Thrombosis Research 148 (2016) 6369

Contents lists available at ScienceDirect

Thrombosis Research

journal homepage: www.elsevier.com/locate/thromres

Review Article

A systematic review of biomarkers for the prediction of

thromboembolism in lung cancer Results, practical issues and
proposed strategies for future risk prediction models
Marliese Alexander, B.Pharm(Hons), Grad Cert Pharm Prac, MPH a,b,,
Kate Burbury, MBBS(Hons) FRACP FRCPA DPhil c
a
Department of Epidemiology and Preventative Medicine Monash University, Commercial Road Melbourne Australia
b
Departments of Pharmacy and Haematology Peter MacCallum Cancer Centre, Grattan Street Melbourne Australia
c
Department of Haematology Peter MacCallum Cancer Centre, Grattan Street Melbourne Australia

a r t i c l e i n f o a b s t r a c t

Article history: Introduction: This review aimed to identify candidate biomarkers for the prediction of thromboembolism (TE) in
Received 30 July 2016 lung cancer.
Received in revised form 13 October 2016 Materials and methods: Systematic review of publications indexed in PubMed or EMBASE databases in the past
Accepted 24 October 2016 5 years (01/05/201101/05/2016) which evaluated baseline and/or longitudinal biomarker measurements as a
Available online 26 October 2016
predictor of subsequent TE (venous and arterial) in lung cancer patients.
Results: Of 1105 studies identied, 18 fullled predened inclusion criteria: 6 prospective and 12 retrospective.
Keywords:
Thromboembolism
The 18 studies included 11,262 patients and 36 unique biomarkers. The combined TE rate was 7% (741/
Thrombosis 10,854), increasing to 11% (294/2612) within prospective studies. All biomarker measurements were baseline
Biomarker only, with no longitudinal assessment reported. The most frequently investigated biomarkers were tumour-re-
Lung cancer lated driver mutations, D-dimer, haemoglobin, white cell, and platelet count; as well as biomarker combinations
Risk prediction previously used in risk prediction models, such as Khorana risk score. Biomarker thresholds rather than contin-
uous variable analyses were generally applied, however thresholds were not consistent across studies. D-dimer
and epidermal growth factor receptor mutation were the strongest and most reproducible predictors of TE.
Conclusion: An important limitation is the lack of prospective data across specic subpopulations of cancer, with
correlative, and preferably longitudinal, biomarker assessments. This would provide insight into the pathophys-
iology, allow patient proling, and the development of personalised decision-making tools that can be used real-
time and throughout the course of the patients' journey, for targeted, risk-adaptive preventative strategies.
2016 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.1. Search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.2. Inclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.3. Data collection and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.1. Quality of included studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2. Biomarkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3. Khorana risk score and included parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4. D-dimer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.5. Driver mutations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Corresponding author.
E-mail address: marliese.alexander@petermac.org (M. Alexander).

http://dx.doi.org/10.1016/j.thromres.2016.10.020
0049-3848/ 2016 Elsevier Ltd. All rights reserved.
64 M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369
Declarations of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
1. Introduction
As a commonly diagnosed and well established thrombogenic can-
cer, lung cancer is a major contributor to cancer associated thromboem-
bolism (TE) [1,2]. TE is a major complication and an important
contributor to morbidity and mortality, however the risk is dynamic
and heterogeneous. Various thrombogenic biomarkers have been
utilised to predict risk in various populations, with discordant success.
The pathophysiology is complex and involves the interaction be-
tween the tumour (oncogenes and proteins), intracellular signalling
pathways, coagulation system and the anticancer treatment [35]. As
a consequence, the overall TE risk is heterogeneous even within lung
cancer populations and moreover, is dynamic over the individual
patient's journey. As such, a generic broad application of
thromboprophylaxis particularly given the potential concomitant
bleeding risks, is not appropriate, as reected by low utilisation of pre-
ventative therapy outside of hospitalisation and/or surgery [6,7].
Clinicians are looking for simple, practical and relevant methods to
risk stratify patients, real time in the clinic and enable a personalised
targeted approach to TP. Despite the existence of validated risk assess-
ment models for the prediction of cancer associated TE, [8,9] they
have not resulted in decision-making algorithms for prophylactic ther-
apy among patient groups. A major limitation has been the substantial
population heterogeneity within the derivation and validation cohorts,
with TE rates across included cancer populations ranging from b1% to
40%, [10,11] and resultant poor sensitivity and potency of biomarkers.
Moreover studies have not measured longitudinal change over time,
or considered competing factors, which results in an overall underesti-
mation of the true effect for high TE risk patients during high risk
periods.
The conventional Khorana risk score for the prediction of TE among
cancer patients receiving ambulatory chemotherapy considers site of
cancer, body mass index ( 35 kg/m2), haemoglobin level (b100 g/L),
platelet count (350 109) and white cell count (11 109) [8]. The
model is simple and contains parameters that can be measured real
time and in routine diagnostic laboratories. However, these clinical
and non-specic laboratory markers achieve only modest sensitivity
and specicity, which is paramount in risk-prediction tools. Adaptations
to improve performance have included additional biomarkers p-selectin
and D-dimer [9], and weighting for patients receiving chemotherapy
regimens associated with the highest rates of TE (cisplatin, carboplatin
or gemcitabine) [12]. Within these models, lung cancer with no addi-
tional risk factors, already achieves an intermediate TE risk score. More-
over, the cited thrombogenic chemotherapy agents are those
commonly used for the treatment of lung cancer, and biomarkers such
as D-dimer are regularly elevated in patients with lung cancer [13
15]. As such, the current models lack stratication power, and would
suggest that all lung cancer patients warrant consideration of
thromboprophylaxis, for the entirety of their treatment. The
FRAGMATIC study assessed this approach in a multi-centre, randomised
controlled trial comparing standard treatment plus low molecular
weight heparin (LMWH) versus standard treatment alone in N2000 pa-
tients with newly diagnosed lung cancer [16]. Non-targeted LMWH pro-
phylaxis reduced the risk of TE by 40% in the entire cohort (hazard ratio
0.57, p = 0.001). A more targeted strategy would have potentially re-
sulted in a greater risk reduction in patients identied at intermediate
to high TE risk, while avoiding unnecessary intervention in those with
the lowest risk. Importantly, supportive treatments should focus on out-
comes such as improved morbidity, quality of life, decreased health
resource utilisation, rather than lack of survival benet. But this result
(hazard ratio for 1 year survival 1.01, p = 0.814) in the FRAGMATIC
study has likely contributed to lack of global uptake for primary TE pro-
phylaxis in lung cancer patients [16].
This systematic review was undertaken to identify candidate bio-
markers in patients with lung cancer and propose future strategies for
the development of a dynamic TE risk prediction tool, which contributes
to appropriate real-time decision-making algorithms.
2. Methods
2.1. Search strategy
Papers indexed in PubMed (including MEDLINE via NCBI) and
EMBASE (via OVID) were systematically searched for the most recent
5-year period (01/05/201101/05/2016). Reference lists of retrieved ar-
ticles were reviewed for additional citations. Broadly, the search strate-
gy combined the following key search terms: thromboembolism/deep
vein thrombosis/pulmonary embolism, AND biomarker/risk factor,
AND lung cancer; full search strategy available as Supplementary
material.
2.2. Inclusion criteria
Studies included in the analysis were required to report i) data anal-
yses on a dened lung cancer population as entire cohort or stratied
subset of mixed cohort; ii) at least one biomarker either as single base-
line or longitudinal measurements, as a predictor of risk for TE; iii) a
measure of association (or ability to calculate) between biomarker
and TE. Patients could receive any or no anti-cancer treatment and
any or no thromboprophylaxis. TE was dened as any venous thrombo-
embolism (VTE) or arterial thromboembolism (ATE) including but not
limited to deep vein thrombosis (DVT), pulmonary embolism (PE), ce-
rebral vascular accident (CVA) or acute myocardial infarction (AMI).
2.3. Data collection and analysis
Two independent authors assessed study inclusion and quality. Data
was extracted by author one and then reviewed and validated by author
two. Principal summary measures extracted were biomarker levels, TE
rates, and TE risk (hazard ratio (HR), odds ratio (OR)). Where appropri-
ate, data was pooled using a random effects model in Review Manager
5.3 software [17]. Risk of bias assessments was conducted using the
Newcastle-Ottawa Scale (NOS) [18], a validated tool for non-
randomised studies [19]. Within the NOS a total of 4 points can be allo-
cated for selection methods, 2 points for comparability of cohorts or
cases and controls, and 3 points for outcomes or exposures [18].
3. Results
The search strategy identied 1105 studies of which 18 (6 prospec-
tive and 12 retrospective) fullled predened inclusion criteria and
were included in the nal review, Fig. 1. The 18 studies included
11,262 patients and 36 unique biomarkers. All studies investigated ve-
nous events (deep vein thrombosis (DVT) and/or pulmonary embolism
(PE)), with no study including arterial thrombotic events. Anticancer
treatments varied: chemotherapy (5 studies); surgery (2 studies); any
chemotherapy, radiotherapy, surgery or biologic therapy (5 studies);
and treatment not specied (6 studies).
 M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369 65

Identification
1103 studies identified through database searching
+ 2 additional studies identified from reference list

Screening
839 excluded-review paper, non-lung
217 duplicates removed and
cancer population, no biomarkers, no
888 titles and abstracts screened
measure of association with TE, population
with existing TE, otherwise irrelevant
Eligibility

49 full papers assessed for eligibility 31 excluded-duplicate data (2), population

with existing TE (5), no biomarker evaluated
(24)
Included

18 included in final review 18 includedprospective (6), retrospective

(12)

Fig. 1. Study ow diagram.

Of the included studies, 14 were cohort (6 prospective/8 retrospec- factor receptor mutation (EGFR) positive) [24]. Delmonte et al. included
tive) [2033] and 4 were case-control studies [3437]. The combined 6 patients with anaplastic lymphoma kinase (ALK) transformations;
TE event rate was 7% (741/10,854), [2033] increasing to 11% (294/ higher EGFR (n = 49) and KRAS (n = 50) mutation positive [25].
2612) considering only prospective studies [2023].

