Internal and Emergency Medicine

, Volume 9, Issue 5, pp 559–567

A prospective study on survival in cancer patients with and without venous thromboembolism

Authors

  • Giancarlo Agnelli
    • Internal and Cardiovascular Medicine-Stroke UnitUniversity of Perugia
    • Internal and Cardiovascular Medicine-Stroke UnitUniversity of Perugia
  • Mario Mandalà
    • Medical Oncology UnitOspedali Riuniti
  • Silvano Gallus
    • Department of EpidemiologyIstituto di Ricerche Farmacologiche Mario Negri
  • Claudio Cimminiello
    • Department of Medicine, Vimercate Hospital AziendaOspedaliera di Desio E Vimercate
  • Giovanni Apolone
    • Direzione Scientifica, Azienda Ospedaliera Arcispedale Santa Maria NuovaIRCCS
  • Giovanni Di Minno
    • Department of Clinical and Experimental MedicineFederico II University Hospital
  • Evaristo Maiello
    • Oncology Unit, IRCCS Casa Sollievo Della Sofferenza
  • Paolo Prandoni
    • Department of Cardiothoracic and Vascular Sciences, Clinica Medica 2University of Padua
  • Armando Santoro
    • Humanitas, Cancer Center
  • Lucio Crinò
    • Division of Medical Oncology, Santa Maria Della Misericordia HospitalAzienda Ospedaliera di Perugia
  • Roberto Labianca
    • Medical Oncology UnitOspedali Riuniti
IM - ORIGINAL

DOI: 10.1007/s11739-013-0985-z

Cite this article as:
Agnelli, G., Verso, M., Mandalà, M. et al. Intern Emerg Med (2014) 9: 559. doi:10.1007/s11739-013-0985-z

Abstract

Retrospective population-based studies showed that in cancer patients venous thromboembolism (VTE) is associated with reduced survival. Master Oncology is a multicenter study in patients with solid advanced cancer aimed at assessing (1) risk factors for VTE using a case–control design, and (2) survival in cases (patients with VTE) and controls (patients without VTE). Survival data were prospectively collected for at least 10 months. Overall, 237 cases and 339 controls were included in the analysis. The following factors were found to be associated with an increased risk of VTE: body mass index (BMI; OR 2.02; 95 % CI 1.31–3.12 for ≥26 vs. <23 kg/m2), ECOG score (OR 2.14; 95 % CI 1.47–3.11 for grade 1, and 3.32; 95 % CI 1.64–6.00 for grade 2–3, compared to grade 0) and recent diagnosis of cancer (OR 1.90; 95 % CI 1.33–2.71 for <12 vs. ≥12 months). After an average prospective observation of 8.3 months, 136 cases (57.4 %) and 127 controls (37.5 %) died with a median survival of 8.7 (95 % CI 7.5–10.9) and 14.3 months (95 % CI 12.2–18.7), respectively, (Wilcoxon = 27.72, p < 0.001; multivariate hazard ratio 1.55; 95 % CI 1.21–2.00). Median survival time was reduced for both patients with symptomatic (Wilcoxon = 35.22, p < 0.001) and asymptomatic VTE (Wilcoxon = 4.63, p = 0.031). Patients with advanced solid cancer, high BMI, high ECOG score, and recent diagnosis of cancer are associated with an increased risk for VTE. Patients with both symptomatic and asymptomatic VTE have a reduced survival compared to those without VTE.

Keywords

Venous thromboembolismCancerDeep venous thrombosisPulmonary embolismSurvival

Introduction

Venous thromboembolism (VTE) is a common complication and a major cause of morbidity and mortality in patients with cancer [1, 2]. The risk for VTE complications is particularly high during the hospital stay and after surgery, and is associated with central vein catheter (CVC) and chemotherapy. Several risk factors for VTE have been identified in patients with cancer [3]. These risk factors for VTE can be categorized either as patient characteristics and comorbidities, or as malignancy-related characteristics, including specific therapeutic intervention [4].

