Professional Documents
Culture Documents
Nadi 2019
Nadi 2019
ANZJSurg.com
Sinan Nadi ,*† Thomas D. Vreugdenburg ,* Yasoba Atukorale ,* Ning Ma ,* Guy Maddern *‡ and
Maroeska Rovers †
*Research and Evaluation, Incorporating ASERNIP-S, Royal Australasian College of Surgeons, Adelaide, South Australia, Australia
†Department of Epidemiology, Biostatistics and HTA, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands and
‡Discipline of Surgery, The University of Adelaide and The Queen Elizabeth Hospital, Adelaide, South Australia, Australia
aspirin as an appropriate prophylaxis for VTE,2 while the current • Study characteristics: study design, country, setting, funding
guideline from the National Institute for Health and Care Excel- and conflicts of interest.
lence (NICE) does not.5 Therefore, the aim of this systematic • Population: sample size, losses to follow up, age, gender, body
review is to assess the risks and benefits of aspirin compared to mass index, risk factors.
enoxaparin for patients undergoing total hip or knee arthroplasty. • Interventions: prophylactic agent used (e.g. dose, duration),
concomitant interventions.
• Outcomes: safety (bleeding events, wound infections, other
Methods adverse events), primary effectiveness outcomes (mortality,
A systematic literature search was conducted. The Population, rates of symptomatic DVT or PE), secondary effectiveness out-
Intervention, Comparator, Outcome (PICO) and study selection cri- comes (radiological evidence of asymptomatic PE or DVT).
teria were defined a priori (Table 1), as defined in a research proto-
col (PROSPERO id CRD42018090050). Data analyses
We conducted a random-effects network meta-analysis (NMA)
Search strategy and study selection (software: R version 3.5.0). The NMA combines direct and indirect
Searches were conducted in the PubMed, Embase and Cochrane evidence across a network of RCTs into a single effect size, under
Library databases up to 21 February 2018. All citations were the assumptions it can increase the precision in the estimates. risk
reviewed by the first author (SN) based on title and abstract; full- ratios (RRs) and 95% confidence intervals (CIs) were produced to
text screening was performed by two independent reviewers estimate the pairwise comparative treatment effect, and I2 was cal-
(SN and BH). Disagreements were resolved by discussion with a culated to indicate overall heterogeneity. Netrank p-scores were
third reviewer (TDV). generated in the NMA, which is an indicator for the ‘best’ treatment
The TRIP database, clinicaltrials.gov, Cochrane central register by ranking the treatments on a continuous scale from 0 to 1. By
of trials, current controlled trials metaRegister and the Australian taking into account treatment superiority, certainty, homogeneity
New Zealand Clinical trials registry were searched to identify ongo- and consistency, p-scores measure the mean extent of certainty that
ing or unpublished randomized controlled trials (RCTs). Reference a treatment is better than the competing treatment.
lists of included studies were pearled to identify relevant articles
that may have been missed by the searches.6
Subgroup analysis and sensitivity analysis
This study planned to perform subgroup analysis and search for
Data extraction and management heterogeneity for the primary outcomes in the following subgroups:
Data extraction was completed by one author (SN) and checked for • VTE risk (high risk versus low risk based on: age, mobiliza-
accuracy by a second reviewer (BH). Discrepancies settled via con- tion, history of VTE, revision procedure)
sensus. The following information was extracted from the articles: • Treatment duration (up to 10 days versus 10–14 days versus
15 days or more)
Table 1 PICO and study selection criteria • Dosage (depending on the dosages from the included studies)
• Type of surgery (knee versus hip arthroplasty)
Population Patients who have undergone elective total hip or
knee arthroplasty
This study also planned to conduct sensitivity analysis to identify
Intervention Two commonly used drugs in Australian clinical influential studies and factors that affect the direction and the size
practice: of the reported effects.