3.1. Quality of included studies
The quality of studies varied with median Newcastle Ottawa Score of
6/9 (range 19), Table 1. Detraction of points was most commonly a re-
sult of lack of reporting detail, short follow-up (Lee et al. required only
4-week follow-up) [29], unspecied follow-up [33], unexplained loss
to follow-up (up to 26%) [32], or failure to control for confounding fac-
tors and selection bias. While implied, many studies did not explicitly
exclude patients with TE at baseline. Berger et al.'s intent was to mea-
sure the association between tumour driver mutations and TE, however
only included 7 cases with the biomarker of interest (epidermal growth
3.2. Biomarkers
The most frequently investigated biomarkers were D-dimer
(6 studies), EGFR mutation (5 studies), platelet count (5 studies),
haemoglobin (4 studies), white cell count (4 studies), and the bio-
marker combination of haemoglobin, white cell count and platelet
count within the Khorana prediction model (3 studies). All bio-
markers were measured at baseline only, with no longitudinal mea-
surements reported. Varied thresholds were applied apart from
when the pre-dened Khorana risk score parameters were utilised.
Findings are summarised in Table 2 with further details on the
most frequently investigated biomarkers below.

Table 1
Quality of included studies using Newcastle Ottawa Scale for non-randomised studies [18].

Retrospective (R) or prospective (P) Follow-up Selection Comparability Outcome Score out of 9

Cohort studies
Ferroni [26] R 7 monthsa 9
Zhang [32] R 47 monthsa 8
Yang [28] R 26 monthsa 8
Lee [30] R 4 weeksb 7
Zhang [23] P NR 7
Kadlec [22] P 6 months 7
Noble [20]c P 1 year 6
Lee [29] R 2 years 6
Yamazaki [31]c R NR 5
Delmonte [25]c R NR 4
Gaspar [27]c R 3 months 4
Zhu [33]3 R NR 4
Berger [24]3 R NR 3
Evmoradis [21]c P 3 months 1

Case-control studies
Wang [35] R NA 8
Corrales-Rodriguez [36] R NA 8
Arslan [34] P NA 7
Davies [37] P NA 6
Overall median 6
a
Median follow-up.
b
Patients with at least 4 weeks follow-up included.
c
Abstract only.
 66
Table 2
Summary ndings from included studies (listed by biomarker and sample size).

Cancer Sample TE events

M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369

Biomarker Author, year Lung cancer treatment TE type size No. (%) Comparator Outcome Summary nding

Blood biomarker
Khorana risk score Noble 2015 NSCLC/SCLC Chemo VTE 910 108 1 vs. 2 vs. 3 vs. 4 TE rate Score 1: 13%, score 2: 11%, score 3: 12%, score 4: 14%; p N 0.05
(12%)
Lee 2015 SCLC Chemo VTE 277 30 (11%) 2 vs. 1 TE risk (HR) 0.3 (95% CI 0.11.3)
3 vs. 1 1.0 (95% CI: 0.17.1)
3 vs. b3 1.0 (95% CI: 0.17.7)
Gaspar 2012 NR Chemo VTE 241 20 (8%) TE (+) vs. () % pts. score 3 35% vs. 14%, p = 0.02
Platelet count Zhu 2012 NR NR VTE 2053 89 (4%) NULN vs. ULN TE rate 6.3% vs. 3.7%, p b 0.05
Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) N299.5 109/L TE risk (OR) 3.9 (95% CI: 2.46.4)
Zhang 2014 NSCLC Any VTE 673 89 (13%) N350 109/L TE risk (OR) NR (p N 0.05)2
Lee 2015 SCLC Chemo VTE 277 30 (11%) 350 vs. TE risk (HR) 0.5 (95% CI: 0.11.9)
b350 109/L
Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. NULN 26% vs. 31%, p = 0.49
Haemoglobin Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) b123.5 vs. 123.5 g/L TE risk (OR) 1.0 (95% CI: 1.01.0)
Zhang 2014 NSCLC Any1 VTE 673 89 (13%) 100 vs. 100 g/L TE risk (OR) 4.6 (95% CI: 1.414.5)
Lee 2015 SCLC Chemo VTE 277 30 (11%) b100 vs. 100 g/L TE risk (HR) NR (no events)2
Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. b100 g/L 38% vs.19%, p = 0.01
White cell count Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) b6.4 vs. 6.4 109/L TE risk (OR) 1.9 (95% CI: 1.33.0)
Zhang 2014 NSCLC Any VTE 673 89 (13%) 10 vs. b10 109/L TE risk (OR) 2.8 (95% CI: 1.55.4)
Lee 2015 SCLC Chemo VTE 277 30 (11%) N11 vs. b11 109/L TE risk (HR) 0.7 (95% CI: 0.22.7)
Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. NULN 49% vs. 19%, p b 0.01
D-dimer Zhu 2012 NR NR VTE 2053 89 (4%) NULN vs. ULN TE rate 19.9% vs. 0.3%, p b 0.05
Yang 2012 NSCLC Surgery VTE 1001 53 (5%) N0.3 vs. 0.3 mg/L TE risk (HR) 7.5 (95% CI: 4.014.3)
TE rate 65.4% vs. 3.7%, p b 0.01
Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) N0.4 vs. 0.4 mg/L TE risk (OR) 1.7 (95% CI: 1.02.7)
Zhang 2013 NSCLC Surgery VTE 232 17 (7%) TE (+) vs. () Continuous 0.35 vs. 0.20 mcg/mL, p = 0.01
Ferroni, 2012 NSCLC/SCLC Chemo VTE 108 16 (15%) TE (+) vs. () Continuous TE risk (HR) 0.35 vs. 0.11 mcg/mL, p b 0.01
N1.5 vs. 1.5 mg/L 11.0 (95% CI: 2.646.2)
Evmoradis 2014 NSCLC Chemo VTE 79 6 (8%) TE (+) vs. () Continuous Increased with TE2
Fibrinogen Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) N3.2 vs. 3.2 g/L TE risk (OR) 1.5 (95% CI: 1.02.4)
ROTEM Davies 2015 NSCLC/SCLC NR VTE 67 6 (9%) TE (+) vs. () Continuous No difference2
Factor VIIa Evmoradis 2014 NSCLC Chemo VTE 79 6 (8%) TE (+) vs. () Continuous NR
Thrombin generation Increased with TE2
Tissue factor activity NR
P-selectin NR
PPL Decreased with TE2
Heparanase Decreased with TE2
Prothrombin time Kadlec 2014 NSCLC/SCLC Any1 VTE 950 91 (10%) b1.1 vs. 1.1 s TE risk (OR) 1.0 (95% CI: 0.71.7)
aPPT N36.0 vs. 36.0 s 1.6 (95% CI: 1.02.5)
 Albumin Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. NULN 61% vs. 39%, p b 0.01
Sodium 13% vs. 3%, p = 0.02
ALT 26% vs. 10%, p b 0.01
AST 21% vs. 13%, p = 0.15
GGT 79% vs. 54%, p b 0.01
Interleukin-1 Zhu 2012 NR NR VTE 2053 89 (4%) NULN vs. ULN TE rate 2.4% vs. 10.3%, p b 0.05

