Abstract
Background
Risk of venous thromboembolism (VTE) recurrence remains high in patients with cancer-associated thrombosis (CAT), despite therapeutic anticoagulation. Identifying patients at risk of treatment failure is still a challenge.
Objectives
We aimed to assess the performance of the Ottawa score in predicting VTE recurrence in a large homogeneous population of patients with CAT treated with the same anticoagulant, tinzaparin, for at least 3 months.
Methods
Individual patient data from 3 prospective cohort studies and 1 randomized controlled trial were pooled (PROSPERO: CRD42019119907). Clinical events of interest were adjudicated by independent central adjudication committees in all 4 studies.
Results
Among the 1413 patients included, the Ottawa score could be calculated for 1088 of whom 646 (59.4%) were classified at high risk of recurrence (Ottawa score ≥ 1). The 6-month cumulative incidence of recurrent VTE was 5.0% (95% CI, 3.2-7.8) in the Ottawa low-risk group and 8.5% (95% CI, 6.6-10.8) in the high-risk group. The area under the receiver operating characteristic curve was 0.56 (95% CI, 0.51-0.62). The sensitivity of the dichotomized Ottawa score (score ≥ 1) was 72.8% (95% CI, 62.6%-83.0%), the specificity was 41.9% (95% CI, 37.8%-45.9%), the positive predictive value was 8.6% (95% CI, 6.4%-10.8%), and the negative predictive value was 95.3% (95% CI, 93.3%-97.4%). Introducing additional predictive factors failed to significantly improve the score’s performance.
Conclusions
Despite the large number of patients and anticoagulant treatment standardization, the Ottawa score failed to accurately predict recurrent VTE in patients with CAT treated with tinzaparin.
Keywords: cancer, individual patient data meta-analysis, Ottawa score, tinzaparin, venous thromboembolism
Essentials
-
•
Ottawa score evaluated in a pooled cohort of 1413 patients with CAT started on tinzaparin.
-
•
Recurrent VTE rates were 5.0% in low-risk (41%) and 8.5% in high-risk (59%) groups at 6 months.
-
•
Ottawa score poorly predicted recurrent VTE (area under the curve, 0.56).
-
•
Adding other factors did not improve the score’s predictive accuracy.
1. Introduction
Compared with noncancer patients, patients with active cancer have both a 3- to 6-fold higher risk of incident venous thromboembolism (VTE) and a 3- to 6-fold higher risk of recurrent VTE despite curative anticoagulation [1]. Current guidelines for anticoagulation in cancer patients recommend use of low-molecular-weight heparin or a direct oral anticoagulant, based on randomized controlled trial results [[2], [3], [4], [5], [6], [7]]. Adequate anticoagulation treatment of cancer patients is complicated by many factors including comorbidities, polymedication, and increased risk of bleeding compared with that of noncancer patients. Ability to predict the risk of recurrent VTE during anticoagulation might facilitate selection of the best treatment option for patients with cancer-associated thrombosis (CAT) or contemplation of a more vigilant approach to care. Among the various prediction rules developed to assess the risk of recurrent VTE in patients with CAT, the most extensively studied index is the so-called Ottawa score, described by Louzada et al. [8]. Several cohort studies have attempted to validate this score with conflicting results [[9], [10], [11], [12], [13]]. Even the conclusions of meta-analyses performed on aggregated data are contradictory. One meta-analysis concluded that the original Ottawa score is an accurate tool for stratifying the risk of recurrent VTE, in particular for reliably identifying patients at high risk of recurrent VTE, but did not assess the predictive performance of the score [14], while the other showed poor performance of the score [13]. In these studies, the Ottawa score was mainly assessed in retrospective cohorts of patients with CAT included between 1995 and 2014 receiving now outdated antitumoral therapies and various types and doses of anticoagulant. Control of the variability induced by anticoagulation is essential to validate a risk score during such treatment. In a prospective cohort study of patients included from 2015 to 2016 and treated with tinzaparin, these 2 limitations were controlled, but the accuracy of the Ottawa score could not be validated [15]. This shortcoming might have been due to inability of the score to adequately capture the factors linked to increased risk of recurrent VTE. Specifically, different cancer sites (eg, lung, colorectal, breast, and hematologic) are associated with disparate VTE risks and involve the use of heterogeneous antitumor treatments varying in their thrombogenicity. In a sufficiently large population with CAT, additional risk factors could be identified permitting enhanced performance of the current score or development of an alternative score.
