Preoperative plasma D-dimer predicts 1-year survival in colorectal cancer patients with absence of venous thromboembolism (VTE): a prospective clinical cohort study

Authors


Mogens Tornby Stender, Department of Surgical Gastroenterology A, Aalborg Hospital, Aarhus University Hospital, Hobrovej 18-22, PO Box 365, DK-9100 Aalborg, Denmark.
Tel.: +45 24437159; fax: +45 99322540.
E-mail: mogens.stender@rn.dk

Abstract

Summary.  Background:  Fibrin formation is required for tumor angiogenesis, metastasis and invasion. Cancer discovered at the same time as or shortly after venous thromboembolism (VTE) tends to be advanced, and the prognosis poor. Previous studies have demonstrated that plasma D-dimer – a degradation product of cross-linked fibrin – correlates with tumor stage and prognosis in patients with colorectal cancer. However, it remains unclear whether D-dimer is of prognostic significance in colorectal cancer patients with absence of VTE.

Objective:  To examine whether the preoperative plasma D-dimer level predicts 1-year survival in pre- and postoperative VTE-negative colorectal cancer patients admitted for surgery.

Methods:  We measured preoperative D-dimer levels in 157 patients, and computed Kaplan-Meier survival curves according to the levels of D-dimer. Cox proportional-hazard regression analysis was used to compute hazard ratio as a measure of 1-year mortality rate ratio, controlling for potential confounding factors. The Aalborg Hospital’s standard cut-off level of 0.3 mg L−1 was used to distinguish negative and positive D-dimer results.

Results:  The overall 1-year survival rate was 87.3% (95% confidence interval (CI), 81.0–91.6%), with 78.1% survival (95% CI, 65.9–86.4%) in the positive D-dimer group compared with 93.6% survival (95% CI, 86.2–97.1%) in the negative D-dimer group. The adjusted hazard ratio of death in the positive D-dimer group compared with the negative D-dimer group was 3.6 (95% CI, 1.3–9.9).

Conclusion:  A positive preoperative D-dimer is associated with a poor prognosis in colorectal cancer patients with absence of VTE.

Background

The association between cancer and thrombosis has been recognized for more than a century [1], and almost all types of cancer are associated with activation of the hemostatic system [2–5]. In addition to induction of thrombosis, there is now solid evidence that tumor-induced coagulation and fibrin formation are required for tumor angiogenesis, metastasis and invasion [6,7], as cross-linked fibrin in the extracellular matrix may serve as a stable framework for endothelial cell and tumor cell migration during angiogenesis and invasion. Moreover, cancer discovered at the same time as or shortly after symptomatic venous thromboembolism (VTE) tends to be advanced, and the prognosis tends to be poor [8]. The extent of coagulation activation, as assessed by the concentration of plasma D-dimer (a degradation product of cross-linked fibrin), has been reported to correlate with tumor stage and prognosis in several cancer types [9–18], including the specific group of colorectal cancer patients [18–23]. However, there is an ongoing debate over whether the increased mortality among cancer patients with elevated D-dimer levels is independent of VTE.

We hypothesized that plasma D-dimer is a predictor of survival, even in the absence of pre- and postoperative VTE. We therefore conducted a prospective cohort study to examine the impact of preoperative plasma D-dimer levels in predicting 1-year survival in colorectal cancer patients with absence of VTE.

Methods

Between October 2003 and November 2005, 157 consecutive patients with colorectal cancer admitted for surgery were enrolled in the study. Exclusion criteria were: previous cancer (within 3 years) or concomitant cancer of any origin, known congenital thrombophilia, a history of arterial or venous thromboembolic events within 3 months prior to the cancer diagnosis, presence of preoperative VTE or development of postoperative VTE during the study period, inflammatory bowel disease, connective tissue disease, severe acute infectious disease (known or suspected infection treated in hospital), stroke or neurosurgery within 6 months, pregnancy, ongoing anticoagulant treatment, or endocarditis. Of 192 patients evaluated for inclusion, 17 were excluded because of presence of preoperative VTE and 18 because of postoperative VTE, leaving 157 patients for final analysis. The study was originally designed to evaluate the capacity of preoperative D-dimer levels to predict postoperative VTE in colorectal cancer patients. The study design has been described in detail previously [24–26].

All patients received standard thromboprophylaxis with tinzaparin 3.500 IU s.c. once daily, starting 2 hours prior to the operation, and graded compression stockings for 7–10 days or until discharge from the hospital, whichever came first.