Tumour and genetic biomarkers

EGFR Lee 2014 NSCLC Any1 VTE 1998 131 (7%) Mutation (+) vs. () TE risk (HR) 0.7, p = 0.2
Yamazaki 2013 NSCLC Any1 PE 1953 18 (1%) Mutation (+) vs. () TE rate 2.3% vs. 0.48%, p b 0.01
Delmonte 2015 NSCLC NR VTE 289 62 (21%) Mutation (+) vs. () TE rate No difference2
Corrales-Rodriguez NSCLC NR VTE 159 57 (36%) Mutation (+) vs. () TE risk (OR) 1.0, p = 1.0
2014
Berger 2014 NSCLC NR VTE 23 5 (22%) Mutation (+) vs. () TE rate 42% vs. 12%, p = 0.5
ALK Lee 2014 NSCLC Any1 VTE 1998 131 (7%) Mutation (+) vs. () TE risk (HR) 2.2, p = 0.2
Delmonte 2015 NSCLC NR VTE 289 62 (21%) Mutation (+) vs. () TE rate No difference2
KRAS Delmonte 2015 NSCLC NR VTE 289 62 (21%) Mutation (+) vs. () TE rate Decreased TE (+) vs. (), p = 0.032
Corrales-Rodriguez NSCLC NR VTE 159 57 (36%) Mutation (+) vs. () TE risk (OR) 2.7, p = 0.01
2014
BRAF Delmonte 2015 NSCLC NR VTE 289 62 (21%) Mutation (+) vs. () TE rate No difference2

M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369

CEA Zhang 2014 NSCLC Any1 VTE 673 89 (13%) Third vs. rst tertile T risk (OR) 2.2 (95% CI: 0.95.4)
Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. NULN 83% vs. 71%, p = 0.19
TNF Zhu 2012 NR NR VTE 2053 89 (4%) NULN vs. ULN TE rate 3.3% vs. 7.7%, p b 0.05
Factor V Leiden Arslan 2011 NR NR DVT 66 33 (50%) TE (+) vs. () % pts. mutation+ 21% vs. 12%, p = 0.3
Prothrombin G20210A 4% vs. 3%, p = 0.6
MTHFR C677T 61% vs. 39%, p = 0.9
MTHFR A1298C 76% vs. 52%, p = 0.04
ACE D 82% vs. 67%, p = 0.2
Glycoprotein IIA 6% vs. 3%, p = 0.6
PAI-I 4G/5G 100% vs. 91%, p = 0.8
Neuron-specic Wang 2015 NSCLC Any1 VTE 183 61 (33%) TE (+) vs. () % pts. NULN 63% vs. 57%, p = 0.54
enolase
Cytokeratin fragment 87% vs. 79%, p = 0.29

Abbreviations: ACE angiotensin converting enzyme; aPPT activated partial thromboplastin time; ALK anaplastic lymphoma kinase; ALT alanine transaminase; AST aspartate aminotransferase; CEA carcinoembryonic
antigen; Chemo chemotherapy; DVT deep vein thrombosis; EGFR epidermal growth factor receptor; GGT aspartate aminotransferase; HR hazard ratio; LLN lower limit normal; MTHFR methylenetetrahydrofolate
reductase; NR not reported; NSCLC non-small cell lung cancer; OR odds ratio; PAI-1 plasminogen activator inhibitor-1; PE pulmonary embolism; PPL procoagulant phospholipids; ROTEM rotational thromboelastometry;
SCLC small cell lung cancer; TE thromboembolism; TNF tumour necrosis factor; ULN upper limit normal; VTE venous thromboembolism.
1
Chemotherapy, radiotherapy, surgery or biologic therapy.
2
Numerical value/HR/RR/OR not reported.

67
68 M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369
3.3. Khorana risk score and included parameters
Three studies assessed the predictive capacity of the pre-dened
Khorana score in lung cancer. The largest study included Non-small
Cell Lung Cancer (NSCLC) patients from the placebo arm of the
FRAGMATIC study which was a randomised controlled trial of low mo-
lecular weight heparin vs. placebo. Within this cohort the TE incidence
was similar across all (Khorana Score-dened) risk groups (Table 2)
[20]. Two further retrospective studies also suggested a lack of TE risk
stratication when using risk score threshold of 3 vs. b 3 (Table 1)
[27,30]. Five studies, which investigated platelet count (using either
ULN or 350 as the threshold) as an individual predictor of TE, also dem-
onstrated discordant results. Two studies (n = 2053 and n = 950) dem-
onstrated increased TE risk with elevated platelets, [22,33] while the
three studies (n = 673, n = 277, n = 183) did not, [29,32,35] Table 1.
Only one of four studies investigating haemoglobin demonstrated in-
creased TE risk with haemoglobin b100 g/L [35]. Two of four studies in-
vestigating WCC demonstrated increased TE risk with elevated WCC
[23,35], one study found increased TE risk with decreased WCC [22],
while the nal study found no relationship with TE [30].
3.4. D-dimer
Six studies evaluated the association between baseline D-dimer and
the development of TE. Four studies applied different thresholds
(1.5 mcg/mL, 0.4 mcg/mL, 0.3 mcg/mL, ULN), [22,26,28,33] while two
measured continuous D-dimer (one reported actual values) [21,32]. De-
spite the varying thresholds, all studies consistently demonstrated in-
creased risk of TE with elevated baseline D-dimer. The association
between increasing TE risk with increments of D-dimer was not ad-
dressed in these studies.
3.5. Driver mutations
Limited data is reported for ALK, KRAS and BRAF mutations with no
clear association with TE yet demonstrated (Table 2). More studies eval-
uated EGFR, with consistent methods and reporting allowing pooled
analyses. Meta-analysis including 1982 patients with 134 TE events
from four studies [24,25,31,36], (one study excluded due to insufcient
published data), [29] demonstrated increased TE risk among patients
harbouring EGFR mutations; HR 1.9 (95% CI 1.23.2).
4. Discussion
The impact of lung-cancer associated TE is substantial. Yet despite
the availability of safe and highly effective (TE risk reduction up to
80% when applied appropriately) anti thrombotic agents, with vast clin-
ical experience, they are not in routine use in the ambulatory care set-
ting where the majority of lung cancer patients are managed. Our
ability to reduce TE burden is not due to ineffective therapy, but rather
our commitment to risk-stratify patients and then make decisions re-
garding appropriate primary thromboprophylaxis, real-time in the clin-
ic. The identication of relevant predictive markers, for incorporation
into simple risk models with directed decision-making algorithms, can
be important enablers for effective targeted management strategies.
This systematic literature review highlighted this area of unmet need
and major limitations to date, particularly with regard to the predictive
power of various biomarkers. While many studies have been undertak-
en, there is a paucity of well-designed prospective lung cancer specic
studies with correlative, and preferably longitudinal, biomarker
assessments.
The burden of TE across studies varied, attributable to study design,
differences in patient-, disease-, and treatment-associated risk factors.
Most of the studies were retrospective and single centre. However the
combined rate of 11% in the prospective studies is substantial and likely
to be underestimated, given only clinically apparent macrovascular
venous events were captured overlooking arterial events, asymptom-
atic TE (which remain clinically relevant) as well as the haemostatic, en-
dothelial and inammatory dysfunction at the microvascular level,
which still contribute to morbidity, mortality and potentially disease bi-
ology. As such, this warrants further efforts for preventative strategies.
The predictive power of clinical and thrombogenic biomarkers have
been widely investigated. However, biomarkers as they have been de-
ned to date, that were predictive of TE in general cancer cohorts,
have not shown the same potency among lung cancer patients. The
combined Khorana risk score as well as the individual biomarker pa-
rameters did not consistently stratify lung cancer patients for TE risk
[20,22,23,27,29,30,32,33,35]. By way of example, in a cohort of patients
with small cell lung cancer, the 95% condence interval for TE risk pre-
diction (hazard ratio) of patients with Khorana score 3 vs. b3 ranged
from 0.1 to 7.7 [30].
The most promising biomarker to date appears to be D-dimer. The
expanded Khorana risk score included D-dimer and p-selectin [9],
with the available published data limited to the derivation cohort in-
cluding patients with mixed solid and haematological malignancies.
To our knowledge sub-analyses of lung cancer patients have not been
performed. While application of varied thresholds within the included
studies prevented pool analyses, ndings across all studies support the
use of D-dimer as a predictor of TE in lung cancer [21,22,26,28,32,33].
The expanded Khorana risk model applied a D-dimer threshold
representing the 75th percentile of the cohort (N1.44 mcg/mL) [9].
One of the included lung cancer studies in this review utilised a similar
threshold (N 1.5 mcg/mL) [26], with all other studies signicantly lower
thresholds (b0.5 mcg/mL) [22,26,28,33]. As such, the optimal threshold
remains unclear, particularly as a signicant proportion of patients with
(lung) cancer will have an elevated D-dimer without clinical conse-
quence. If we consider D-dimer levels of the 182 lung cancer patients
within the Vienna Cancer and Thrombosis Study, nearly three quarters
of patients had basal-dimer above the normal reference range of
0.5 mcg/mL (median 0.84 mcg/mL, interquartile range 0.461.71) [13].
Furthermore, basal D-dimer levels differ according to cancer type [13],
which may reect the varying thrombogenicity and suggests that a sin-
gle threshold for all cancers may not be applicable. Further investiga-
tions are warranted to further guide the appropriate threshold for TE
risk assessment in lung cancer patients.
Driver mutations in lung cancer have changed the landscape of prog-
nostication and treatment strategies over the past decade with im-
proved overall survival and increased treatment options [38,39]. Given
our increased understanding of oncogenes, intracellular signalling and
the coagulome [35], there has been a similar interest in correlating mu-
tation status with TE. Studies assessing EGFR mutational status have
demonstrated that positive status confers an almost two-fold increased
risk of TE; pooled HR 1.9 (95% CI 1.23.2) [24,25,31,36]. However,
whether this association is related to the mutational status or the active
intervention is unknown, as studies did not report timing of TE relative
to disease or anticancer treatment. Patients harbouring EGFR mutations
increasingly will receive targeted therapies either as rst-line treatment
or following disease progression after cytotoxic chemotherapy. It is im-
portant to understand these competing TE risk factors in order to make
appropriate decision regarding preventative strategies. Notably, pa-
tients receiving rst-line targeted therapies are not currently incorpo-
rated within existing risk prediction models developed for patients
receiving cytotoxic chemotherapy.
The longitudinal TE risk prole, including dynamic biomarker
changes in response to disease and anticancer treatments, is yet to be
comprehensively evaluated but remains important in our clinical de-
cision-making. A study published earlier this year did report longitudi-
nal evaluation of thrombogenic biomarkers and demonstrated a
dynamic prole but included a heterogonous population and ndings
were not correlated with TE events for individual cancer types, and
therefore was not included in this review [40]. This was considered an
important inclusion criterion for this review recognising that subclinical
 M. Alexander, K. Burbury / Thrombosis Research 148 (2016) 6369
activation of coagulation, which contributes to the cancer-associated
hypercoagulable state, can occur without incidence of TE, and may differ
by cancer type. Furthermore, the importance of recognising and quanti-
fying the magnitude of TE risk extends beyond not only the prevention
of TE but also to disease related outcomes including the prevention of
tumour growth and progression [41,42].
One of the difculties in navigating and assessing the data is the sub-
stantial heterogeneity within and across studies including, patient
population, biomarkers assessed, thresholds used, follow-up schedules
and observation periods, measurement of TE, and the association with
TE. While these factors limited the application of pooled analyses,
there are important insights and hypothesis generation from this sys-
tematic review to guide future, purpose-designed and constructive re-
search, alongside intervention studies, to prevent unnecessary
repetition and optimise use of resources and patient engagement.
Lung cancer contributes to the largest burden of cancer-associated
thrombosis clinically and economically. This review highlights the
importance of study design and outcome reporting, which will hopeful-
ly be taken into consideration in the planning of future studies.
High TE rates in lung cancer warrant consideration of appropriate
preventative strategies. The heterogeneity in risk, favours a targeted,
risk-adaptive approach, incorporating sensitive (and practical) bio-
markers into a risk-model with optimal predictive power, that can be
applied real-time and repeated sequentially to personalise appropriate
decision-making throughout the course of the patients' journey.
Declarations of interest
None declared.
Appendix A. Supplementary data
Supplementary data to this article can be found online at doi:10.
1016/j.thromres.2016.10.020.
References
[1] C. Vitale, M. D'Amato, P. Calabro, A.A. Stanziola, M. Mormile, A. Molino, Venous
thromboembolism and lung cancer: a review, Multidiscip. Respir. Med. 10 (1)
(2015) 28.
[2] L. Corrales-Rodriguez, N. Blais, Lung cancer associated venous thromboembolic dis-
ease: a comprehensive review, Lung Cancer 75 (1) (2012) 18.
[3] L.G. Lima, R.Q. Monteiro, Activation of blood coagulation in cancer: implications for
tumour progression, Biosci. Rep. 33 (5) (2013).
[4] C. Gardiner, P. Harrison, M. Belting, A. Boing, E. Campello, B.S. Carter, et al., Extracel-
lular vesicles, tissue factor, cancer and thrombosis discussion themes of the ISEV
2014 Educational Day, J. Extracell. Vesicles 4 (2015) 26901.
[5] J. Rak, J.L. Yu, J. Luyendyk, N. Mackman, Oncogenes, trousseau syndrome, and can-
cer-related changes in the coagulome of mice and humans, Cancer Res. 66 (22)
(2006) 1064310646.
[6] G.H. Lyman, K. Bohlke, A.A. Khorana, N.M. Kuderer, A.Y. Lee, J.I. Arcelus, et al., Venous
thromboembolism prophylaxis and treatment in patients with cancer: American
Society of Clinical Oncology clinical practice guideline update 2014, J. Clin. Oncol.
33 (6) (2015) 654656.
[7] M. Mandal, A. Falanga, F. Roila, Group ObotEGW, Management of venous thrombo-
embolism (VTE) in cancer patients: ESMO Clinical Practice Guidelines, Ann. Oncol.
22 (Suppl. 6) (2011) vi85vi92.
[8] A. Khorana, N.M. Kuderer, E. Culakova, G.H. Lyman, C.W. Francis, Development and
validation of a predictive model for chemotherapy- associated thrombosis, Blood
111 (10) (2008) 49024907.
[9] C. Ay, D. Dunkler, C. Marosi, A.L. Chiriac, R. Vormittag, R. Simanek, et al., Prediction of
venous thromboembolism in cancer patients, Blood 116 (24) (2010) 53775382.
[10] H.K. Chew, T. Wun, D. Harvey, H. Zhou, R.H. White, Incidence of venous thromboem-
bolism and its effect on survival among patients with common cancers, Arch. Intern.
Med. 166 (4) (2006) 458464.
[11] G.H. Lyman, L. Eckert, Y. Wang, H. Wang, A. Cohen, Venous thromboembolism risk
in patients with cancer receiving chemotherapy: a real-world analysis, Oncologist
18 (12) (2013) 13211329.
[12] M. Verso, G. Agnelli, S. Barni, G. Gasparini, R. LaBianca, A modied Khorana risk as-
sessment score for venous thromboembolism in cancer patients receiving chemo-
therapy: the Protecht score, Intern. Emerg. Med. 7 (3) (2012) 291292.
[13] C. Ay, D. Dunkler, R. Pirker, J. Thaler, P. Quehenberger, O. Wagner, et al., High D-
dimer levels are associated with poor prognosis in cancer patients, Haematologica
97 (8) (2012) 11581164.
69
[14] National Comprehensive Cancer Network, NCCN Clinical Practice Guidelines in On-
cology, Non-small Cell Lung Cancer, NCCN, 2016 (Version 4).
[15] National Comprehensive Cancer Network, NCCN Clinical Practice Guidelines in On-
cology, Small Cell Lung Cancer, NCCN, 2016 (Version 1).
[16] F. Macbeth, S. Noble, J. Evans, S. Ahmed, D. Cohen, K. Hood, et al., Randomized phase
III trial of standard therapy plus low molecular weight heparin in patients with lung
cancer: FRAGMATIC trial, J. Clin. Oncol. 34 (5) (2016) 488494.
[17] Review Manager (RevMan) [Computer program], Version 5.3. Copenhagen: The
Nordic Cochrane Centre, The Cochrane Collaboration, 2014.
[18] G. Wells, B. Shea, D. O'Connell, J. Peterson, V. Welch, M. Losos, et al., The Newcastle-
Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-
analysesAccessed 4th June 2016. Available http://www.ohri.ca/programs/clinical_
epidemiology/oxford.asp2014.
[19] Cochrane Handbook for Systematic Reviews of InterventionsVersion 5.1.0 [updated
March in: J.P.T. Higgins, S. Green (Eds.),The Cochrane Collaboration, 2011 (Accessed
01/03/2015. Available from www.cochrane-handbook.org).
[20] S. Noble, A. Robbins, R. Alikhan, K. Hood, F. Macbeth, Prediction of venous thrombo-
embolism in lung cancer patients receiving chemotherapy, J. Thromb. Haemost. 13
(2015) 143.
[21] I. Evmoradis, P. Boura, A. Rousseau, A. Charpidou, G. Giozos, P. Van Dreden, et al.,
Differential inuence of lung cancer stage, time from diagnosis and chemotherapy
on plasma and cellular biomarkers of hypercoagulability, The Roadmap Study.
Blood Conference: 56th Annual Meeting of the American Society of Hematology,
ASH, 124, 2014, p. 21.
[22] B. Kadlec, J. Skrickova, Z. Merta, L. Dusek, J. Jarkovsky, The incidence and predictors
of thromboembolic events in patients with lung cancer, TheScienticWorldJournal
2014 (2014) 125706.
[23] Y. Zhang, Y. Yang, W. Chen, L. Guo, L. Liang, Z. Zhai, et al., Prevalence and associations
of VTE in patients with newly diagnosed lung cancer, Chest 146 (3) (2014)
650658.
[24] N. Berger, D. Datta, S. Tannenbaum, Venous Thromboembolism and EGFR Mutation
Status in Advanced Adenocarcinoma of Lung, Chest Conference: CHEST, 146, 2014
(4 Meeting Abstract).
[25] A. Delmonte, M. Mariotti, E. Scarpi, P. Ulivi, G. Gavelli, D. Calistri, et al., Adenocarci-
noma of the lung and venous thromboembolic events: a monoinstitutional, retro-
spective evaluation, Journal of Clinical Oncology Conference 33 (15 Suppl. 1) (2015).
[26] P. Ferroni, F. Martini, I. Portarena, G. Massimiani, S. Riondino, F. La Farina, et al.,
Novel high-sensitive D-dimer determination predicts chemotherapy-associated ve-
nous thromboembolism in intermediate risk lung cancer patients, Clin. Lung Cancer
13 (6) (2012) 482487.
[27] C. Gaspar, D. Ferreira, S. Simoes, P. Alves, A. Costa, Risk assessment of venous throm-
boembolism in lung cancer-utility of Khorana model, Eur. Respir. J. Conf. 40 (2012)
(no pagination).
[28] Y. Yang, Z. Zhou, X.H. Ai, X.M. Niu, Z.M. Li, Z.W. Chen, et al., Postoperative venous
thromboembolism in patients with lung cancer. [Chinese], Tumor 32 (5) (2012)
361367.
[29] Y.G. Lee, I. Kim, E. Lee, S.M. Bang, C.H. Kang, Y.T. Kim, et al., Risk factors and prognos-
tic impact of venous thromboembolism in Asian patients with non-small cell lung
cancer, Thromb. Haemost. 111 (6) (2014) 11121120.
[30] Y.G. Lee, E. Lee, I. Kim, K.W. Lee, T.M. Kim, S.H. Lee, et al., Cisplatin-based chemother-
apy is a strong risk factor for thromboembolic events in small-cell lung cancer, Can-
cer Res. Treat. 47 (4) (2015) 670675.
[31] S. Yamazaki, T. Okano, R. Ishikawa, K. Kobayashi, Pulmonary thromboembolism in
lung cancer patients and EGFR mutations, Respirology 18 (2013) 22.
[32] P.P. Zhang, J.W. Sun, X.Y. Wang, X.M. Liu, K. Li, Preoperative plasma D-dimer levels
predict survival in patients with operable non-small cell lung cancer independently
of venous thromboembolism, Eur. J. Surg. Oncol. 39 (9) (2013) 951956.
[33] R. Zhu, J. Liu, Venous thromboembolism in lung cancer with clinical analysis of 89
cases, Eur. Respir. J. Conf. 40 (2012) (no pagination).
[34] S. Arslan, S. Manduz, K. Epozturk, O. Karahan, I. Akkurt, Association of deep venous
thrombosis with prothrombotic gene polymorphism identied in lung cancer cases,
Mol. Biol. Rep. 38 (4) (2011) 23952400.
[35] Z. Wang, H.H. Yan, J.J. Yang, B.C. Wang, H.J. Chen, Q. Zhou, et al., Venous thromboem-
bolism risk factors in Chinese non-small cell lung cancer patients, Support Care Can-
cer 23 (3) (2015) 635641.
[36] L. Corrales-Rodriguez, D. Soulieres, X. Weng, M. Tehfe, M. Florescu, N. Blais, Muta-
tions in NSCLC and their link with lung cancer-associated thrombosis: a case-control
study, Thromb. Res. 133 (1) (2014) 4851.
[37] N.A. Davies, N.K. Harrison, A. Sabra, M.J. Lawrence, S. Noble, S.J. Davidson, et al., Ap-
plication of ROTEM to assess hypercoagulability in patients with lung cancer,
Thromb. Res. 135 (6) (2015) 10751080.
[38] A. El-Telbany, P.C. Ma, Cancer genes in lung cancer: racial disparities: are there any?
Genes Cancer 3 (78) (2012) 467480.
[39] S.Y. Luo, D.C. Lam, Oncogenic driver mutations in lung cancer, Transl. Respir. Med. 1
(1) (2013) 18.
[40] E.M. Reitter, A. Kaider, C. Ay, P. Quehenberger, C. Marosi, C. Zielinski, et al., Longitu-
dinal analysis of hemostasis biomarkers in cancer patients during antitumor treat-
ment, J. Thromb. Haemost. 14 (2) (2016) 294305.
[41] L.-X. Yu, L. Yan, W. Yang, F.-Q. Wu, Y. Ling, S.-Z. Chen, et al., Platelets promote tu-
mour metastasis via interaction between TLR4 and tumour cell-released high-mo-
bility group box1 protein, Nat. Commun. 5 (2014).
[42] C.H. Liu, P. Gong, J. Yang, Relationship between circulating tumor cells and coagula-
tion factors in primary lung cancer patients. Zhonghua zhong liu za zhi, Chin. J.
Oncol. 38 (5) (2016) 368371.