By analyzing individual patient data from prospective studies in cancer patients treated with a standardized tinzaparin regimen for a VTE episode, we aimed to improve the ability to predict recurrent VTE in the CAT population.
2. Methods
2.1. Protocol registration
The study protocol was prospectively registered in the International prospective register of systematic reviews (PROSPERO, https://www.crd.york.ac.uk/PROSPERO, registration number CRD42019119907). For the individual patient data meta-analysis, we adhered to the Preferred Reporting Items for Systematic Review and Meta-Analyses of Individual Participant Data [16].
2.2. Literature search and study identification
We sought to identify all relevant published and unpublished prospective studies, that is, observational cohorts or randomized controlled trials, including patients with CAT treated with tinzaparin after an episode of VTE and followed up for at least 6 months. An exhaustive literature search was performed without any restriction on language or publication period. The computer-assisted search was carried out on electronic databases (Medline, Web of Science, and Google Scholar), using the combination of keywords and medical subject headings terms related to treatment exposure (tinzaparin [Innohep]), outcome (eg, venous thromboembolism, deep vein thrombosis [DVT], and pulmonary embolism [PE]), and population (eg, cancer, acute venous thromboembolism, and cancer-associated thrombosis). Reference lists of retrieved articles and review were manually scanned to identify all relevant additional studies. The search was updated on December 26, 2023. The full electronic search equation is presented in Supplementary Table 1. Individual patient data were requested and obtained from the coordinating investigator of each trial included.
2.3. Study selection
Studies were eligible for inclusion if (1) they were observational, prospective, cohort studies or randomized controlled trials; (2) with a recruitment period after 2003 (excluding studies over 20 years old); (3) including patients with CAT treated with tinzaparin for at least 3 months and followed up for at least 6 months; (4) with individual patient data available; and (5) with adjudication of clinical events of interest (recurrent VTE and major bleeding) by an independent central adjudication committee as part of the study. For randomized controlled trials, only the tinzaparin arm was considered.
2.4. Data extraction
Electronic versions of the data sets were requested from the coordinating investigator in SAS format if possible, or otherwise in EXCEL or CSV format, with written details of the dictionary and coding of variables. The data required were defined and selected a priori to correspond to the minimum data required to meet the primary research objectives. The statistician collated and cleaned the data, and harmonized the databases. Additional data checks (eg, relating to data generation and harmonization, data consistency, and completeness) were performed by comparing the generation of the acquired data to published results or statistical reports. Individual patient data from each study were then pooled in a common database. The final analyses were performed on the collated and merged data set after the abovementioned steps.
Race and ethnicity data were only available in one of the included data sets [2]. In the French cohorts, collection of these data is restricted by law, and they were not recorded in the Spanish cohort. Consequently, these variables were not included in the pooled analysis.
2.5. End points
The primary end point of the meta-analysis was objectively confirmed recurrent VTE (DVT or PE) within the first 6 months of follow-up, either symptomatic or incidentally detected. Secondary outcomes included major bleeding and all-cause deaths during the same follow-up period. The definitions of recurrent VTE and major bleeding adjudicated by the independent central adjudication committee of each study were identical or very similar and were maintained for this meta-analysis.
2.6. Methodological quality of the studies
The Newcastle-Ottawa scale (NOS) was used to assess the quality of the cohort studies (or tinzaparin arms of randomized controlled studies) included in the meta-analysis [17]. The NOS is a star system containing 8 items categorized into 3 dimensions comprising (1) study selection, (2) comparability of the cohorts according to their design or analysis, and (3) study outcome. A maximum of 9 stars can be awarded for each study based on these 3 dimensions. In this meta-analysis, the second dimension of the scale concerning the comparability of cohorts was not applicable. A maximum of 7 stars could therefore be awarded for each study. Two authors independently assessed the quality of the studies included, with discrepancies between authors being resolved by discussion.
2.7. Statistical analysis
Individual data concerned all included patients with evaluable data for recurrent VTE. The Ottawa score was calculated by adding or subtracting the stated number of points for the following baseline data: female gender (+1 point), TNM stage I (−2 points), lung cancer (+1 point), breast cancer (−1 point), and a history of previous VTE (+1 point) [8]. Patients with a score of <1 were classified as being at low risk of recurrence, and those with a score of ≥1 as being at high risk of recurrence. To assess overall discrimination, the area under the receiver operating characteristic (ROC) curve of the Ottawa score was calculated with its 95% CI. The performance of the dichotomized Ottawa score (score ≥ 1) was assessed according to its sensitivity, specificity, and positive and negative predictive values. To screen for additional potential risk factors for recurrent VTE up to 6 months, the Fine and Gray regression model, considering death as a potential competing risk, was used, adjusted on the variables of the Ottawa score. This model generated subdistribution hazard ratios (subHR) with their corresponding 95% CI. The following additional variables were also tested: age, Eastern Cooperative Oncology Group index, type of tumor, evolutionary stage, renal function, and anticancer therapies. Univariate analyses were first implemented for each of these factors. Factors with a P value <.15 and a frequency > 3% were then introduced into a multivariate model adjusted on the variables of the Ottawa score. Variables significant at a threshold of 5% were considered for construction of the risk score for recurrent VTE. A multivariate model, including the variables significant at a threshold of 5% in the step-by-step model, was then constructed, and its Wolbers c-statistic was calculated [18]. An enhanced score was subsequently developed by adding the points assigned to each significant variable based on the regression coefficients calculated in the multivariate model. The predictive power of this score was evaluated in the study cohorts, and the area under the ROC curve associated with the score was determined. This newly constructed enhanced score was internally validated, with its predictive power being evaluated by the bootstrapping method. The same methods were used to develop an alternative score designed to improve the prediction of recurrence risk. All analyses were performed using SAS software (version 9.4; SAS Institute). R software (version 4.2.1; R Core Team) was used to generate graphs.
3. Results
3.1. Study selection
The search procedure identified 949 references (Figure 1). After exclusion of 939 studies, including 59 duplicates, 10 studies were potentially eligible for analysis. For 6 of these studies, individual patient data were not sought owing to the lack of blinded central event adjudication [13,[19], [20], [21], [22]] and the fact that the patients were recruited >20 years ago [23]. Finally, individual patient data were available for 3 prospective cohort studies—TiCAT [24], PREDICARE [15], and AXA (not yet published) and 1 randomized controlled study (CATCH [2]) (Table 1). All these studies were multicenter. Two studies were sponsored by LEO Pharma [2,15], and the others by University Hospitals in France (AXA) and Spain (TiCAT) [24]. In total, 1413 patients were treated with tinzaparin, with the number of treated patients per study varying from 247 [24] to 449 [2]. All studies attained the maximum of 7 stars on the NOS.
Figure 1.
Flow chart. IPD, individual patient data; RCT, randomized controlled trial; VTE, venous thromboembolism.
Table 1.
Design of studies included in the meta-analysis.
| Study | TiCAT [24] | CATCH [2] | PREDICARE [15] | AXA (not yet published) |
|---|---|---|---|---|
| Country | Spain | International | France | France |
| Sponsor | Hospital Universitario Virgen del Rocío | LEO Pharma (NCT01130025) | LEO Pharma (NCT03099031) | AP-HP (NCT02898051) |
| Inclusion period | January 2009 to September 2015 | August 2010 to November 2013 | May 2015 to April 2016 | August 2011 to November 2015 |
| Design | Prospective cohort study of tinzaparin-treated patients | Tinzaparin arm of an open RCT | Prospective cohort study of tinzaparin-treated patients | Prospective cohort study of tinzaparin-treated patients |
| No. of centers | 3 | 164 | 75 | 8 |
| No. of patients | 247 | 449 | 409 | 308 |
AP-HP, Assistance Publique – Hôpitaux de Paris; RCT, randomized controlled trial.