Diagnosis of DVT and follow-up visits

We performed compression ultrasonography (CUS), including the femoral, popliteal and calf veins, according to standard procedures (grey scale, B-mode, color Doppler) within 48 h prior to the initial scheduled day of surgery, as well as 1 week, 1 month and 1 year after surgery, using a high-end ultrasound scanner (Logiq 9; GE Medical Systems, Milwaukee, WI, USA). In patients with rectal cancer admitted for preoperative radiotherapy, an additional CUS was performed within 48 h of starting radiotherapy. All examinations were performed by one of two experienced ultrasonographers at the Department of Radiology, Aalborg Hospital. The radiologists were unaware of the results of the D-dimer analysis. After 3, 6 and 9 months, the patients were clinically examined by one of the authors (MTS) for signs of DVT and in the case of clinical signs of DVT we performed an additional CUS. Radiologic records were reviewed in patients admitted to neighbouring hospitals by their general practitioner because of suspected DVT in the follow-up period. The patients were classified as having a DVT if so stated in the radiologic records from other hospitals.

In order to detect cancer recurrence, Dukes’ A–C patients underwent physical examination 1, 3, 6, 9 and 12 months postoperatively (including carcinoembryonic antigen (CEA) and hemoglobin levels), computed tomography of the chest and abdomen 12 months postoperatively and colonoscopy within 3 months postoperatively (if not performed preoperatively).

D-dimer assay

Blood samples for D-dimer analysis were obtained at enrollment. The blood samples were obtained according to European Concerted Action on Thrombosis (ECAT) procedures [27] and immediately analyzed as routine samples at the Department of Clinical Biochemistry, Aalborg Hospital. The samples were tested using the Auto Dimer® assay (Biopool® International, Umeå, Sweden), which is the standard assay at Aalborg Hospital. The Auto Dimer® is a quantitative latex test for cross-linked fibrin degradation products. All blood samples were processed according to instructions from the manufacturer. The samples were analysed using the BCS Coagulation Analyser (Dade Behring, Marburg, Germany). One of the authors (TBL) has validated the analytical performance of the Auto Dimer® assay at our institution [28]. The Aalborg Hospital’s standard cut-off level of 0.3 mg L−1 was used to distinguish negative and positive results.

Statistical analysis

Using the Kaplan-Meier method we constructed survival curves according to the cut-off level of plasma D-dimer. Cox proportional-hazard regression analysis was performed to compute hazard ratio as a measure of the 1-year mortality rate ratio, controlling for potential confounding factors. The estimated hazard ratios were adjusted for the following possible confounding factors: age (minimum – 60, 61–80, 81 – maximum), Dukes’ stage (A, B, C or D), gender and tumor site (colon, rectum). The follow-up time was calculated considering the day before surgery as the time of entry into the study. Exit from the study occurred at the end of the study or death. No results of the plasma D-dimer analyses were considered outliers. Data were analysed using stata 9.1 (StataCorp LP, College Station, TX, USA).

Ethics

Oral and written informed consent was obtained from all patients. The study was approved by the Ethics Committee of the Counties of Northern Jutland and Viborg (Jr. No. 2-16-4-0001-03).

Results

One hundred and fifty-seven patients were followed until the end of the study period. Twenty patients died during follow-up (endpoint reached). No patients were lost to follow-up. The overall follow-up time was 1712 months. Twelve patients suffered from Dukes’ stage D or T4 (tumor invasion of surrounding structures) disease and died without a subsequent autopsy. One patient died from postoperative myocardial infarction, one from arrhythmia-induced cardiac arrest during readmission for wound infection, one from cardiac arrest during anesthesia induction prior to reoperation for severe intra-abdominal hemorrhage, one from a postoperative intracranial hemorrhage, and two from postoperative bowel ischemia. One patient, who underwent surgery with curative intent, died at home from unknown reasons (according to the death certificate, the general practitioner stated the cause of death as ‘myocardial infarction’), and one patient who underwent surgery with curative intent died immediately postoperatively due to aspiration-induced cardiac arrest. Aside from standard thromboprophylaxis, none of the patients suffering from major hemorrhage were on anticoagulant drugs.

Characteristics of the study population according to 1-year mortality are shown in Table 1. The overall 1-year survival rate was 87.3% (95% confidence interval (CI), 81.0–91.6%), with 78.1% survival (95% CI, 65.9–86.4%) in the positive D-dimer group compared with 93.6% survival (95% CI, 86.2–97.1%) in the negative D-dimer group (Fig. 1). The crude and adjusted hazard ratios of death in the positive D-dimer group compared with the negative D-dimer group were 3.7 (95% CI, 1.4–9.7) and 3.6 (95% CI, 1.3–9.9), respectively. The comparison of 1-year mortality by predictor variables is presented in Table 2.

Table 1.   Characteristics of the study population
 Overall (n = 157)Survivors (= 137)Non-survivors (= 20)
Gender, female/male, No. (per cent)55/102 (35/65)47/90 (34/66)8/12 (40/60)
D-dimer, positive/negative, No. (per cent)64/93 (41/59)50/87 (36/64)14/6 (70/30)
Age, median (range), years68 (33–94)66 (33–94)74 (49–90)
Dukes’ stages, A–C/D, No. (per cent)130/27 (83/17)122/15 (89/11)8/12 (40/60)
Cancer, colon/rectum, No. (per cent)86/71 (55/45)71/66 (52/48)15/5 (75/25)
Figure 1.

