SEARCH

SEARCH BY CITATION

Keywords:

  • deep vein thrombosis;
  • post-thrombotic syndrome;
  • risk factors;
  • venous insufficiency;
  • venous thromboembolism;
  • Villalta scale

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Background

Post-thrombotic syndrome (PTS) is the most frequent complication of deep vein thrombosis (DVT). Its diagnosis is based on clinical characteristics. However, symptoms and signs of PTS are non-specific, and could result from concomitant primary venous insufficiency (PVI) rather than DVT. This could bias evaluation of PTS.

Methods

Using data from the REVERSE multicenter study, we assessed risk factors for PTS in patients with a first unprovoked unilateral proximal DVT 5–7 months earlier who were free of clinically significant PVI (defined as absence of moderate or severe venous ectasia in the contralateral leg).

Results

Among the 328 patients considered, the prevalence of PTS was 27.1%. Obesity (odds ratio [OR] 2.6 [95% confidence interval (CI) 1.5–4.7]), mild contralateral venous ectasia (OR 2.2 [95% CI 1.1–4.3]), poor International Normalized Ratio (INR) control (OR per additional 1% of time with INR < 2 during anticoagulant treatment of 1.018 [95% CI 1.003–1.034]) and the presence of residual venous obstruction on ultrasound (OR 2.1 [95% CI 1.1–3.7]) significantly increased the risk for PTS in multivariable analyses. When we restricted our analysis to patients without any signs, even mild, of contralateral venous insufficiency (n = 244), the prevalence of PTS decreased slightly to 24.6%. Only obesity remained an independent predictor of PTS (OR 2.6 [95% CI 1.3–5.0]). Poor INR control and residual venous obstruction also increased the risk, but the results were no longer statistically significant (OR 1.017 [95% CI 0.999–1.035] and OR 1.7 [95% CI 0.9–3.3], respectively).

Conclusions

After a first unprovoked proximal DVT, obese patients and patients with even mild PVI constitute a group at increased risk of developing PTS for whom particular attention should be paid with respect to PTS prevention. Careful monitoring of anticoagulant treatment may prevent PTS.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Post-thrombotic syndrome (PTS) is the most common complication of deep vein thrombosis (DVT), and develops in 20–50% of patients after a proximal DVT [1-3]. In the absence of an available objective test, its diagnosis is clinical [1]. However, it is acknowledged that the clinical presentation of PTS may be non-specific, and conditions other than DVT, such as primary venous insufficiency (PVI), may produce similar signs or symptoms in the lower extremities, leading to potential misdiagnoses [4-6]. In addition, most identified risk factors for PTS are also reported to be risk factors for venous insufficiency [4, 7]. Therefore, previous venous insufficiency might bias the evaluation of PTS. Determining predictors of PTS in patients free from PVI, the prevalence of which is at least three to four times higher in the population than secondary chronic venous disorders such as PTS, would be of interest to determine true predictors of PTS that might be amenable to measures to prevent PTS [7, 8].

In this perspective, we analyzed data from the REcurrent Venous thromboembolism Risk Stratification Evaluation (REVERSE) study, to assess the risk factors for PTS in patients with a first, unilateral, DVT 5–7 months previously and free from significant PVI [9].

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

The REVERSE study is a large, international (Canada, France, Switzerland, and the USA), multicenter, prospective cohort study of patients with a first unprovoked venous thromboembolism (VTE) (proximal DVT or pulmonary embolism [PE]). Its primary objective was to derive a clinical decision rule to identify unprovoked VTE patients at low risk of recurrence who could potentially discontinue anticoagulant therapy. The study protocol has been extensively described elsewhere, and is available at ClinicalTrials.gov (registration number: NCT00261014) [9]. Approval from the institutional research ethics board was obtained at all participating centers. In this substudy, we focused on REVERSE study patients free of significant PVI whose index VTE event was a first unilateral proximal DVT.