3.2. Patients
Baseline patient characteristics and the 6-month cumulative incidences of recurrent VTE, major bleeding, and death from any cause are presented by individual study and for the overall study population in Table 2. The median age of the patients was 65.0 years (IQR, 56.0-73.5 years), and 51.9% were females. Regarding fragilities, 21.3% of patients were aged ≥75 years, 9.3% had a body weight of ≤50 kg, and 13.9% had a creatinine clearance <50 mL/min. The qualifying VTE event was symptomatic in 75.0% of patients, 23.0% presenting both DVT and PE. The primary tumor site was gastrointestinal in 27.0% of patients, pulmonary in 18.0%, genitourinary in 14.1%, gynecologic in 14.0%, breast in 11.7%, hematologic in 7.4%, and elsewhere in 7.8%. The cancer was metastatic in 61.7% of patients.
Table 2.
Baseline characteristics of patients with CAT enrolled in each study and in the overall meta-analysis population.
| Characteristic | TiCAT [24] (n = 247) | CATCH [2] (n = 449) | PREDICARE [15] (n = 409) | AXA (n = 308) | Total (n = 1413) |
|---|---|---|---|---|---|
| Age (y), median (IQR) | 64.8 (56.0-72.5) | 60.0 (52.0-69.0) | 68.3 (59.4-76.7) | 67.5 (59.9-75.7) | 65.0 (56.0-73.5) |
| Age ≥ 75 y | 42 (17.1) | 50 (11.1) | 123 (30.1) | 86 (27.9) | 301 (21.3) |
| Female gender | 112 (45.3) | 262 (58.4) | 199 (48.7) | 160 (51.9) | 733 (51.9) |
| BMI (kg/m2), median (IQR) | NA | 24.8 (21.7-28.6) | 24.5 (21.7-28.0) | 24.9 (22.0-28.1) | 24.7 (21.9-28.3) |
| Body weight ≤ 50 kg | 4 (1.9) | 67 (14.9) | 33 (8.1) | 25 (8.1) | 129 (9.3) |
| Creatinine clearance < 50 mL/min | 17 (7.8) | 66 (14.9) | 57 (14.5) | 49 (15.9) | 189 (13.9) |
| History of previous VTE | NA | 27 (6.0) | 49 (12.0) | 61 (19.8) | 137 (11.7) |
| Qualifying thrombotic event | |||||
| DVT only | 128 (51.8) | 252 (57.1) | 162 (39.6) | 64 (20.8) | 606 (43.2) |
| PE only | 75 (30.4) | 48 (10.9) | 175 (42.8) | 177 (57.7) | 475 (33.8) |
| DVT and PE | 44 (17.8) | 141 (32.0) | 72 (17.6) | 66 (21.5) | 323 (23.0) |
| Symptomatic event | 168 (68.0) | 441 (100.0) | 271 (66.3) | 174 (56.5) | 1054 (75.0) |
| ECOG 0-1 | 206 (85.1) | 343 (76.4) | 239 (58.7) | 146 (60.8) | 934 (69.8) |
| Primary tumor site | |||||
| Gastrointestinala | 45 (18.2) | 125 (27.8) | 108 (26.4) | 104 (33.8) | 382 (27.0) |
| Upper | 6 (2.4) | 29 (6.5) | 15 (3.7) | 13 (4.2) | 63 (4.5) |
| Lower | 29 (11.7) | 66 (14.7) | 59 (14.4) | 65 (21.1) | 219 (15.5) |
| Digestive tract | 10 (4.0) | 30 (6.7) | 34 (8.3) | 26 (8.4) | 100 (7.1) |
| Lung | 41 (16.6) | 48 (10.7) | 128 (31.3) | 37 (12.0) | 254 (18.0) |
| Genitourinaryb | 53 (21.5) | 53 (11.8) | 42 (10.3) | 51 (16.6) | 199 (14.1) |
| Gynecologicc | 22 (8.9) | 101 (22.5) | 33 (8.1) | 42 (13.6) | 198 (14.0) |
| Breast | 35 (14.2) | 37 (8.2) | 57 (13.9) | 37 (12.0) | 166 (11.7) |