 Kaplan-Meier survival estimates up to 1 year post-surgery, according to the preoperative plasma D-dimer level (Log-rank test for equality of survivor functions, P = 0.004).

Table 2.   Comparison of 1-year mortality by predictor variables
ExposureAdjusted HR
(95% CI)
  1. CI, confidence interval; HR, hazard ratio.

D-dimer
 Negative1.0 (ref.)
 Positive3.6 (1.3–9.9)
Age
 ≤ 601.0 (ref.)
 61–803.0 (0.8–10.8)
 ≥ 814.8 (1.0–23)
Dukes’ stage
 A1.0 (ref.)
 B1.0 (0.1–8.8)
 C2.6 (0.3–25)
 D15 (1.9–117)
Gender
 Male1.0 (ref.)
 Female0.9 (0.3–2.4)
Tumor site
 Colon1.0 (ref.)
 Rectum0.5 (0.2–1.4)

Discussion

We examined the impact of preoperative plasma D-dimer levels in predicting 1-year survival in 157 colorectal cancer patients without pre- or postoperative VTE. We showed that patients with elevated preoperative D-dimer levels were more likely to have a poor prognosis than patients with normal D-dimer levels.

Our findings accord with previous studies addressing the clinical and prognostic significance of D-dimer in cancer patients in general [9,13,14,16,17,29]. Regarding colorectal cancer patients specifically, Kilic et al. [23] found significantly shorter postoperative survival time among patients with elevated preoperative D-dimer levels, and Oya et al. [21] reported that elevated D-dimer levels were associated with relatively advanced tumor stage and short postoperative survival after resection with curative intent. Among patients with metastatic colorectal cancer, Blackwell et al. [19] demonstrated that pretreatment levels of D-dimer were a strong predictor of overall survival among 104 patients treated in a Phase II randomized trial comparing bevacizumab plus 5-fluorouracil/leucovorin with 5-fluorouracil/leucovorin alone. In a new study by Ay et al. [18], assessing the prognostic value of D-dimer in a large group of cancer patients (= 1178), high D-dimer levels were, independently of symptomatic VTE, associated with poor overall survival and increased mortality in general, as well as in the subgroup of colorectal cancer patients (n = 133). Modelling symptomatic VTE as a time-dependent covariate, the study by Ay et al. is, as far as we know, the only study addressing whether the association of high D-dimer levels with poor prognosis is independent of VTE. The present study, which, to our knowledge, is the largest to address the issue of the prognostic significance of D-dimer in colorectal cancer patients, thus confirms the findings of Ay et al., which, very interestingly, suggest that the increased mortality among patients with elevated D-dimer levels is not necessarily mediated by complications of VTE.

As D-dimer seems to be a promising biomarker of aggressive tumor biology, very important clinical perspectives of D-dimer are arising. Studies are urgently needed to establish the role of D-dimer in selecting patients for adjuvant anti-cancer therapy, disease monitoring and as a marker of treatment response.

Our study has strengths and limitations. The prospective recording of data and consecutive cohort of patients reduced the risk of selection and information bias.

With the exception of one patient, who died immediately postoperatively, autopsies were not performed among patients who died in hospital (eight), as the cause of death could explicitly be identified based on the clinical presentation of the patients. The death certificates were filled out accordingly. Eleven out of 12 patients who died at home suffered from Dukes’ D cancer and did not undergo autopsy. We did not review the death certificates of these patients, as it seems unlikely that the general practitioners filling out the death certificates were able to distinguish VTE positive and negative patients. Only an autopsy would have been able to do so. One out of 12 patients who died at home suffered from Dukes’ C disease and received adjuvant chemotherapy. Autopsy was not performed. According to the death certificate, the general practitioner stated the cause of death as ‘myocardial infarction’. As CUS for asymptomatic DVT (between 1 and 12 months postoperatively) and autopsy were not routinely performed, some cases of asymptomatic DVT and fatal PE might have been overlooked, potentially impairing the validity of our findings.

We did not have data regarding co-morbidity. As the number of events was small, however, we had to restrict the number of possible confounding factors included in the Cox regression model. The validity of the results is compromised by the lack of standardization of D-dimer assays. This is due primarily to heterogeneity of the fibrin degradation products, the lack of specificity of antibodies used in the assays, variability of calibrators, and different formats of the assay systems [30]. However, D-dimer is available as a routine analysis at most hospitals. The D-dimer test was applied to a homogeneous patient material, which increased test performance through noise reduction but made it difficult to apply the results generally.

We evaluated an endpoint different from the original study design, which is a limitation that weakens the results.

In conclusion, the present study showed that D-dimer level is an important prognostic feature of colorectal cancer, even in the absence of VTE. However, the clinical implications of this finding remain a major challenge.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary