The activation of the clotting-fibrinolytic system in cancer patients is common and represents an unfavorable clinical sign. D-dimer (DD) is a sensitive marker of fibrinolysis.
The activation of the clotting-fibrinolytic system in cancer patients is common and represents an unfavorable clinical sign. D-dimer (DD) is a sensitive marker of fibrinolysis.
The current study comprised 826 new lung carcinoma patients seen consecutively in a single institution over a 10-year period (1992–2001). For each patient, 31 variables, including DD and survival duration, were available for analysis.
Only weak relationships between DD and the other variables were found. The DD variable correlated best with the level of lactate dehydrogenase, performance status, tissue polypeptide antigen, stage of disease, and the number of metastases (rho = 0.33, −0.25, 0.18, 0.18, and 0.15, respectively). The D-dimer distinguished patients with different prognoses. The median survival periods were 154 days (95% confidence interval [CI], 122–189 days) and 308 days (95% CI, 227–409 days; log rank statistic, 26.56; P < 0.01), respectively, for abnormally elevated and normal values. The difference was greater in patients with adenocarcinoma and in patients presenting with a less advanced disease, especially in patients with pathologic Stage Ia disease. The best multivariate survival model selected 10 significant covariates, including DD.
The authors recommend measuring the plasma level of DD in all new lung carcinoma patients. This measurement may help to formulate individual prognoses and can be used to indicate adjuvant treatment for surgical patients. Cancer 2003;97:3044–52. © 2003 American Cancer Society.
After the first historical paper by Trousseau, which dates back to the nineteenth century,1 the relationship between cancer and blood coagulation has never lost its theoretical and practical interest.2–4 It is now well accepted that patients with cancer may present with one or more circulating markers of hemostatic activation, which underlies a hypercoagulable state.5 This state might have, in turn, a pivotal role in tumor growth regulation and cancer cell dissemination.6, 7 For example, survival benefits have been achieved for patients with certain types of tumors when anticoagulant therapy was used in combination with chemotherapy.8–10
The systemic activation of the clotting-fibrinolytic system in patients with lung carcinoma occurs frequently and is correlated with important clinical variables. A thrombophilic state often is associated with a large tumor burden, clinical progression, low rates of response to chemotherapy, and a poor prognosis.3, 9, 11–17 The evidence, however, was derived from several small or medium studies, and no marker of hemostasis was identified, even in the most recent and expert review of lung carcinoma prognosis.18
The purpose of the current study was to confirm the prognostic significance of hemostatic activation in lung carcinoma patients and to delineate the relationships between markers of hemostasis and other clinical and laboratory characteristics. We limited our attention to the degradation product of fibrin, D-dimer (DD), which is probably the most reliable prognostic indicator of this group.16, 17
From January 1992 to December 2001, 826 consecutive unselected patients (727 men and 99 women; median age, 67 years [range, 35–89 years]) with a cytologically or histologically proven bronchogenic carcinoma were referred to the Unit of Respiratory Medicine at the S. Croce and A. Carle Hospital (Cuneo, Italy), a third referral institution for a community of 500,000 people in northwest Italy. The accrual was distributed uniformly across the 10 years of study. The median accrual date was November 13, 1996. The pathologic diagnosis of primary lung carcinoma was carried out in accordance with the revised World Health Organization classification of lung tumors.19 The lung tumors included 296 squamous cell carcinomas, 210 adenocarcinomas, and 93 small cell carcinomas (the other, less common pathologic diagnoses are listed in Table 1). All patients were classified according to the 1997 International System for Staging Lung Carcinoma,20 either retrospectively21 or prospectively. Pretreatment clinical evaluation was based on a few clinical tests and examinations and mostly remained unchanged during the decade of study. The pretreatment evaluation consisted of a physical examination, assessment of both present and past (6 months before the diagnosis) body weight, the evaluation of Karnofsky performance status (PS).22 A battery of laboratory tests, including the measurement of two key tumor markers with prognostic significance (carcinoembryonic antigen [CEA]23 and tissue polypeptide antigen [TPA]),24 were also performed. The instrumental evaluation consisted of X-rays, bronchoscopy, and computed tomography scans of the chest, upper abdomen, and brain. In potentially resectable tumors, a radiologic finding equivocal for mediastinal involvement was considered an indication to mediastinoscopy. Patients with dubious distant metastases were investigated further with other, more appropriate imaging studies or with targeted biopsies or needle aspirations. Bone scans and bone marrow biopsies were performed in the majority of small cell lung carcinoma patients. Based on the results of the clinical evaluation, surgery was appropriate for 155 patients (18.8%). In contrast, consistent with our early recognition of its potential benefits,25 chemotherapy was the most frequent treatment modality for 385 patients (46.7%), either alone or in combination with thoracic irradiation. Curative radiotherapy was the only treatment in 28 patients (3.4%). The remaining 258 patients (31.2%) received symptomatic or supportive care alone. Details of the treatment protocols used during the last decade, including the type and scheme of cytostatic drugs, have been reported already.21
|Characteristic||No. of observations||Median (range) or frequency|
|Age (yrs)||826||67 (35–89)|
|Karnofsky PS (≤ 40/50/60/70/80/90/100)||826||20/46/94/215/246/136/69|
|Weight loss in the 6 mo before diagnosis (> 10/≤ 10%)||826||444/382|
|Hemoglobin (g/dL)||824||13.6 (7.0–18.8)|
|White blood cells (no./mm3)||825||8530 (2890–36,400)|
|Neutrophils (no./mm3)||821||5790 (832–32,600)|
|Platelets (no./mm3 × 1000)||826||280 (72–982)|
|Lactate dehydrogenase (mg/dL)||798||399 (140–4985)|
|Alkaline phosphatase (mg/dL)||820||92 (35–2171)|
|Alanine transaminase (mg/dL)||825||19 (2–259)|
|Aspartate transaminase (mg/dL)||826||19 (5–208)|
|Total serum proteins (g/dL)||817||7 (4.4–9.09)|
|Creatinine (mg/dL)||826||0.94 (0.5–4.80)|
|Sodium (mEq/mL)||820||140 (114–148)|
|Carcinoembryonic antigen (ng/mL)||822||13 (0–7584)|
|Tissue polypeptide antigen (U/L)||813||130 (30–4000)|
|D-dimer (μg/mL)||826||0.5 (0.0–17.4)|
|Stage of disease (0/Ia/Ib/IIa/IIb/IIIa/IIIb/IV)||826||5/58/91/7/50/101/209/305|
|T factor (0/1/2/3/4)||826||5/118/290/117/296|
|N factor (0/1/2/3)||826||330/81/275/140|
|M factor (0/1)||826||521/305|
|Lung metastases (Y/N)||826||121/705|
|Brain metastases (Y/N)||826||96/730|
|Liver metastases (Y/N)||826||85/741|
|Adrenal gland metastases (Y/N)||826||40/786|
|Bone metastases (Y/N)||826||81/745|
|Other metastases (Y/N)||826||33/793|
|No. of metastases per patient (0/1/2/3/4/5/6)||826||521/191/71/29/9/4/1|
|Main treatment (surgical/nonsurgical)||826||155/671|
The survival period was recorded from the time of histologic diagnosis to the date of death or to the closure of the study (December 2001). The status of the patient (i.e., deceased or living) was verified by a telephone contact with the patient (if living) or with the family, physician, or the municipal office of the registry. This information was available for all patients in the study. As of December 2001, 173 patients (20.9%) were still alive after a median follow-up time of 62 weeks (range, 2–484 weeks). The median follow-up time (up to death or the last follow-up contact) was 34 weeks for the 826 patients enrolled in to the study.