Emerging research indicates a detrimental effect of VTE on survival in patients with cancer, even after adjusting for comorbid conditions. It is estimated that in cancer patients who experience a VTE, the risk of death at 2 years is about double in comparison with cancer patients without VTE [5, 6]. The reduced survival associated with VTE has been reported in association with several types of cancer, [511] and it seems to be more pronounced in patients with early-stage cancer (local or regional cancer) and when the VTE occurs during the first year since cancer diagnosis [5, 8].

Most of the survival data related to VTE and cancer come from retrospective analyses performed in population-based or single-hospital registries [58, 10, 11]. We report the results of Master Oncology, a prospective multicenter study in cancer patients with and without VTE.

Patients and methods

Study aims

The aims of Master Oncology were to assess (1) risk factors for VTE using a case–control design, and (2) survival in cases (patients with VTE) and controls (patients without VTE) after at least 10-month prospective follow-up.

Patients

Cases and controls were patients of both genders of 18 years or older affected by histology confirmed cancer (metastatic or locally advanced) of the lung, stomach, colon, pancreas, kidney, ovary, breast, prostate, and other genito-urinary organs. Cases were cancer patients who had a symptomatic or an asymptomatic VTE diagnosed <2 months before the inclusion in the study. Deep venous thrombosis (DVT) was diagnosed by compression ultrasonography or pelvic-abdominal CT scan; pulmonary embolism (PE) was diagnosed by lung CT angiography or lung scan. Controls were cancer patients without VTE enrolled in the same oncology units of the cases. Controls did not undergo imaging studies to exclude a VTE.

Cancer patients with a life expectancy of <3 months or who had surgery <2 months before enrolment were excluded from the study.

Data collection and follow-up

Study patients (cases and controls) were seen at the baseline (at the time of their inclusion in the study), and at the follow-up visits. Demographic and clinical characteristics of cases and controls were collected using an electronic data network. Patient data collection included information on height and weight (used to derive body mass index, BMI, kg/m2); ECOG (Eastern Cooperative Oncology Group) performance status [12]; cancer features, including site, stage and tumor histology, and cytology; anticancer therapies as chemotherapy, radiotherapy, hormonal and anti-angiogenic treatment; concomitant therapies, personal and family history of VTE. For cases, data on the qualifying VTE event were also collected.

During the long-term follow-up, information was collected on vital status and date of death.

Ethics and regulatory issues

The study protocol was submitted to the ethics committees and institutional review boards of the participating study centers for review and written approval. The protocol complied with recommendations of the Helsinki declaration and all its amendments. Written informed consent was obtained prior to the conduct of any study-related procedures. The patient consent form was complying with local regulations and requirements.

Statistical analysis

Odds ratios (OR) for VTE and corresponding 95 % confidence intervals (CI) were estimated using unconditional multiple logistic regression models [13], after allowance for selected covariates, including study center and demographic characteristics. Covariates were selected according to a backward stepwise selection analysis, performed using VTE as the dependent variable and a set of center, patient and clinical characteristics as the independent variables. At the beginning step, variables included in the model were geographic area (North, Center or South of Italy), sex (men or women), age group (<60, 60–69, or ≥70 years), BMI class (<23.0, 23.0–25.9, or ≥26.0), ECOG performance status (Grade 0, 1, or 2–3), cancer site (lung, colon, breast, other sites), tumor histology (adenocarcinoma, squamous cell carcinoma, carcinoma NOS, undifferentiated carcinoma), cancer stage (locally advanced or metastatic), time since cancer diagnosis (<12 months or ≥12 months), chemotherapy (yes or no), radiotherapy (yes or no), anti-angiogenic therapy (yes or no), hormonal therapy (yes or no), and other cancer therapy (yes or no). Variables included in the final model were those whose corresponding χ2 test reached 95 % statistical significance (i.e., age group, BMI class, ECOG score, time since cancer diagnosis).

Using the multivariate OR and the distribution of the main risk factors among cases, population attributable risks (PARs) were computed [14]. PARs represent the proportion of VTE cases that would be avoided if all subjects were not exposed to the factor considered or exposed to the level associated with the lowest risk. PARs were expressed in percentage terms, and for each separate risk factor 95 % CI were obtained as described by Benichou and Gail [15].

The survival analyses were performed through the Kaplan–Meier curves. The Wilcoxon test was used to compare survival curves of cases and controls. Hazard ratios (HR) of mortality and corresponding 95 % CIs were estimated using Cox proportional hazard regression models. Two models were considered: (1) an unadjusted model, and (2) a model including covariates selected according to a backward stepwise selection analysis, performed using VTE condition (cases/controls) plus all the covariates considered in the beginning step of the previously described backward selection, including demographic data, anthropometric measures, tumor characteristics and cancer therapies, as independent variables. Also in this case, variables included in the final model were those who’s HR estimates reached 95 % statistical significance (i.e., VTE condition, ECOG performance status, cancer site, histology and stage, and use of chemotherapy). Only estimates from the multivariate adjusted model are shown in tables.

All the analyses were performed for both the enrolled population and the evaluable population. As no substantial difference was observed, only estimates for the evaluable population are shown in tables.

The original plan anticipated the inclusion of 600 cases and 1,200 controls in the study in more than 50 Italian oncology units. Each case was planned to be individually matched to two controls according to study center, age, sex, cancer site, histology, and stage. The study was then interrupted at the end of December 2009 due to the difficulties in enrolling cases and, more importantly, in finding matched controls. Thus, the matching between cases and controls was not assured, and the sample size was re-computed. A sample of at least 500 patients, 250 cases, and 250 controls, was required to appreciate a statistically significant OR equal or greater than 1.80, assuming an exposure rate among controls of 0.20, α = 0.05 and β = 0.20.

Results

Baseline patient characteristics

Overall, 611 patients with advanced solid cancer (266 cases and 345 controls) were enrolled in 51 Italian oncology units (enrolled population). After exclusion of patients not satisfying eligibility criteria (Fig. 1), data on 576 cancer patients (evaluable population), including 237 cases with VTE and 339 controls were analyzed.
https://static-content.springer.com/image/art%3A10.1007%2Fs11739-013-0985-z/MediaObjects/11739_2013_985_Fig1_HTML.gif
Fig. 1

Included and evaluable population

Among cases, 132 (55.7 %) had a symptomatic and 51 (21.5 %) an asymptomatic DVT, and 26 (11.0 %) had a symptomatic and 53 (22.4 %) an asymptomatic PE. All cases received anti-thrombotic therapy; 89.9 % of cases received low molecular weight heparin (LMWH).

Table 1 shows the distribution of cases and controls and overall according to selected characteristics, originally identified as matching variables. Patients were more frequently women (57.6 %) and had a mean age at enrolment of 64.6 years (age range 23–87 years). The most frequent cancer sites were colon (33.2 %), lung (21.2 %), breast (20.0 %), pancreas (6.6 %), and stomach (6.3 %). More than 75 % of study patients had an adenocarcinoma, and 95 % had a metastatic disease. Although at the interruption of the study matching for study center, age, sex, cancer site, histology, and stage was not assured, cases and controls had a similar distribution for all these characteristics (p value >0.05).
Table 1

Distribution of cancer patients with VTE (cases) and patients without VTE (controls) according to the matching variables

Characteristics

Total

Cases

Controls

p value*

N

%

N

%

N

%

Total

576

100.0

237

100.0

339

100.0

 

Geographic area

 North

370

64.2

144

60.8

226

66.7

0.202

 Center

110

19.1

46

19.4

64

18.9

 South

96

16.7

47

19.8

49

14.5

Sex

 Men

244

42.4

102

43.0

142

41.9

0.783

 Women

332

57.6

135

57.0)