Patients

Unselected consecutive patients with a first objectively proven symptomatic unprovoked VTE who had been treated for at least 5 days with heparin followed by 5–7 months of oral anticoagulant therapy were eligible for the REVERSE study. Patients were excluded if: (i) their VTE was a recurrent unprovoked VTE; (ii) if their VTE was provoked by a transient risk factor – immobilization for > 3 days, surgery with the use of a general anesthetic in the 3 months prior to the index VTE event, leg fracture or lower extremity plaster cast; (iii) their VTE was provoked by a chronic risk factor – diagnosis of malignancy in the previous 5 years, high-risk thrombophilia (homozygous for factor V Leiden or prothrombin gene mutation [PGM] or compound heterozygous for FV Leiden and PGM or any of antithrombin, protein C or protein S deficiencies, and lupus anticoagulant/antiphospholipid antibody positivity); or (iv) they refused to provide consent.

For this substudy, we included REVERSE patients whose index VTE event was a first unilateral proximal DVT (with or without PE) and who did not have evidence of significant PVI, defined as absence of moderate or severe venous ectasia in the leg contralateral to the DVT, and who had a PTS assessment performed at study baseline (or at the 6-month follow-up visit, if study baseline PTS assessment was missing). We excluded patients with PE without DVT, bilateral DVT, or a history of previous secondary VTE, in order to focus on a population at risk for incident PTS. Patients in whom a PTS assessment was only available for the 6-month follow-up visit and who had experienced DVT recurrence between the study baseline and the 6-month follow-up visit were also excluded.

Study protocol

At study baseline, i.e. 6 months after the index VTE event, trained research nurses collected data on demographic characteristics, risk factors for VTE at the time of the index event, thrombophilia testing, medications, and compression stocking use, and reviewed imaging reports confirming the index VTE. Imaging of the leg that was symptomatic at the time of the index event was also performed, with compression ultrasonography (CUS). Patients attended follow-up visits every 6 months thereafter for up to 4 years. An evaluation for PTS was performed by research nurses at each study visit. Patients were instructed to contact study personnel if they developed symptoms of recurrent VTE during follow-up, so that objective testing could be performed [9]. All recurrent VTE events were adjudicated by the study's expert adjudication committee.

Risk factors

The influence of the following potential risk factors or confounders for PTS was analyzed: (i) demographic characteristics – age, sex, income, ethnicity (Caucasian vs. non-Caucasian), and mild venous insufficiency (e.g. mild contralateral venous ectasia); (ii) characteristics of the index VTE – pregnancy/postpartum/hormonal therapy related VTE, clinical presentation of the index VTE event (DVT vs. DVT + PE); (iii) thrombophilia – family history of VTE, biological thrombophilia (FV Leiden or PGM mutation, elevated FVIII, hyperhomocysteinemia, presence of a lupus anticoagulant or anticardiolipin antibodies, serum lipoprotein A); (iv) hemogram and inflammatory markers – platelet count, hemoglobin, D-dimer after discontinuing anticoagulants and C-reactive protein levels; (v) residual venous obstruction (wall thickening or worse) on CUS; (vi) quality of International Normalized Ratio (INR) control assessed with to Rosendaal's method [10]; and (vii) concomitant medications/devices – statin, antiplatelet and non-steroidal anti-inflammatory drugs, drugs for venous insufficiency (horse chestnut seed extract and garlic pills), and use of elastic compression stockings.

Study endpoint

The outcome for the present study was the presence of a PTS at the study baseline visit or 6-month follow-up visit (if baseline scores were missing), defined according to ISTH standards, namely Villalta score of > 4 in the leg ipsilateral to the DVT [1]. The Villalta scale is a clinical measure that incorporates ratings of six clinician-rated venous signs (venous ectasia, pretibial edema, skin induration, hyperpigmentation, redness, and pain on calf compression) and five patient-rated venous symptoms (pain, cramps, heaviness, paresthesia, and pruritus) in the leg ipsilateral to the DVT. For each symptom or sign, the severity is graded from 0 to 3 points (no, mild, moderate, and severe), and the points are summed; PTS was categorized as mild if the score was 5–9, moderate if the score was 10–14, and severe if the score was > 14 or an ulcer was present.