The current retrospective study was based on a prospective database that was used to assess the clinical significance of blood coagulation tests in lung carcinoma patients. We recorded dozens of variables, but only 31 variables were used for patient evaluation in the current study. Besides DD levels, we also analyzed age, gender, PS, weight loss in the 6 months preceding the diagnosis (defined as the pecentage loss of the previous body weight), hemoglobin blood content, total white cell/neutrophil counts, tumor cell type, TNM clinical stage, and the number of metastatic sites. In addition, we analyzed the serum concentration of each of the following enzymes and substances: lactate dehydrogenase (LDH), alkaline phosphatase, alanine and aspartate transaminases, creatinine, sodium (Na+), CEA, and TPA. Follow-up programs were the same for all patients and remained substantially unchanged during the study period. They consisted of clinical, laboratory, and radiologic reassessments performed at 3–4-week intervals during chemotherapy and every 3–6 weeks for patients who received palliative radiotherapy or no anticancer treatment. Patients treated with radical surgery were scheduled for clinical evaluation at intervals that ranged from 3 to 6 months.
Commercially available reagents (Diagnostica Stago; Boehringer Mannheim, Mannheim, Germany) were used to measure DD (Sta Liatest D-DI kit; Boehringer Mannheim). The manufacturer's instructions were followed for all determinations. Patients' blood samples were obtained at presentation and processed immediately or they were stored for a maximum of 2 days at −20 °C. The reference value for DD was less than 0.5 μg/mL.
Statistical analysis was performed using the SPSS package for Windows, version 9.0 (SPSS Inc., Chicago, IL). Nonparametric methods26 were used for descriptive purposes and to assess statistical relationships between DD levels and the other variables (i.e., the Spearman rank and Kruskall–Wallis tests).
The 31 parameters listed in Table 1 were considered for survival analysis. Survival time was the dependent variable. Survival functions were obtained using the Kaplan–Meier method27 and the PROC LIFETEST of SAS.28 Survivals were calculated in days and curves were plotted graphically by years. Differences among survivals were tested statistically with the log rank test.29
To control the effect of potential confounders, multivariate analyses based on the Cox proportional hazards regression model30 were performed. Variables included in the models were all significant in univariate analysis. For each variable, the proportional hazards assumption was tested graphically.
The exponent of the coefficient estimated from the regression model was assumed as the hazard ratio (HR) of dying during the follow-up period for subjects in the exposed category of each variable compared with the reference category and after having allowed for the other factors entered in the model. The PROC PHREG of SAS was used for Cox regression.28 Ninety-five percent confidence intervals (95% CI) of HRs were calculated by exponentiating the estimated coefficient and adding or subtracting 1.96 times its standard error.
We fit several Cox models. The goodness-of-fit of each model was assessed using the equation D = −2 · ln(likelihood ratio). Because the D statistic provides a measure of the unexplained variability, the lower its value, the better is the goodness-of-fit of the model to the data. When feasible, the comparison between the two models was made by measuring the difference between the values of the D statistic (likelihood ratio test [LRT]). The LRT has an asymptotic chi-square distribution; i.e., the degrees of freedom are equal to the difference between the degrees of freedom of the D statistic of the two models compared.31 A P value of less than 0.05 was statistically significant. All tests were two-sided.
Table 1 lists the anthropometric and clinical characteristics of the 31 variables. The mean value (standard deviation) of the DD measured in the 826 pretreatment assays was 1.1 μg/mL (1.4) and the 25th, 50th, 75th, 96th, and 99th percentiles were 0.1, 0.5, 1.6, 4, and 4.7 μg/mL, respectively. Only 306 patients (37%) exhibited normal DD values.
