY, USA). The search string was [(aorta) or (aortic) or (dissection) or (aneurysm) and/or (fibrin fragment D) or (D-dimer)]. Editorial comments, review articles, and abstracts as publications in other languages than English, German, French, Dutch, Danish, Spanish, Italian, or Portuguese were excluded from further review. We predefined a variety of endpoints for assessment of clinical validity: study design, number of patients, classification of AAD, types of assay used, predefined thresholds, levels of D-dimer, sensitivity, and specifity of D-dimer measurement in diagnosis of AAD.

Two reviewers (G.H.S. and H.D.) independently abstracted data from each trial and entered the data in a predefined database. We compared the results and resolved disagreement by discussion among reviewers.

Validation of test sensitivity in a prospective cohort of acute aortic dissection patients

All patients with acute Stanford A aortic dissection presenting to a tertiary care non- trauma emergency department (Department of Emergency Medicine, Medical University Vienna, Austria) were eligible for study participation. The inclusion period was set between January 2003 to January 2007.

Demographic data and time from onset of clinical symptoms to presentation to the emergency department were recorded in a comprehensive database.

D-dimer was measured immediately after establishment of correct diagnosis. For determination of quantitative D-dimer values, STA® latex-agglutination test (Roche Diagnostics Vienna, Austria) was used according to the manufacturer's recommendations. The test has a lower detection limit of 0.04 µg/mL with a coefficient of variation of 10%.

All study-related procedures were in accordance with the ethical standards of the responsible committee of the Vienna Medical School.

Statistics and study endpoint

Systematic review

Continuous data are presented as the median and the interquartile range and the mean and the corresponding standard deviation, as appropriate. In articles reporting the median and the interquartile range, we took the median to be representative of the mean and converted the interquartile range into a standard deviation by dividing it by 1.35.7

We calculated a summary estimate for the true positive rate (sensitivity) using the ‘meta’ command in STATA. To handle statistical heterogeneity and to yield comparable results, we used random effect models to produce summary true positive rates. Metaregression using the ‘metareg’ command in STATA was performed to extend the random effects meta-analysis. We estimated the extent to which the cut-off value explains heterogeneity in the true positive rate. To estimate the additive between studies variance tau2 a moment estimator, using a non-iterative procedure was employed to comply with estimates of the random-effects meta-analysis. We used MS Excel 2003, STATA 8.2 (Intercooled Stata 8.2 for Windows, StataCorp LP, Houston, TX, USA), and Meta-DiSc V 1.4 (Unidad de Biostatistica Clinica, Hospital Ramon y Cajal, Madrid, Spain) for data management and calculations. Statistical heterogeneity was assessed using the I2 statistic.8 Metaregression was only performed for sensitivity as this is the parameter of main interest and there were enough studies to perform sufficient analysis. Likelihood ratio parameters, DOR, and summary-ROC estimates are based on studies where both diseased and non-diseased populations were available.

Reporting and grading of quality of evidence were performed according to the MOOSE, GRADE, QUADAS statement, respectively911 (Tables 1 and 2).