Statistical analysis

Categorical variables were expressed as frequency and percentage, and continuous variables were expressed as mean and standard deviation. We hypothesized that the primary predictors for PTS would be age, female gender, CUS abnormalities, and poor INR control [4]. The last of these variables was a continuous variable that was defined as percentage of time with INR < 2. Its calculated odds ratio (OR) associated with the risk of PTS corresponds to the excess of risk of PTS per additional 1% of time with INR < 2 during anticoagulant treatment. A multivariable model using a backward elimination procedure with sex, age, INR control and compression stocking use in the model in addition to all variables listed in the risk factor section achieving a P-value of ≤ 0.10 in univariate analyses was used. To test the robustness of our model (sensitivity analyses), we estimated the ORs of the variables retained in our final multivariable model in two subgroups of patients, using more restrictive definitions of venous insufficiency: those without any contralateral venous ectasia and those with a Villalta score of ≤ 4 in the leg contralateral to the DVT. This latter subgroup was chosen in order not to misclassify patients who are without significant venous signs (moderate or severe venous ectasia) but who have significant venous symptoms (or other signs of venous insufficiency) as being free of PVI. Two-sided P-values of ≤ 0.05 were considered to be statistically significant. Statistical analyses were performed with sas for Windows (version 9.2; SAS Institute, Cary, NC, USA).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Of the 664 REVERSE patients, 328 patients fulfilled our criteria for participation in this substudy (Fig. 1). Their demographic and clinical characteristics are shown in Table 1. Among those 328 patients, 244 patients were free of any contralateral venous ectasia, and 65 patients had mild contralateral venous ectasia. Information on the severity (none or mild) of venous ectasia was missing for 19 patients.

Table 1. Demographic and clinical characteristics of the study population
  1. DVT, deep vein thrombosis; PE, pulmonary embolism; SD, standard deviation. *Defined as body mass index of ≥ 30 kg/m2.

Number of patients, n328
Age (years), mean (SD)52.7 (17.2)
Male, % (n)57.0 (187)
Caucasian ethnicity, % (n)91.8 (301)
Obesity, % (n)*39.9 (131)
Mild contralateral venous ectasia, % (n)25.6 (84)
Characteristics of index DVT
Isolated DVT, % (n)77.7 (255)
DVT with PE, % (n)22.3 (73)
Any use of compression stockings, % (n)49.1 (161)
image

Figure 1. Study flow chart. DVT, deep vein thrombosis; PE, pulmonary embolism; PTS, post-thrombotic syndrome; PVI, primary venous insufficiency; VTE, venous thromboembolism.

Download figure to PowerPoint

Prevalence of PTS at baseline

Among patients without significant PVI 5–7 months after the index DVT event (n = 328), 27.1% (n = 89) showed PTS, which was mild, moderate and severe in 79.7% (n = 71), 16.9% (n = 15) and 3.4% (n = 3) of cases, respectively. One patient developed an ipsilateral leg ulcer.

In the subgroup of patients without any contralateral venous ectasia (n = 244), the prevalence of PTS was 24.6% (n = 60), and it was mild, moderate and severe in 81.7% (n = 49), 15.0% (n = 9) and 3.3% (n = 2) of cases, respectively.

In the subgroup of patients with a Villalta score of ≤ 4 in the contralateral leg (n = 298, 274 patients analyzed), the prevalence of PTS was 21.2% (n = 58), and it was mild, moderate and severe in 89.7% (n = 52), 6.7% (n = 4) and 3.4% (n = 2) of cases, respectively.

Risk factors for PTS

Obesity, mild contralateral venous ectasia, poor INR control and the presence of residual venous obstruction on ultrasonography significantly increased the risk of PTS in univariable and multivariable analyses (Table 2).