• Low-molecular-weight heparin (i.e. enoxaparin,
traded as clexane)
• Antiplatelet (i.e. aspirin) Publication bias
Comparator Effectiveness and safety outcomes of both aspirin
and enoxaparin will be compared with each Publication bias was planned to be investigated using funnel plot
other, placebo and no treatment asymmetry testing for meta-analysis with greater than 10 studies.7
Outcomes Safety outcomes: bleeding events, wound In addition, clinical trial databases were searched to identify can-
infections or other adverse events
Primary effectiveness outcomes: mortality, rates of celled or unpublished clinical trials.
symptomatic DVT or PE
Secondary effectiveness outcomes: radiological
evidence of DVT or PE Risk of bias and level of evidence (quality
Design RCTs appraisal)
Publication date All RCTs published in the last 20 years, from 1998
and type† to 2018 Two reviewers (SN and BH) independently assessed the risk of bias
Grey literature resources for unpublished RCTs using the Cochrane ‘Risk of bias’ tool for RCTs.8 Any discrepan-
All languages
cies were discussed with a third reviewer (TDV).
†Literature searches for EMBASE and PubMed databases were performed The strength of the body of evidence was assessed per outcome
on 20 February 2018 and the search in the Cochrane database was per-
formed on 21 February 2018. DVT, deep vein thrombosis; PE, pulmonary using the GRADE approach.9 Scores for the five domains in the
embolism; PICO, Population, Intervention, Comparator, Outcome; RCT, GRADE approach for direct and indirect evidence were combined
randomized controlled trial.
to get a GRADE score to indicate the strength of the body of
evidence for the NMA. Scoring was performed in the online GRA- Quantitative analysis
DEpro tool.10 Sufficient data were only available for total DVT rates (combining
both symptomatic and asymptomatic DVT rates), symptomatic PE
rates, major haemorrhage and wound complication; hence the
NMA was performed on the four outcomes. All the comparisons
Results investigated by the NMA are based on data combining THA
Study selection and TKA.
The flowchart of the search strategy is shown in Figure 1. After the
full-text assessment of all citations, 13 RCTs were excluded and Total DVT rates
only nine RCTs were eligible for inclusion in the quantitative anal- There was no statistical significance in total DVT rates between
ysis. No additional references were identified through pearling the aspirin and enoxaparin (RR = 1.27 (0.84, 1.96)) reported within the
reference lists of the included studies. direct evidence (Table 3). Also, no statistical significance was
observed for indirect evidence (RR = 1.09 (0.56, 2.12)) or the
NMA, incorporating direct and indirect evidence (RR = 1.21 (0.86,
1.72)) on total DVT. The I2 is zero (P-value = 0.48) indicating
Study characteristics reported outcomes are very consistent (Table 3). For total DVT
The nine studies included a total of 2336 patients receiving aspi- rates, the Netrank p-score was higher for aspirin (p-score = 0.94)
rin, 897 patients receiving enoxaparin and 2625 patients receiving than enoxaparin (p-score = 0.55). The higher the p-score, the more
placebo (Table 2). Two studies were direct comparisons,11,12 that likely the treatment is more reliable overall regarding superiority,
investigated the effect of enoxaparin compared to aspirin. The consistency and certainty.
other seven studies compared enoxaparin with placebo13–17 or
aspirin with placebo.18,19 Although fewer trials investigated aspi-
rin the sample size investigating aspirin is significantly larger. Symptomatic PE rates
From the aspirin trials, only two studies reported the use of enteric For direct evidence RRs and 95% CIs could not be calculated due to
coating.12,19 All studies excluded high-risk patients and therefore zero events. For indirect evidence RR and 95% CI could not be cal-
all the evidence is based on low risk patient groups. High-risk culated as no trial comparing aspirin with placebo reported symptom-
patients were mostly excluded due to history of VTE, history of atic PE rates. The NMA did not find a significant difference between
haemorrhagic stroke or gastro-intestinal bleeding, severe periph- aspirin compared to enoxaparin (RR = 1.02 (0.00, 242.90)). The evi-
eral vascular diseases; chronic heart failure or severe varicose dence base for aspirin in the NMA is based on one study,11 reporting
veins. zero events. The wide range of the CI implies the point estimate is
very uncertain. There was moderate heterogeneity for symptomatic
PE rates (I2 = 58.6, P-value = 0.06). Assessment of heterogeneity
‘between designs’ was not applicable because the outcome was not
Total studies identified sufficiently reported in the direct evidence. The Netrank p-score was
Identification
(N=5158)
0.68 for aspirin and 0.44 for enoxaparin.
EMBASE (n=3267)
PubMed (n=1199) Total studies excluded
Cochrane (n=692) (N = 1205)
Wound complication
Duplicates
(n = 708) Wound complication was only reported in one direct RCT
Studies prior 1998 (RR = 0.68 (0.12, 3.98)),11 and the indirect evidence (RR = 0.66
(n=497)
Screening
Total studies screened by title (0.05, 8.86)). Neither of the RRs were statistically significant. The
and abstract NMA, incorporating direct and indirect evidence, non-significant
(N= 3953) Total studies excluded
(N = 3931) differences were observed between aspirin compared to enoxaparin
Non RCT ( n= 43) (RR = 0.73 (0.17, 3.20)). Heterogeneity was low (I2 = 19.6%, P-
Cost analysis ( n= 104) value = 0.29). The highest Netrank p-score was found for aspirin
Reviews ( n=271)
Guidelines ( n= 24) (p-score = 0.79) compared to enoxaparin (p-score = 0.46).
Eligibility
Total studies included in reported (0.09, 50.99)). The wide range of the 95% CI indicates the point
quantitative analysis
(N= 9)
estimate is highly uncertain.
There was no evidence of heterogeneity (I2 = 0%, P-value =
0.6). No differences were found between the Netrank p-scores for
Fig. 1. PRISMA study selection flowchart. aspirin (p-score = 0.51) and enoxaparin (p-score = 0.50).
Author, year Total Type of Age mean (range) BMI mean (range) Daily dose in mg Follow Funding and Conflict of
Country no. patients procedure up, support interest†
randomized weeks
Intervention Comparator Intervention Comparator Intervention Comparator
†Conflict of interest is based on the reported acknowledgement of conflict of interest reported and reported funding and support. ‡Fuji et al. investigated THA and TKA and reported results separately.
NR, not reported; SD, standard deviation; THA, total hip arthroplasty; TKA, total knee arthroplasty; UK, United Kingdom.
Table 3 Inconsistency for total, within design and between design comparisons for effectiveness and safety outcomes
Total I † = 0%
2 2
I = 58.6% 2
I = 19.6% I2 = 0%
Q = 6.56(7) Q = 7.25(3) Q = 6.22(5) Q = 3.69(5)
P-value = 0.48 P-value = 0.06 P-value = 0.29 P-value = 0.60
Within design Q = 6.38(6) Q = 7.25(3) Q = 6.16(4) Q = 3.69(5)
P-value = 0.38 P-value = 0.06 P-value = 0.19 P-value = 0.60
Between design Q = 0.18(1) Q = 0.00(0) Q = 0.05(1) Q = 0.00(0)
P-value = 0.67 P-value = NA P-value = 0.82 P-value = NA
†I2 indicates high (75%), moderate (50%) and low (25%) heterogeneity. DVT, deep vein thrombosis; NA, not applicable; PE, pulmonary embolism; Q, degrees of
freedom.
Publication bias
Because of the low number of studies that are incorporated in the
NMA it was inappropriate to perform funnel plot asymmetry testing
to investigate publication bias. Two trials identified on clinical trial
databases were noted as completed; however, no publication was
found indicating possible publication bias. Details for the trials are
shown in Appendix I.
Discussion
The NMA did not find statistically significant differences for the
four outcomes comparing aspirin with enoxaparin.