View this table:
Table 1

MOOSE statement

Provided at page no.
Reporting of background
 Problem definition3
 Hypothesis statement3
 Description of study outcome(s)3
 Type of exposure or intervention used3
 Type of study designs used3
 Study population3
Reporting of search strategy
 Qualifications of searchers (e.g. librarians and investigators)n.a.
 Search strategy, including time period included in the synthesis and keywords4
 Effort to include all available studies, including contact with authors4
 Databases and registries searched4
 Search software used, name and version, including special features used (e.g. explosion)4
 Use of hand searching (e.g. reference lists of obtained articles)n.a.
 List of citations located and those excluded, including justification4
 Method of addressing articles published in languages other than English4
 Method of handling abstracts and unpublished studiesn.a.
 Description of any contact with authorsn.a.
Reporting of methods
 Description of relevance or appropriateness of studies assembled for assessing the hypothesis to be tested4
 Rationale for the selection and coding of data (e.g. sound clinical principles or convenience)4
 Documentation of how data were classified and coded (e.g. multiple raters, blinding, and interrater reliability)4
 Assessment of confounding (e.g. comparability of cases and controls in studies where appropriate)n.a.
 Assessment of study quality, including blinding of quality assessors; stratification or regression on possible predictors of study results4
 Assessment of heterogeneity5
 Description of statistical methods (e.g. complete description of fixed or random effects models, justification of whether the chosen models account for predictors of study results, dose-response models, or cumulative meta-analysis) in sufficient detail to be replicated5
 Provision of appropriate tables and graphics16ff
Reporting of results
 Graphic summarizing individual study estimates and overall estimate16ff
 Table giving descriptive information for each study included16ff
 Results of sensitivity testing (e.g. subgroup analysis)8
 Indication of statistical uncertainty of findings8
Reporting of discussion
 Quantitative assessment of bias (e.g. publication bias)n.a.
 Justification for exclusion (e.g. exclusion of non-English-language citations)10
 Assessment of quality of included studies26
Reporting of conclusions
 Consideration of alternative explanations for observed resultsn.a.
 Generalization of the conclusions (i.e. appropriate for the data presented and within the domain of the literature review)10
 Guidelines for future research10
 Disclosure of funding source11
  • n.a.: non applicable.

View this table:
Table 2

QUADAS statement

Sbarouni et al. (2007)37Kiernan (2007)39Spinner et al. (2006)40Hazui et al. (2006)38D́Aloia et al. (2006)29Ohlmann et al. (2006)30Weber et al. (2006)3Alshakarchi et al. (2005)28Hazui et al. (2005)31Akutsu et al. (2005)32Eggebrecht et al. (2004)4Perez et al. (2004)33Shimazaki et al. (2003)34Weber et al. (2003)5Arrich et al. (2004)35Prattichizzo et al. (2001)36
Was the spectrum of patients representative of the patients who will receive the test in practice?++++++++++++++++
Were selection criteria clearly described?+O++O++O++++++OO
Is the reference standard likely to correctly classify the target condition?++++++++++++++++
Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests?++++++++++++++++
Did the whole sample or a random selection of the sample, receive verification using a reference standard of diagnosis?++++++++++++++++
Did patients receive the same reference standard regardless of the index test result?++++++++++++++++
Was the reference standard independent of the index test (i.e. the index test did not form part of the reference standard)?
Was the execution of the index test described in sufficient detail to permit replication of the test?+++++++++++++++
Was the execution of the reference standard described in sufficient detail to permit its replication?OOOOOOOOOOOOOOOO
Were the index test results interpreted without knowledge of the results of the reference standard?
Were the reference standard results interpreted without knowledge of the results of the index test?
Were the same clinical data available when test results were interpreted as would be available when the test is used in practice?++++++++++++++++
Were uninterpretable/ intermediate test results reported?
Were withdrawals from the study explained?OOOOOOOOOOOOOOOO
  • +: no; −: yes; o: undetermined.

Prospective cohort study

Sample size estimations for the prospective cohort study were precision based, since no formal hypothesis was tested. At a given sample of 60 patients and assuming one negative patient among those, the yielded exact confidence interval ranged from 0.9106 to 0.9996. Discrete data are given as counts and percentages. Linear (Pearson) correlation was applied to test for correlation between admission levels and time of onset of symptoms. We applied sensitivity analysis to validate proposed cut-off levels of D-dimer in AAD in our study population. We sought to determine the optimal threshold level to exclude the presence of AAD on the base of D-dimer measurement at the emergency department. Finally, we calculated negative predictive values for AAD for the different D-dimer cut-off levels based on the calculated test sensitivities and a previously published prevalence of AAD in unselected patients with chest pain at our institution.12 Calculations were performed with SPSS for Windows (version 10.0). A two-sided P-value less than 0.05 was considered statistically significant.

Results

Systematic review

The electronic search of databases resulted in 43 hits, and we retrieved 32 publications for closer inspection after deleting double-hits. Of these, 16 studies were either editorial comments,2,6,13 reviews,5,1419 studies in other languages as predefined,2,20,21 or non-diagnostic;5,2227 therefore 16 articles were then potentially eligible for further review35,2840 (Figure 1).

Figure 1

Progress through the stages in selecting studies for final analysis.

The clinical and methodological characteristics of the trials that reported at least one of the endpoints of interest are presented in Table 3. The diagnostic value of elevated D-dimer levels in AAD was evaluated in an overall population of 437 patients. In most of the instances, a cohort study design was applied;35,2934,38 in nine trials, a control group of patients with chest pain was included.4,5,3032,34,37,40

View this table:
Table 3

Clinical and methodological characteristics of trials, assessing the diagnostic value of D-dimer in acute aortic dissection

ReferenceStudy designAADNo. of control patientsCut-offType of assay
Sbarouni et al. (2007)37Prospective cohort study1829>0.7 µg/mLELISA
Kiernan (2007)39Case report1Not declaredNot declared
Spinner et al. (2006)40Not declared; cohort study2656>0.3 µg/mLLatex agglutination
Hazui et al. (2006)38Retrospective cohort study113>0.4 µg/mLLatex agglutination
D́Aloia et al. (2006)29Case report1Not declaredNot declared
Ohlmann et al. (2006)30Retrospective cohort study9494>0.4 µg/mLImmunoturbidimetric
Weber et al. (2006)3Prospective/retrospective cohort study27>0.5 µg/mLImmunoturbidimetric
Alshakarchi et al. (2005)28Case report1>0.5 µg/mLELISA
Alshakarchi et al. (2005)28Case report1>0.3 µg/mLLatex agglutination
Hazui et al. (2005)31Prospective cohort study2949>0.9 µg/mLLatex agglutination
Akutsu et al. (2005)32Prospective cohort study3048>0.5 µg/mLImmunoturbidimetric
Akutsu et al. (2005)32Prospective cohort study3048>0.1 µg/mLELISA
Eggebrecht et al. (2004)4Prospective cohort study1680>0.62 µg/mLLatex agglutination
Perez et al. (2004)33Retrospective cohort study7>0.5 µg/mLLatex agglutination
Shimazaki et al. (2003)34Prospective cohort study2931>0.8 µg/mLNot declared
Weber et al. (2003)5Prospective/retrospective cohort study2435>0.5 µg/mLImmunoturbidimetric
Arrich et al. (2004)35Case report1>0.5 µg/mLLatex agglutination
Prattichizzo et al. (2001)36Case report1>0.5 µg/mLNot declared
  • Note that Weber et al. published data in 2006 from a previous study in 2003; Akutsu et al. applied two different assays simultaneously. D-dimer has been evaluated in a total of 437 patients.

Measurement of D-dimer levels was performed using three different types of assays; ELISA in three,28,32,37 immunoturbidimetric in four,3,5,30,32 and latex agglutination test in another seven studies.4,12,23,25,30,33,36 In one study, two different assays were evaluated simultaneously.32

The reported diagnostic cut-off values ranged between 0.1 and 0.9 µg/mL. Absolute D-dimer levels in patients with AAD are given in detail in Table 4 (ranged between mean 0.98 and 10.10 µg/mL).

View this table:
Table 4

As total quantitative D-dimer values differed among the studies, cut-off values were highly variable, ranging from 0.1 to 0.9 µg/mL

ReferenceD-dimer in AADD-dimer negative
Sbarouni et al. (2007)37Mean 4.63 µg/mL; SD 3.011
Kiernan (2007)39>1.00 µg/mL
Spinner et al. (2006)40Mean 3.12 µg/mL; SD 2.202
Hazui et al. (2006)38Mean 3.69 µg/mL; SD 7.319
D́Aloia et al. (2006)29Positive
Ohlmann et al. (2006)30Mean 8.61 µg/mL; SD 12.601
Weber et al. (2006)3Mean 3.55 µg/mL; SD 6.40
Alshakarchi et al. (2005)28>5.00 µg/mL
Alshakarchi et al. (2005)2814.36 µg/mL
Hazui et al. (2005)31Mean 0.97 µg/mL; SD n.a.2
Akutsu et al. (2005)32Mean 1.25 µg/mL; SD 1.31
Akutsu et al. (2005)32Mean 1.80 µg/mL; SD 1.18
Eggebrecht et al. (2004)4Mean 2.36 µg/mL; SD 1.82
Perez et al. (2004)33Positive (>0.50 µg/mL)
Shimazaki et al. (2003)34Mean 6.80 µg/mL; SD 10.80
Weber et al. (2003)5Mean 9.40 µg/mL; SD 40.05
Arrich et al. (2004)354.40 µg/mL
Prattichizzo et al. (2001)363.20 µg/mL
  • An overall of 15 patients was reported to be D-dimer negative.

Fifteen of 437 patients with AAD were reported to be D-dimer negative,30,31,37,38,40 test sensitivity ranged from 92 to 100%; further details are presented in Table 5.

View this table:
Table 5

Sensitivity and specifity of D-dimer measurement in diagnosis of acute aortic dissection

ReferenceCut-off level (µg/mL)Sensitivity in AAD (%)Specifity in AAD (%)
Sbarouni et al. (2007)37>0.79459
Spinner et al. (2006)40>0.392
Hazui et al. (2006)38>0.492
Ohlmann et al. (2006)30>0.49934
Weber et al. (2006)3>0.5
Hazui et al. (2005)31>0.993.1
Akutsu et al. (2005)32>0.510054
Akutsu et al. (2005)32>0.110054
Eggebrecht et al. (2004)4>0.6210073
Perez et al. (2004)33>0.5100
Shimazaki et al. (2003)34>0.8
Weber et al. (2003)5>0.510068. 9

Overall, the observer agreement regarding the various components of the data extraction form reached nearly 100%. Both reviewers assigned a high clinical grade of evidence of measurement of D-dimer levels in exclusion of AAD.

Meta analysis

D-dimer testing provided high sensitivity (0.97 95% CI 0.94–0.98, I2 = 47.1%) and negative likelihood ratios (0.06 95% CI 0.02–0.13, I2 = 0.0%) within narrow confidence intervals. Summary receiver operating characteristic (SROC) curve yielded a high certainty for excluding AD on base of negative results (AUC 0.94) (Figure 2).

Figure 2

Summary receiver operating characteristic (SROC) curve for excluding AAD on base of negative D-dimer testing (AUC 0.94).

Pooled specifity (0.59 95% CI 0.53–0.64, I2 = 0.0%) and positive likelihood ratios (2.58 95% CI 1.76–3.78, I2 = 0.0%) did not greatly increase the certainty of diagnosis of AD. The pooled diagnostic odds ratio was 21.27 (95% CI 11.64–38.88, I2 = 0.0%).

Meta regression revealed no influence of the cut-off value on sensitivity (DOR 1.00 95% CI 0.72–1.40, P = 0.99).

Validation of test sensitivity in a prospective cohort of acute aortic dissection patients

We identified 65 out of 70 patients with AAD diagnosed at our institution who met the inclusion criteria for further analysis (36 male, 55%; median age 59 years, IQR 49–67) (Figure 3). The median time from onset of leading symptoms to hospital admission was 4.8 h (IQR 2.4–16). Acute chest pain was the leading symptom in 43 (66%) patients. D-dimer levels at initial presentation were scattered over a wide range from 0.24 to 137.88 µg/mL (median 3.47; IQR 1.55–14.49), respectively (Figure 4). There was no correlation between admission level and time of onset of symptoms (corrected R2: 0.008, P = 0.48). Further demographic details are depicted in Table 6.

Figure 3

Trial flow of the prospective cohort study.

Figure 4

There was no correlation between D-dimer levels and time of onset of symptoms (corrected R2: 0.008, P = 0.48; n = 65).

View this table:
Table 6

Descriptive characteristics of the cohort

Descriptive characteristics of the cohort
n overall = 65
Demographics
 Age, median (IQR)59 (49–67)
 Female, n (%)29 (45)
Chronic health conditions and risk factors
 Arterial hypertension, n (%)51 (78)
 Chronic obstructive pulmonary disease, n (%)12 (19)
 Connective tissue disease, n (%) 8 (12)
 Preknown coronary artery disease, n (%) 6 (9)
 Diabetes, n (%) 5 (8)
 Serum creatinine >200 mmol/L, n (%) 5 (8)
 Extracardiac arteriopathy, n (%) 3 (5)
Leading symptoms
 Chest pain, n (%)43 (66)
 Syncope, n (%)19 (29)
 Other, n (%) 3 (5)
 Onset (h), median (IQR) 4.8 (2.4–16)
In-hospital clinical assessment
 Critical preoperative state, n (%)12 (19)
 Pericardial effusion, n (%)25 (38)
 D-dimer (µg/mL), median (IQR) 3.47 (1.55–14.49)
  • Unless otherwise indicated, data are number (percentage). IQR, interquartile range; critical preoperative state defined any one or more of the following conditions: ventricular tachycardia or fibrillation or aborted sudden death, preoperative cardiac massage, preoperative ventilation before arrival in the anaesthetic room, preoperative inotropic support, or preoperative acute renal failure.

Sensitivity analysis

Applying the lowest reported cut-off value of 0.1 µg/mL, 100% sensitivity to exclude AAD based on D-dimer measurement could be achieved in our study population. Applying 0.5 µg/mL, as proposed by 7 out of 16 trials,3,5,28,32,35,36,38 diagnosis of AAD would have been missed in one patient (sensitivity 98%) in our cohort. Diagnostic sensitivity further decreases to 86% at highest reported threshold level of 0.9 µg/mL;31 applying this cut-off would have resulted in nine false-negative patients in our cohort.

Negative predictive values for different D-dimer cut-off levels ranged from 92 to 100%, assuming a prevalence of 0.23% of AAD in unselected patients presenting with acute chest pain to an emergency department12 (Figure 5).

Figure 5

Negative predictive value analysis of D-dimer testing, stratified to different cut-off values, arbitrary to individual test specifity.

Discussion

We present a systematic review assessing the diagnostic sensitivity of serum D-dimer measurement in the clinical setting of AAD. We found that AAD can be ruled out safely by negative D-dimer testing.

Serum D-dimer is elevated due to endogenous fibrinolytic activity naturally counteracting the activation of the extrinsic pathway of the coagulation cascade.36 Routine measurement has already been recommended as part of the pre-operative work-up with proven AAD from the Task Force of the European Society of Cardiology.2 Recognized to be a reliable biomarker to exclude pulmonary embolism,41 it was first to Weber et al.5 in 2003 to report in a small case series about the clinical value of D-dimer testing in AAD. In subsequent studies, the value of D-dimer testing in AAD has been evaluated in a total of 437 patients so far.

Currently, diagnosis and exclusion of AAD remain challenging, since clinical examination is insufficient to exclude AAD and no biomarker is available for clinical routine;1,42 guidelines recommend to perform transesophageal echocardiography, computed tomography, or even angiography in any case of suspected aortic dissection.2 These diagnostic procedures are more or less invasive and expensive, with limited availability at different settings, and bear potential serious side effects. Therefore, D-dimer measurement for exclusion of AAD represents an advance in diagnostic pathways in patients with suspected AAD.

Our systematic review, meta-analysis, and prospective cohort study and a very recent meta-analysis by Marill43 show that D-dimer may represent a highly sensitive biomarker: exclusion of AAD based on D-dimer levels at a threshold of 0.1 µg/mL may be achieved with 100% sensitivity. At a cut-off level of 0.5 µg/mL, which represents the most commonly clinical used value for exclusion of pulmonary embolism,41 a negative predictive value of 99% will be still achieved. However, we found the reported diagnostic cut-off values to be heterogeneous, ranging between 0.1 and 0.9 µg/mL. Further-on, three different assay types have been applied in measurement of quantitative D-dimer values in the reported studies.35,28,3033,35,37,38,40 This might explain, why a total of 15 patients were reported to be D-dimer false negative. It should be therefore kept in mind that particular D-dimer assays with predefined cut-off levels will differ in sensitivity and negative predictive value in patients with suspected AAD.41

Applying D-dimer testing as a rapid point of care test will help to risk stratify patients at a very early time-point and might result in a significant reduction for advanced imaging procedures. It has to be noticed that D-dimer measurement is only unidirectional due to its lack of specificity; a negative result should be used in the diagnostic pathway to exclude AAD: in patients with a low and intermediate suspicion of AAD, a negative D-dimer may be sufficient to rule out disease. In patients with a high clinical likelihood of AAD, however, appropriate imaging should be still performed, even in presence of negative D-dimer testing, since prospective controlled studies for this strategy are missing.

A positive result will be found in patients with AAD and other life-threatening conditions like pulmonary embolism or non-serious clinical conditions. This fact somewhat might limit the clinical utility of the D-dimer testing in AAD.

Limitations

Only two papers were excluded because there was no abstract available in English or in another predefined language.20,21 From their titles, it was doubtful that these were clinical trials reporting about the diagnostic sensitivity of D-dimer measurement in AAD, but we could not confirm. Some limitations of the drawn conclusions from the meta-analysis, however, have to be acknowledged. D-dimer assays will differ in sensitivity and negative predictive value in patients with suspected AAD. Even more, some D-dimer assays coefficients of variation will be high at our proposed 0.1 µg/mL cut-off value. Therefore, the most straightforward approach for determining the accuracy and cut-off of a diagnostic test would be to carry out a prospective, cross sectional study in unselected patients, with independent, blinded assessments of test and reference methods.

In patients with intermediate suspicion of AAD sample size of our own study population is modest when it comes to random variability; therefore, our findings ideally need confirmation in a larger cohort of patients. Selection bias might be a minor confounder, since 93% of the screening population could be included in our study design. However, this was a prospective, observational case series of consecutive patients with proven AAD and sensitivity of D-dimer measurement in exclusion of AAD in our study population is in line with previous investigations.

Conclusion

Up to now, various non-invasive and invasive diagnostic steps are required to diagnose or to rule-out AAD in case of clinical suspicion. Determination of D-dimer in clinical routine may represent a valuable addition to the current diagnostic work-up of patients with suspected AAD. The measurement of this biomarker is likely to help triage patients and stratify them according to the urgency for more sophisticated diagnostic steps. Appropriate imaging is always required for definite diagnosis and confirmation of clinically suspected AAD. A negative D-dimer testing however will exclude the presence of AAD in up to 100% of instances.

Contributions

G.H.S.: study design, data management, statistics, first draft, final draft, and approval; H.D.: study design, scientific and statistic advisory, study monitoring, final draft, and approval; M.S.: statistics, final draft, and approval; M.P.E.; data collection, final draft, and approval; G.E.: analysis, data collection, final draft, and approval; H.H.: design, statistics, final draft, and approval; A.N.L.: design, scientific and statistic advisory, final draft, and approval.

Conflict of interest statement. None declared.

References

View Abstract