Table 2. Predictors of post-thrombotic syndrome (PTS) in patients without clinically significant contralateral venous insufficiency (univariate and multivariable analyses)
Predictors of PTS Univariate analysis OR (95% CI) Multivariable analysis‡‡ OR (95% CI)
  1. CI, confidence interval; CRP, C-reactive protein; DVT, deep vein thrombosis; NA, not assessable; INR, International Normalized Ratio; NSAID, non-steroidal anti-inflammatory drug; OR, odds ratio; PE, pulmonary embolism. †The influences of the following risk factors or confounders were also studied, but the results were not statistically significant: height, pregnancy and postpartum, inferior vena cava filter, family history of venous thromboembolism, dyslipidemia, levels of hemoglobin, or anticardiolipin antibody, apolipoprotein A, estroprogestative drugs use prior to DVT, clopidogrel, dipyridamole, and drugs for venous insufficiency. ‡The final multivariable model was selected by use of a backward elimination procedure with sex, age, poor INR control, and compression stocking use in the model, in addition to all variables achieving a P-value of ≤ 0.10 in univariate analyses. §Body mass index of ≥ 30 kg/m2. ¶Mild venous ectasia. ††Per additional 1% of time for which INR < 2 during the full period of anticoagulant treatment, which has a range of 0–88%. ‡‡For any 25% increase in percentage time with poor INR control, the risk of developing PTS increases by 56%, which is represented by an OR of 1.56 = exp(log[1.018] × 25). §§Minimal wall thickening or worse. *P ≤ 0.05. **0.05 < P ≤ 0.10.

Age (per year)0.998 (0.984–1.012)1.004 (0.987–1.022)
Female sex1.345 (0.825–2.194)1.596 (0.890–2.862)
Income ($/year)
< 25 000 vs. > 80 0003.309 (1.472–7.440)*
25 000–49 999 vs. > 80 0002.149 (0.929–4.968)**
50 000–80 000 vs. > 80 0001.688 (0.738–3.860)
Caucasian1.070 (0.436–2.624)
Index DVT + PE (vs. DVT alone)1.210 (0.683–2.146)
Obesity§2.186 (1.333–3.586)*2.627 (1.469–4.699)*
Mild contralateral venous insufficiency1.917 (1.075–3.419*2.222 (1.142–4.322)*
Thrombophilia
FV Leiden0.743 (0.387–1.426)
Prothrombin gene mutation1.372 (0.535–3.519)
Elevated FVIII1.087 (0.706–1.674)
Hyperhomocysteinemia0.994 (0.967–1.023)
Presence of lupus anticoagulantNA
Platelet count1.003 (1.000–1.007)**
Inflammatory markers at baseline
Elevated CRP0.978 (0.936–1.022)
Elevated D-dimer1.000 (0.999–1.001)
Poor INR control††1.015 (1.002–1.029)*1.018 (1.003–1.034)*,‡‡
Residual venous obstruction on baseline ultrasound§§1.666 (0.995–2.788)**2.054 (1.126–3.744)*
Medications
Any NSAID or aspirin use1.874 (0.889–3.948)**
Any statin use1.695 (0.906–3.170)**
Any use of compression stockings1.153 (0.709–1.877)1.357 (0.761–2.420)

In the subgroup of patients without any contralateral venous ectasia (n = 244), obesity (OR 2.5 [95% CI 1.4–4.6]), residual venous obstruction (OR 1.6 [95% CI 0.9–2.9]) and poor INR control (OR 1.02 [95% CI 1.00–1.03]) increased the risk of PTS in univariate analyses. Multivariable analyses confirmed that obesity significantly increased the risk of PTS (OR 2.6 [95% CI 1.3–5.0]). Residual venous obstruction and poor INR control also increased the risk of PTS, but the results were not statistically significant (OR 1.7 [95% CI 0.9–3.3] and OR 1.02 [95% CI 0.99–1.04], respectively).

Among the 298 patients with a Villalta score of ≤ 4 in the contralateral leg, obesity (OR 2.8 [95% CI 1.5–5.0]), mild contralateral venous ectasia (OR 1.9 [95% CI 1.0–3.7]), residual venous obstruction (OR 2.1 [95% CI 1.1–3.9]) and poor INR control (OR 1.01 [95% CI 1.00–1.03]) increased the risk of PTS in univariate analyses. These results were confirmed in multivariable analyses: obesity (OR 2.9 [95% CI 1.5–5.6]), mild contralateral venous ectasia (OR 2.6 [95% CI 1.2–5.6]) and residual venous obstruction (OR 2.4 [95% CI 1.2–4.8]) significantly increased the risk of PTS. Poor INR control also increased the risk of PTS, but the results were not statistically significant (OR 1.01 [95% CI 1.00–1.03]).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

We found that, among patients with a first, unprovoked, unilateral proximal DVT who were free of clinically significant PVI, one-quarter showed PTS 5–7 months after the index event. Obesity, mild contralateral venous ectasia, poor INR control and the presence of ultrasonographic residual venous obstruction significantly increased the risk of PTS in this population.

To our knowledge, PTS in DVT patients free of previous venous insufficiency has been studied in only one small single-center cohort (95 patients) [11]. PTS was assessed in the same way as in the REVERSE study (i.e. at one time point, with the Villalta scale), but the follow-up was longer (51 months) and one-third of DVTs were distal. The prevalence of PTS was similar to ours (28.4%), but no risk factors for PTS could be identified. Therefore, our study is the first to detect clinical predictors for PTS in such patients. We found that obesity was a risk factor for PTS, irrespective of the presence of any degree of concomitant venous insufficiency. In addition, even mild clinical expressions of venous insufficiency increased the risk of PTS, suggesting that underlying PVI could play a key role in PTS pathophysiology. This is in agreement with recent findings from a Dutch cohort, where the presence of varicosities was the strongest risk factor for PTS after acute DVT (OR 13.4 [95% CI 3.0–59.1]) [12].

Our results could have important implications for the routine management of patients with acute DVT. As they constitute a group at high risk of developing PTS, obese patients or patients with even mild venous insufficiency should deserve particular attention with respect to PTS prevention (e.g. patient education and closer surveillance for compression stocking compliance). Our results also confirm the importance of careful monitoring of anticoagulant treatment [13]. This may explain why, at least in part, by providing an optimal and constant anticoagulant intensity, heparin was found to be superior to vitamin K antagonists in preventing PTS [5]. This also suggests that new oral direct inhibitors could have a significant impact on the PTS disease burden [14, 15].

Finally, and as previously described, ultrasound evidence of residual venous obstruction was also associated with PTS [16]. However, as residual venous obstruction is part of the pathophysiology of PTS, it may be a marker of the disease rather than a true predictor [4].

Our population deserves some comment. It was extracted from a large multicenter international cohort, and an important number of potential risk factors (n = 34), collected at study baseline and identified a priori, were assessed. As data were censored after a first recurrent VTE event, which occurred in 12.2% (n = 40) of our population during the first 2 years, we chose to only assess PTS at study baseline (i.e. 5–7 months after the acute DVT). Therefore, we may have failed to identify some cases of PTS that would have subsequently developed during follow-up. However, our assessment is in agreement with ISTH standards [1]. In addition, as patients were recruited 5–7 months after DVT, our results need to be confirmed by studies assessing venous status before or at least at the time of the index DVT event. However, we believe that the impact on our results should be minimal, because PVI, the leading cause of venous insufficiency, is typically bilateral [7, 17, 18]. Hence, assessment of venous status in the contralateral leg should provide a reliable estimate of venous status in the ipsilateral leg before the occurrence of the index DVT in our population of patients without a history of previous DVT. We also cannot exclude the possibility that some patients who were excluded because of contralateral venous insufficiency may have had undiagnosed contralateral asymptomatic DVT, e.g. at the time of the index event, as the cause of their contralateral venous insufficiency. However, estimates from the published literature suggest that bilateral DVT in patients with unprovoked DVT is rare (< 6%) [19]. We therefore think that this would be unlikely to modify our results, as < 10 patients would have been wrongly excluded.

We believe that our definition of significant clinical PVI (moderate to severe contralateral venous ectasia without including patients with only mild contralateral venous ectasia) is valid and not too restrictive, and it allowed us to study predictors of PTS in a large population with minimum overlap between PTS and chronic PVI. Indeed, venous ectasia is the most common sign of venous insufficiency, and its mildest form, telangiectasia, is not usually considered to be an important sign of venous insufficiency according to the CEAP classification [17, 20]. We performed a sensitivity analysis on two separate subgroups of patients corresponding to more restrictive definitions of venous insufficiency: patients without any signs of contralateral venous ectasia (n = 84 excluded) and those with a Villalta score of < 5 in the contralateral leg (n = 30 excluded). The prevalence of PTS decreased slightly, particularly in the latter subgroup, which is consistent with previous findings of the REVERSE study showing that Villalta scores were strongly correlated between legs [21]. However, the OR associated with each risk factor was similar in magnitude (i.e. only CI differed) between main analysis and sub-group analyses.

Our definition of venous insufficiency, namely moderate or severe contralateral venous ectasia, has the advantage of being based on a single criterion, and is easy to apply by any physician. This kind of evaluation is in widespread use by both general practitioners and specialists in vascular medicine in everyday clinical practice. Furthermore, it takes into account only well-recognizable clinical signs (telangiectasia or the absence of any ectasia). As our definition did not exclude many patients (11%, vs. 34% if patients with any venous ectasia are excluded), our findings are likely to apply to a large population.

In conclusion, if confirmed, our results suggest that, after a first unprovoked proximal DVT, obese patients or patients with even mild venous insufficiency constitute a group at high risk of developing PTS for whom particular attention should be paid with respect to PTS prevention. They also confirm the importance of careful monitoring of anticoagulant treatment, and underline the potential impact of the choice of the type of anticoagulant on PTS occurrence.

Addendum

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

M. A. Rodger, M. J. Kovacs, S. R. Kahn, G. Le Gal, P. S. Wells, and D. R Anderson: conception and design of the REVERSE study; J. P. Galanaud, C. Holcroft, and S. R. Kahn: conception and design of the manuscript; P. S. Wells, M. A. Rodger, and R. H. White: administrative support; M. A. Rodger, M. J. Kovacs, S. R. Kahn, P. S. Wells, and D. R Anderson: grant funding; all authors: data acquisition; J. P. Galanaud, C. Holcroft, M. Betancourt, and S. R. Kahn: data analysis; J. P. Galanaud, C. Holcroft, and S. R. Kahn: drafting of the article; M. J. Kovacs, D. Anderson, M. A. Rodger, and G. Le Gal: critical revision for important intellectual content; all authors: final approval of the manuscript.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

We thank the staff and patients from the thrombosis clinics that participated in the study.

Disclosure of Conflict of Interests

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

J. P. Galanaud was funded by a grant from the French Society of Vascular Medicine. M. A. Rodger is the recipient of the Maureen Andrew New Investigator Award and a Career Investigator Award from the Heart and Stroke Foundation of Canada. P. S. Wells is a recipient of Canada research chair in venous thromboembolism. M. A. Crowther holds a Career Investigator Award from the Heart and Stroke Foundation. S. R. Kahn is a recipient of a National Investigator Award from the Fonds de la Recherche en Santé du Québec. This study was funded by the Canadian Institutes of Health Research (Grant no. MOP 64319) and BioMerieux (through an unrestricted research grant). J.-P. Galanaud had full access to all of the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis. BioMerieux had no role in the design or conduct of the study, the collection, management, analysis or interpretation of the data, or the preparation, review or approval of the manuscript. The Canadian Institutes of Health Research reviewed the design of the study, but had no role in the conduct of the study, the collection, management, analysis or interpretation of the data, or the preparation, review or approval of the manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References