197

58.1

Age (year)

 <60

158

27.4

69

29.1

89

26.3

0.262

 60–69

227

39.4

84

35.4

143

42.2

 ≥70

191

33.2

84

35.4

107

31.6

 Mean (±SD)

64.6 (±9.4)

64.7 (±10.4)

64.5 (±8.7)

0.808°

Cancer site

 Lung

122

21.2

47

19.8

75

22.1

0.451

 Stomach

36

6.3

18

7.6

18

5.3

 Colon

191

33.2

71

30.0

120

35.4

 Pancreas

38

6.6

18

7.6

20

5.9

 Kidney

13

2.3

7

3.0

6

1.8

 Breast

115

20.0

45

19.0

70

20.6

 Ovary

25

4.3

11

4.6

14

4.1

 Prostate

12

2.1

6

2.5

6

1.8

 Other genito-urinary

24

4.2

14

5.9

10

2.9

Cancer histology

 Adenocarcinoma

441

76.6

173

73.0

268

79.1

0.292

 Squamous cell carcinoma

16

2.8

6

2.5

10

2.9

 Carcinoma NOS

99

17.2

49

20.7

50

14.7

 Undifferentiated carcinoma

20

3.5

9

3.8

11

3.2

Cancer stage

 Locally advanced

29

5.0

16

6.8

13

3.8

0.115

 Metastatic

547

95.0

221

93.2

326

96.2

VTE venous thromboembolism, SD standard deviation, NOS not otherwise specified

* p values estimated through χ2 test

° p values estimated from t test

Evaluation of VTE risk factors in cancer patients

The BMI significantly increased the risk of VTE in cancer patients, the OR being 2.02 (95 % CI 1.31–3.12) for the highest in comparison to the lowest level of BMI (p for trend = 0.002). As compared to ECOG 0, the OR was 2.14 (95 % CI 1.47–3.11) for grade 1, and 3.32 (95 % CI 1.84–6.00) for grade 2 (p for trend <0.001). A recent diagnosis of cancer significantly increased the risk of VTE (OR 1.90; 95 % CI 1.33–2.71 for <12 vs. ≥12 months). The distribution of cases and controls, with corresponding multivariate ORs, according to BMI level, ECOG performance status and time since cancer diagnosis is shown in Table 2.
Table 2

Distribution of cancer patients with VTE (cases) and patients without VTE (controls), according to body mass index (BMI), ECOG score and time since cancer diagnosis

Patient’s characteristics

Cases

Controls

ORa

95 % CI

N

%

N

%

BMI (kg/m2)b

 <23.0

71

30.0

127

37.6

1c

 

 23.0–25.9

74

31.2

115

34.0

1.16

0.75–1.77

 ≥26.0

92

38.8

96

28.4

2.02

1.31–3.12

χ2 (p for trend)

10.1 (0.002)

ECOG performance statusb

 Grade 0

94

39.7

198

58.8

1c

 

 Grade 1

107

45.1

114

33.8

2.14

1.47–3.11

 Grade 2–3

36

15.2

25

7.4

3.32

1.84–6.00

χ2 (p for trend)

23.3 (<0.001)

Time since cancer diagnosis (months)

 >12 months

116

48.9

207

61.1

1c

 

 ≤12 months

121

51.1

132

38.9

1.90

1.33–2.71

Corresponding odds ratios (OR) and 95 % confidence intervals (CI)

VTE venous thromboembolism, ECOG Eastern cooperative oncology group

aEstimated using unconditional multiple logistic regression models after allowance for age group, BMI class, ECOG and time since cancer diagnosis (covariates obtained through a backward stepwise selection analysis)

bThe sum does not add up to the total because of missing values

cReference category

Radiotherapy was used by 10.1 % of cases and 5.9 % of controls; no association with VTE risk was observed (OR 1.63; 95 % CI 0.84–3.16). No significant association was also observed for chemotherapy (OR = 0.95), anti-angiogenic therapy (OR 0.89), hormonal therapy (OR 0.75), and other cancer therapies (OR 1.68). The use of analgesics (OR 2.45; 95 % CI 1.54–3.88) and corticosteroids (OR 1.50; 95 % CI 1.04–2.15) increased the risk of VTE, whereas no association was observed for other supportive therapies, including antiemetic and antinauseant (OR 1.44). The use of therapies related to other relevant comorbidities at baseline was not associated to the risk of VTE (OR 1.18).

Attributable risks

The PAR for those factors showing a significant association was estimated, assuming all patients could be moved to the exposure levels with the lowest risk. BMI accounted for 20.0 (95 % CI 1.0–39.1) of cases, ECOG score for 32.1 (95 % CI 19.8–44.4) and their combination for 45.7 %. Months from diagnosis accounted for 19.8 (95 % CI 7.4–32.3). The combination of these three factors explained 56.5 % of cases in this population.

Survival analysis

The survival curves for patients with VTE (cases) and patients without VTE (controls) are reported in Fig. 2. After an average follow-up of 8.3 months, 136 cases (57.4 %) and 127 controls (37.5 %) died. The median overall survival was 8.7 months (95 % CI 7.5–10.9) for cases and 14.3 months (95 % CI 12.2–18.7) for controls (Wilcoxon = 27.72, p < 0.001).
https://static-content.springer.com/image/art%3A10.1007%2Fs11739-013-0985-z/MediaObjects/11739_2013_985_Fig2_HTML.gif
Fig. 2

Survival curves for cancer patients with VTE (cases) and cancer patients without VTE (controls)

Figure 3 shows the survival curves related to cancer patients with symptomatic VTE, cancer patients with asymptomatic VTE, and control patients. Among symptomatic VTE cases, 92 (61.3 %) died and 58 (39.7 %) censored, with a median overall survival of 7.6 months (95 % CI 5.6–9.0). Among asymptomatic VTE cases 44 died (50.6 %) and 43 censored (49.4 %), with a median overall survival of 11.4 months (95 % CI 8.7–17.9). The differences in median survival time between controls and symptomatic (Wilcoxon = 35.22, p value <0.001) or asymptomatic VTE patients (Wilcoxon = 4.63, p value = 0.031) were both statistically significant.
https://static-content.springer.com/image/art%3A10.1007%2Fs11739-013-0985-z/MediaObjects/11739_2013_985_Fig3_HTML.gif
Fig. 3

Survival curves for cancer patients with symptomatic VTE, cancer patients with asymptomatic VTE, and cancer patients without VTE (controls)

In the univariate analysis (raw model), the mortality HR for cases in comparison to controls was 1.79 (95 % CI 1.40–2.28; data not shown). After adjustment for selected covariates, the HR for patients with VTE in comparison to patients without VTE was 1.55 (95 % CI 1.21–2.00).

A high ECOG score was associated to mortality, the multivariate HR being 1.53 (95 % CI 1.16–2.02) for grade 1 and 2.61 (95 % CI 1.79–3.82) for grade 2–3, compared with grade 0. Among other considered factors, age, sex, and BMI did not influence survival. Concerning the cancer site, as compared to lung, colon (HR 0.64; 95 % CI 0.44–0.93) and breast cancer localizations (HR 0.35; 95 % CI 0.22–0.57) were inversely associated to mortality, whereas no significant difference was shown for other sites combined, as well as for time since cancer diagnosis. Compared to patients with adenocarcinoma, those with carcinoma NOS (HR 1.85; 95 % CI 1.26–2.72) and undifferentiated carcinoma (HR 1.89; 95 % CI 1.09–3.29) had an increased risk of mortality. The HR for metastatic cancers was 2.07 (95 % CI 1.08–3.99) compared with locally advanced carcinomas. Concerning therapies, chemotherapy increased the hazard of death (HR 1.90; 95 % CI 1.32–2.75), whereas radiotherapy, anti-angiogenic therapy, hormonal and other cancer therapies had no significant effect on survival. However, these observations are coloure by the large heterogeneity of chemotherapy treatments and no analysis has been performed for single agent treatment and in particular for platinum based regimen. Among other therapies supportive to cancer patients, the use of analgesics (HR 1.71; 95 % CI 1.28–2.28) and corticosteroids (HR 1.31; 95 % CI 1.01–1.70) increased, while the use of antiemetic and antinauseant did not show a significant association with the hazard of death (HR 0.75).

Discussion

Master Oncology showed a reduced survival in patients with metastatic or locally advanced cancer and VTE in comparison with cancer patients without VTE who had similar distribution in terms of demographic and clinical features.

A reduced survival in cancer patients with VTE has been already shown in large population-based studies and retrospective series [511]. However, in these studies patients with and without VTE were hardly comparable according to factors that can influence survival. In our prospective study, cancer patients with and without VTE were comparable in terms of age, sex, cancer site, histology and stage, thus limiting potential interference related to prognostic confounding factors.

This study showed that cancer patients with both symptomatic and asymptomatic VTE have a reduced survival compared to those without VTE. Most of previous studies retrieved only symptomatic VTE and this did not allow an estimation of the impact on survival of asymptomatic VTE patients. Moreover, data on the prognostic value of asymptomatic VTE were mainly achieved in retrospective studies in single center studies [1618].

In our study, the large majority of patients died at home and an autopsy was almost never performed. Therefore, it remains unclear whether the thromboembolic events are the direct cause of death or reflect the presence of a biologically more aggressive cancer that in turn leads to an increased mortality.

We attempted to identify risk factors associated with the development of VTE. Among the variables analyzed by multivariate analysis, high ECOG score, high BMI, and number of months from diagnosis, were significantly associated to VTE risk. In several studies, low performance status was reported as an independent risk factor for the development of VTE [19]. Obesity is a known risk factor for VTE both in the general population and in cancer patients [20]. Patients with <12 months from the diagnosis were at twofold increased risk of VTE as compared to those more than 12 months from diagnosis. This is in agreement with retrospective studies [7], as well as with a recent large population-based study [18].

In our study, we observed that the use of analgesics and steroids is associated with an increase of risk of VTE. However, this observation probably reflects a sicker population.

Our study has some limitations. The study was interrupted before the inclusion of the planned number of patients. This was mainly due to difficulties in recruiting cases and finding matched controls. The reduction in the study sample size mainly affected the statistical power to identify the risk factors for VTE in patients with cancer included in the study. It is, therefore, possible that the role of some VTE risk factors was underestimated. Furthermore, our study was performed in patients with a broad representation of cancer site and type, with locally advanced and metastatic disease, and receiving a wide range of chemotherapy regimens. The observation of an increased mortality in cancer patients with VTE despite the heterogeneity of the study population strengthens the results of the study. However, studies including specific cancer site and type as well as specific chemotherapy regimens would reinforce the value of our observation on the association between VTE and increased mortality.

Our study has some methodological strengths. Survival was assessed in a prospective multicenter study. Cancer patients with and without VTE were comparable in terms of several potential confounding factors for the risk of death.

In summary, Master Oncology showed in a prospective study that VTE, both symptomatic and asymptomatic, is associated with reduced survival in locally advanced and metastatic cancer patients. Further studies are needed to clarify the reasons for this reduced survival and whether preventing VTE in cancer patients is associated with improved survival.

Acknowledgements

The Italian MASTER Oncology study was supported by Sanofi; Sanofi had no role in analyzing the data or preparing the manuscript.

Conflict of interest

As for the present study, Drs. Silvano Gallus and Giovanni Apolone served as statistical consultants for Sanofi. All other authors declare no conflicts of interest related to this article.

Copyright information

© SIMI 2013