There was no relationship between the DD level and the cell type (P = NS, Kruskall–Wallis test). Similarly, a strong association was not found between the DD level and each of the other variables recorded, although the correlation tests were often significant, due to the high number of observations (Table 2). The DD correlations with LDH (Spearman rho [rs] = 0.33), Karnofsky PS (rs = −0.25), TPA (rs = 0.18), and the stage of disease (rs = 0.18) were the most appreciable. The associations between DD level and the presence, type, and number of metastatic sites or between the DD level and other hematologic and biochemical parameters were also statistically significant (Table 2). In essence, elevated plasma DD levels were associated with advanced-stage disease and with poor clinical and biochemical status.
|Gender (male/female)||− 0.04|
|Karnofsky PS (<40/50/60/70/80/90/100)a||− 0.25|
|Weight loss (Y/N)b||0.07|
|Hemoglobin (g/dL)a||− 0.12|
|White blood cells (no./mm3)a||0.10|
|Platelets (no./mm3 × 1000)a||0.10|
|Lactate dehydrogenase (mg/dL)a||0.33|
|Alkaline phosphatase (mg/dL)||0.02|
|Alanine transaminase (mg/dL)a||0.11|
|Total serum proteins (g/dL)||− 0.06|
|Creatinine (mg/dL)||− 0.03|
|Carcinoembryonic antigen (ng/mL)a||0.10|
|Tissue polypeptide antigen (U/L)a||0.18|
|Stage of disease (0/Ia/Ib/IIa/IIb/IIIa/IIIb/IV)a||0.18|
|T factor (0/1/2/3/4)a||0.11|
|N factor (0/1/2/3)||0.07|
|M factor (0/1)a||0.14|
|Lung metastases (Y/N)||0.03|
|Brain metastases (Y/N)a||0.13|
|Liver metastases (Y/N)a||0.13|
|Adrenal gland metastases (Y/N)||0.06|
|Bone metastases (Y/N)||0.09|
|Other metastases (Y/N)||0.07|
|No. of metastases per patienta||0.15|
|Main treatment (surgical/nonsurgical)a||− 0.17|
Univariate analyses of survival confirmed the already known associations between survival and stage of disease, TNM factors, Karnofsky PS, and weight loss (P < 0.01). Also prognostically meaningful were age at diagnosis; the concentrations of TPA, LDH, and Na+; the number of white blood cells, neutrophils, and platelets; and the serum concentration of hemoglobin (Table 3).
|Variable||No.||Quartile estimates (days)||P (log rank test)||Variable||No.||Quartile estimates (days)||P (log rank test)|
|≤ 63||289||146||326||774||< 0.01||≤ 80||290||215||418||1460||< 0.01|
|> 70||268||79||237||455||> 200||270||58||174||305|
|≤ 70||375||61||168||340||< 0.01||< 3||414||137||303||698||< 0.01|
|> 70||451||208||385||936||≥ 3||408||96||256||512|
|≤ 12.7||278||77||219||415||< 0.01||≤ 0.3||294||162||347||953||< 0.01|
|White cell count||Histology|
|≤ 8500||411||146||326||765||< 0.01||Squamous||296||56||204||608||< 0.05|
|> 8500||414||90||224||468||Small cell||93||158||317||463|
|≤ 10%||382||160||332||765||< 0.01||Stage|
|> 10%||444||89||234||463||I||154||418||1342||5837||< 0.01|
|≤ 5800||412||168||350||804||< 0.01||III||310||143||280||451|
|Platelet count||Tumor status|
|≤ 241||280||155||333||772||< 0.01||T1||123||236||692||2252||< 0.01|
|Alkaline phosphatase||Lymph node status|
|≤ 81||279||148||332||745||< 0.01||N0–1||411||168||377||1336||< 0.01|
|≤ 326||267||199||373||1019||< 0.01||Absent||521||180||361||889||< 0.01|
|> 484||264||72||194||367||CNS metastases|
|≤ 6.6||284||72||210||431||< 0.01||Present||96||57||121||240|
|> 6.7||272||158||300||616||Surgical||155||211||759||5837||< 0.01|
|< 139 or > 141||475||95||243||512|
Abnormally elevated concentrations of DD were a strong predictor of poor outcome (the median survival period of patients with normal levels of DD was 308 days [95% CI, 227–409 days] vs. 154 days [95% CI, 122–189 days] for patients with abnormal values; log rank statistic, 26.56; P < 0.01). The difference remained significant in several strata of established prognostic factors and in a few specific cell types. In particular, DD showed good prognostic ability in patients with the least advanced stages of disease (Stages Ia–IIb; Fig. 1), as well as in patients with many other subsets of early disease, such as Stage T1 disease (Fig. 2). In Stage Ia–IIb disease, the median survival period for patients with normal plasma levels of DD was 1343 days (95% CI, 485–2201 days) compared with 642 days (95% CI, 310–982 days) for patients with abnormal values (log rank statistic, 4.44; P < 0.05). In Stage T1 disease, the median survival periods were 850 days (95% CI, 453–1252 days) and 456 days (95% CI, 186–726 days) in patients with normal and abnormal DD concentrations, respectively (log rank statistic, 10.83; P < 0.05). Even in the small group of 58 patients with Stage Ia disease, DD levels were still highly predictive, being able to separate patients at different outcomes (median survival period for patients with normal DD levels, 337 days [95% CI, 289–385 days] vs. 236 days [95% CI, 211–261 days] for patients with abnormal DD levels; log rank statistic, 26.56; P < 0.01). The DD level was most prognostically efficient in patients with adenocarcinoma (Fig. 3).
Table 4 summarizes the best multivariate model. In this model, the plasma levels of DD emerged among the 10 significant covariates. Finally, as shown in Table 5, the addition of DD increased the goodness-of-fit of the model based on four classic prognostic factors (stage of disease, Karnofsky PS, age, and weight loss).
|Variable||Hazard ratio||95% CI|
|Lymph node status|
|Variable||Model 1b Hazard ratio (95% CI)||Model 2b Hazard ratio (95% CI)|
|II||2.0 (1.4–3.1)||2.0 (1.3–3.0)|
|III||4.2 (3.2–5.5)||4.1 (3.1–5.5)|
|IV||5.6 (4.3–7.4)||5.4 (4.1–7.2)|
|> 70||2.1 (1.8–2.5)||2.0 (1.7–2.4)|
|63–70||1.2 (1.0–1.5)||1.2 (1.0–1.5)|
|> 70||1.3 (1.1–1.6)||1.3 (1.1–1.6)|
|Present||1.1 (0.9–1.2)||1.1 (0.9–1.2)|
|> 1.0||—||1.4 (1.2–1.7)|
|Chi-square||—||11.5e (D1 − D2)|
Approximately 90% of cancer patients with metastatic disease and one-half of all cancer patients have 1 or more abnormal coagulation parameters.7 Several laboratory abnormalities have been described, including prolonged and shortened prothrombin time (PT), partial thromboplastin time (PTT), increased and decreased levels of Factor II (F-II), Factor V, Factor VIII, Factor IX, Factor XI, Factor XII, fibrinogen (F), F/fibrin degradation products, the thrombin-antithrombin III complex (TAT), and thrombocytosis.7 In addition, changes in the clotting/fibrinolytic system in lung carcinoma patients are common.3, 9, 11–17 Clinical and experimental evidence supports the idea that the activation of coagulation/fibrinolysis may play an important role in the invasiveness of cancer.6, 7 For example, Kwaan and Keer6 demonstrated a correlation between coagulation changes and the natural history of malignancies.
D-dimer is the smallest degradation product of fibrin, resulting from the proteolytic action of plasmin.32 It is a sensitive marker of the fibrinolytic enhancement.32 As in many other human malignancies,33–38 including lung carcinoma, DD might have prognostic significance.12, 15, 39 In 1993, Seitz et al.12 reported that TAT and DD were significantly increased in patients with metastatic disease and that elevated TAT values at diagnosis might be prognostically significant. Taguchi et al.15 measured the plasma levels of DD in 70 lung carcinoma patients. They found that low values of DD were significantly predictive of both good prognosis and longer survival times. Pavey et al.39 studied fibrinolytic markers in 166 lung carcinoma patients. Their principal aim was to evaluate the effects of cutoff values determined in three ways. First, Model 1 assigned patients to one of three equal groups (low, medium, or high) based on the fibrinolytic measurements made at diagnosis. Second, Model 2 divided patients into low/high groups using median values. Third, Model 3 grouped patients according to whether the parameter was above or below the normal range. Relative risk estimates indicated that elevated levels of plasma F, soluble fibrin, and DD were associated with increased risk of death. The authors recommended the use of the normal/above-normal cutoff to provide the maximum number of significant parameters relating to prognosis. They also emphasized that increased plasma DD levels were among the parameters related more closely to survival and prognosis.
The correlation between clotting activation and the clinical behavior of lung carcinoma was investigated by our group in two consecutive studies.16, 17 In the earlier study,16 we reported that several clotting tests were predictive of prognosis, according to both univariate and multivariate models. Univariate tests showed that low values of plasma PT, high values of F, and abnormally elevated concentrations of DD were associated significantly with an adverse outcome. The multivariate models confirmed the significance of several clotting tests, including DD. The later study17 assessed the potential prognostic significance of a wider selection of coagulation parameters. Platelet counts and the plasma levels of PT, PTT, antithrombin III (AT-III), F, DD, F-II, Factor VII (F-VII), Factor X (F-X), protein C clotting, plasminogen, and antiplasmin were obtained from 343 consecutive new lung carcinoma patients. The study confirmed that a prethrombotic state often is present in patients with lung carcinoma. The most frequent coagulation abnormalities were DD, F, AT-III (55%, 42%, and 28% of patients, respectively), F-VII (27%), F-X (20%), and F-II (16%). Confirming our previous report,16 univariate analyses of survival showed that a prolonged value of PT and high values of platelet count, F, and DD were associated with a poor prognosis. However, in this second study,17 the multivariate model did not confirm the prognostic relevance of any coagulation factor. We concluded that new studies should address the issue in specific subgroups of lung carcinoma patients.
The current study continues the work of these studies. To increase the number of observations and to improve the simplicity of the message (a prerequisite for any message to be accepted), we reviewed our decennial experience with coagulation abnormalities and lung carcinoma. We focused on DD, a unique marker of coagulation and fibrinolytic activation. Our previous experience16, 17 and other studies12, 15, 39 indicated that DD was the most useful variable. The weakness of the current study is that it does not provide any completely new evidence (approximately 7 of the 10 years encompassed by the current report, and 629 of 826 patients [76%] were included in the 2 previous publications). Conversely, consolidating all the data would have greatly enhanced the message and, more importantly, would have permitted the analysis of subgroups of particular interest.
In summary, we have confirmed conclusively that DD levels in lung carcinoma patients increased in parallel with the severity of the illness, as demonstrated by the correlation with stage of disease, performance status, and other biochemical markers of severity of disease (e.g., LDH and TPA). Increased plasma levels of DD strongly predict a poor prognosis, especially in patients with adenocarcinomas.
Even adjusting for stage of disease, DD differentiates patients with opposite outcomes, especially in earlier stages of disease (e.g., in patients with Stage T1 tumors). D-dimer significantly increases the prognostic information of the four classic prognostic factors (stage of disease, Karnofsky PS, age, and weight loss). Even with the availability of 31 variables, all with certain or very likely prognostic significance, DD emerges as an important independent covariate of the best-fit Cox proportional hazards model.
Based on this evidence, we advocate the use of a DD test in any new lung carcinoma patient to provide better physician advice after a prognosis is indicated, as well as to encourage the initiation of adjuvant treatment trials for surgically cured patients who manifest a pathologic increase in their plasma levels of DD. Future reviews of prognostic factors for lung carcinoma should evaluate the clinical implication of the activation of the clotting-fibrinolytic system, as signaled by DD.
The authors thank Anna Merlo for her invaluable help and support in the outpatient unit.