Outcome Direct evidence (two studies) Indirect evidence‡ (seven studies) Network meta-analysis (nine studies)
Relative risk ratio GRADE scores§ Relative risk ratio GRADE scores§ Relative risk ratio GRADE scores§
(95% CI) (95% CI) (95% CI)
†Only one study was available; event rates were 0 so a RR could not be calculated. ‡Risk ratios and the quality of evidence for indirect evidence were generated
by combining enoxaparin versus placebo and aspirin versus placebo using R statistical package §GRADE scores are based on evidence graded according to the
GRADE approach including the following components: risk of bias, inconsistency, indirectness and imprecision. The GRADE scores for these components are
obtained separately for all three comparisons investigated in this study. For grading evidence of the NMA scores for those five components are combined accord-
ing to the GRADE approach to obtain a new GRADE score. ¶The network meta-analysis included one direct study of clexane versus aspirin, and one indirect study
of clexane versus placebo ††One study comparing clexane to placebo was identified; an indirect RR could not be calculated. CI, confidence interval; DVT, deep
vein thrombosis; NA, not applicable; PE, pulmonary embolism. RR values <1 indicate the size of a protective effect of the treatment described compared to enoxa-
parin. RR values >1 indicate the size of a certain risk of the treatment described compared to enoxaparin.
study. The exclusion of other drugs also minimizes the risk of this statement; the largest trial included in this review (n = 4088)
including potentially heterogeneous populations from other trials. comparing aspirin with placebo was not able to detect sufficient PE
Even though Kapoor et al.20 included more comparative arms events.19
and used all low-molecular-weight heparin as the active compara- Due to the limited data available, the planned subgroup analyses
tor, which made them able to expand their evidence base, they were could not be performed. Future studies investigating the use of aspi-
not able to draw firm conclusions about aspirin. This corresponds rin without an enteric coating would also be valuable, as only two
with our findings, which showed that the evidence base around aspirin trials reported the use of enteric coating,12,19 which reduces
aspirin as a VTE prophylaxis is limited. the bio availability of aspirin.26
We encountered the same restrictions for type of procedure, drug
dose, treatment duration, ethnicity, conflict of interest and the use
Strengths and limitations mechanical prophylaxis.
This review did not find sufficient data for symptomatic DVT rates It would be clinically valuable to compare the effects between
and was forced to investigate total DVT rates (including both high and low risk populations. However, high-risk populations were
symptomatic and asymptomatic DVT rates). Debate on the clinical systematically excluded from all current RCTs, meaning the results
relevance of asymptomatic DVT is ongoing,21 and disagreement cannot be extrapolated to patients at high-risk for VTE or complica-
tions due to prophylactic treatments. While this review only looked
exists on the relationship between asymptomatic DVT and clini-
at RCTs, future reviews should consider the inclusion of other levels
cally important VTE. This is the main issue leading to conflicting
of evidence to identify data from patients at greater risk of VTE. Risk
recommendations on the use of aspirin.22 Struijk-Mulder et al.
levels for VTE and postoperative complications, based on patient-
stated that disagreement on the relevance of different endpoint
specific comorbidities, will likely affect clinical decisions as to which
(e.g. asymptomatic DVT) should be solved before agreement
drug to use in specific risk groups. The individual complexity of
around aspirin as a VTE prophylaxis can be reached. This review
weighing the best approach for managing VTE depends on each
was not able to distinguish between both outcomes.
patient, which makes this area of research open for interpretation.
Although this review investigated both effectiveness (e.g. DVT)
Although the data are limited, this review does show a lack of a
and safety (e.g. major haemorrhage) outcomes, it does not place a
statistical significant difference between aspirin and enoxaparin for
greater emphasis on one or the other when comparing aspirin with
VTE prophylaxis after THA and TKA. This is an important finding
enoxaparin. Current guidelines, which consider both effectiveness
as it directly contributes to the current discussion around whether to
and safety outcomes to be important, have different views on the
or not to use aspirin as a prophylactic agent. Also, this review
weight of importance when looking at effectiveness and safety.22
offers insights in the possible reasons for not finding statistical sig-
This review did not find any evidence for the effectiveness and
nificance, which is necessary in helping future studies to obtain bet-
safety of aspirin in the prevention of PE. Interestingly, evidence
ter and more convincing evidence.
around PE was also absent for enoxaparin. This can be explained
by the rare occurrence of PE; according to two register-based
studies,23,24 the 90-day non-fatal PE rate after THA and TKA with-
out prophylaxis is 0.93% and 0.41% respectively. According to the
Conclusion
American Academy of Orthopaedic Surgeons (AAOS) rates of PE This review did not find statistical significant results when compar-
cannot be assessed adequately.25 Results from this review support ing aspirin with enoxaparin as a VTE prophylaxis. No evidence
was identified for PE and subgroup analysis was not possible due disease prophylaxis in total knee arthroplasty. J. Arthroplasty 2006; 21:
to limited data. Future studies should identify more evidence, par- 139–43.
ticularly for PE and high-risk groups. This might help decision- 13. Chin PL, Amin MS, Yang KY, Yeo SJ, Lo NN. Thromboembolic pro-
makers to get consensus on the use of aspirin as VTE prophylaxis. phylaxis for total knee arthroplasty in Asian patients: a randomised con-
trolled trial. J. Orthop. Surg. (Hong Kong) 2009; 17: 1–5.
14. Fuji T, Ochi T, Niwa S, Fujita S. Prevention of postoperative venous
thromboembolism in Japanese patients undergoing total hip or knee arthro-
Acknowledgements
plasty: two randomized, double-blind, placebo-controlled studies with
Funding support for this project was provided by the Royal Austral- three dosage regimens of enoxaparin. J. Orthop. Sci. 2008; 13: 442–51.
asian College of Surgeons. Sinan Nadi was supported by a research 15. Intiyanaravut T, Thongpulsawasdi N, Sinthuvanich N, Teavirat S,
scholarship by the Royal Australasian College of Surgeons. The Kunopart M. Enoxaparin versus no anticoagulation prophylaxis after
total knee arthroplasty in Thai patients: a randomized controlled trial.
authors are grateful to Dr Alun Cameron and Dr David Tivey for
J. Med. Assoc. Thai. 2017; 100: 42–9.
their assistance during the preparation of this manuscript. Also, the
16. Jain V, Dhal AK, Dhaon BK, Pradhan G. Deep vein thrombosis after
authors offer their special thanks to Bridget Heijkoop, who has
total hip arthroplasty in Indian patients with and without enoxaparin.
been the second reviewer in this manuscript. J. Orthop. Surg. (Hong Kong) 2004; 12: 173–7.
17. Kim SM, Moon YW, Lim SJ, Kim DW, Park YS. Effect of oral factor
Xa inhibitor and low-molecular-weight heparin on surgical complica-
Conflicts of interest tions following total hip arthroplasty. Thromb. Haemost. 2016;
115: 600–7.
None declared.
18. Kim YH, Choi IY, Park MR, Park TS, Cho JL. Prophylaxis for deep
vein thrombosis with aspirin or low molecular weight dextran in Korean
patients undergoing total hip replacement. A randomized controlled
References trial. Int. Orthop. 1998; 22: 6–10.
1. White RH, Henderson MC. Risk factors for venous thromboembolism 19. Rodgers A, MacMahon S, Collins R, Prentice C. Prevention of pulmo-
after total hip and knee replacement surgery. Curr. Opin. Pulm. Med. nary embolism and deep vein thrombosis with low dose aspirin: pulmo-
2002; 8: 365–71. nary embolism prevention (PEP) trial. Lancet 2000; 355: 1295–302.
2. Arthroplasty Society of Australia. Arthroplasty Society of Australia 20. Kapoor A, Ellis A, Shaffer N et al. Comparative effectiveness of venous
Guidelines for VTE Prophylaxis for Hip and Knee Arthroplasty. 2017. thromboembolism prophylaxis options for the patient undergoing total
[Cited 20 Jan 2018.] Available from URL: https://www.aoa.org.au/ hip and knee replacement: a network meta-analysis. J. Thromb. Haemost.
3. Ricket AL, Stewart DW, Wood RC et al. Comparison of postoperative 2017; 15: 284–94.
bleeding in total hip and knee arthroplasty patients receiving rivaroxa- 21. Quinlan DJ, Eikelboom JW, Dahl OE, Eriksson BI, Sidhu PS, Hirsh J.
ban or enoxaparin. Ann. Pharmacother. 2016; 50: 270–5. Association between asymptomatic deep vein thrombosis detected by
4. Guijarro R, Montes J, San Roman C et al. Venous thromboembolism venography and symptomatic venous thromboembolism in patients under-
and bleeding after total knee and hip arthroplasty. Findings from the going elective hip or knee surgery. J. Thromb. Haemost. 2007; 5: 1438–43.
Spanish National Discharge Database. Thromb. Haemost. 2011; 22. Struijk-Mulder MC, Ettema HB, Verheyen CC, Buller HR. Comparing
105: 610–5. consensus guidelines on thromboprophylaxis in orthopedic surgery.
5. National Institute of Clinical Excellence. Venous Thromboembolism: J. Thromb. Haemost. 2010; 8: 678–83.
Reducing the Risk for Patients in Hospital. 2015. [Cited 22 Jan 2018.] 23. Katz JN, Losina E, Barrett J et al. Association between hospital and sur-
Available from URL: https://www.nice.org.uk/Guidance/CG92 geon procedure volume and outcomes of total hip replacement in the United
6. Cooke A, Smith D, Booth A. Beyond PICO: the SPIDER tool for quali- States medicare population. J. Bone Joint Surg. Am. 2001; 83-a: 1622–9.
tative evidence synthesis. Qual. Health Res. 2012; 22: 1435–43. 24. SooHoo NF, Lieberman JR, Ko CY, Zingmond DS. Factors predicting
7. Sterne J, Becker B, Egger M. The funnel plot. In: Publication Bias in complication rates following total knee replacement. J. Bone Joint Surg.
Meta-Analysis: Prevention, Assessment and Adjustments. Chichester, Am. 2006; 88: 480–5.
UK: John Wiley & Sons, 2005. 25. American Academy of Orthopedic Surgeons Clinical Guideline on Pre-
8. Higgins J, Sterne J, Savovic J et al. A revised tool for assessing risk of vention of Symptomatic Pulmonary Embolism in Patients Undergoing
bias in randomized trials. Cochrane Database Syst. Rev. 2016; 10 Total Hip or Knee Arthroplasty. 2007. [Cited 22 Jan 2018.] Available
(Suppl 1): 29–31. from URL: http://www.aaos.org/Research/guidelines/PE_guideline.pdf
9. Balshem H, Helfand M, Schunemann HJ et al. GRADE guidelines: 26. James MJ, Foreman RK, Burnett JR, Cleland LG. Inhibition of human
3. Rating the quality of evidence. J. Clin. Epidemiol. 2011; 64: 401–6. endothelial prostacyclin synthesis by different aspirin formulations.
10. GRADEpro GDT: GRADEpro Guideline Development Tool [Soft- Aust. N. Z. J. Surg. 1991; 61: 849–52.
ware]. McMaster University, 2015 (developed by Evidence Prime,
Inc.). Available from gradepro.org
11. Zou Y, Tian S, Wang Y, Sun K. Administering aspirin, rivaroxaban and Supporting information
low-molecular-weight heparin to prevent deep venous thrombosis after Additional Supporting Information may be found in the online ver-
total knee arthroplasty. Blood Coagul. Fibrinolysis 2014; 25: 660–4.
sion of this article at the publisher’s web-site:
12. Westrich GH, Bottner F, Windsor RE, Laskin RS, Haas SB, Sculco TP.
VenaFlow plus Lovenox vs VenaFlow plus aspirin for thromboembolic Appendix S1. Supporting Information.