| Hematological | 20 (8.1) | 44 (9.8) | 22 (5.4) | 18 (5.8) | 104 (7.4) |
| Other sites | 31 (12.6) | 41 (9.1) | 19 (4.6) | 19 (6.2) | 110 (7.8) |
| Metastatic cancer | 161 (65.7) | 247 (55.0) | 268 (67.0) | 179 (61.3) | 855 (61.7) |
| Cancer treatment | |||||
| Chemotherapy | 154 (66.1) | 133 (29.6) | 263 (64.3) | 199 (64.8) | 749 (53.6) |
| Radiotherapy | 11 (7.1) | 51 (11.4) | 48 (11.7) | 21 (6.8) | 131 (9.9) |
| Recurrent VTE | 11 | 31 | 28 | 15 | 85 |
| Cumulative incidence (%) (95% CI)d | 4.6 (2.6-8.0) | 7.2 (5.0-10.3) | 7.3 (5.3-10.2) | 5.0 (2.9-8.3) | 6.2 (5.0-7.7) |
| Major bleeding | 7 | 13 | 15 | 12 | 47 |
| Cumulative incidence (%) (95% CI)d | 3.0 (1.5-6.0) | 3.0 (1.6-5.7) | 3.7 (2.3-6.1) | 3.9 (2.2-7.0) | 3.4 (2.7-4.5) |
| All-cause deaths | 38 | 156 | 144 | 88 | 426 |
| Cumulative incidence (%) (95% CI)e | 16.2 (11.5-20.9) | 36.4 (31.8-41.0) | 36.2 (31.3-41.0) | 29.4 (24.2-34.6) | 31.3 (28.9-33.8) |
Values are n (%) unless stated otherwise.
BMI, body mass index; ECOG, Eastern Cooperative Oncology Group; NA, not available; VTE, venous thromboembolism.
Colorectal, pancreas, stomach, biliary tract, esophagus, small intestine, peritoneum, jejunum, and ampulla of Vater.
Prostate, kidney, bladder, testicular, penile, and ureter.
Ovary, cervix, uterus, endometrium, vagina, vulva, fallopian tube, and female reproductive system.
Estimated by the Kalbfleisch and Prentice method taking into account the competing risk of death.
Estimated by the Kaplan-Meier method.
The 6-month cumulative incidence of recurrent VTE varied from 4.6% in TiCAT to 7.3% in PREDICARE (Supplementary Figure 1A). Overall, recurrent VTE occurred in 85 patients, yielding a 6-month cumulative incidence of recurrent VTE of 6.2% (95% CI, 5.0-7.7). Regarding major bleeding, the 6-month cumulative incidence of major bleeding varied from 3.0% in the TiCAT and CATCH studies to 3.9% in the AXA study (Supplementary Figure 1B). Overall, major bleeding occurred in 47 patients, yielding a 6-month cumulative incidence of 3.4% (95% CI, 2.7-4.5). A total of 426 patients died, mainly from the underlying cancer (31.3%; 95% CI, 28.9-33.8) (Supplementary Figure 1C).
3.3. Performance of the Ottawa score
Data for the variables in the multivariate model used to calculate the Ottawa score were available for 1088 of the 1413 patients included, as data on tumor stage and a personal history of VTE were not available in the TiCAT study. The c-statistic associated with this model was 0.61 (95% CI, 0.54-0.67) (Figure 2).
Figure 2.
Wolbers c-statistic associated with the 3 models.
The Ottawa score classified 442 patients (40.6%) as at low risk of recurrence (Ottawa score < 1), with an estimated 6-month cumulative recurrent VTE incidence of 5.0% (95% CI, 3.2-7.8), and 646 patients (59.4%) as at high risk of recurrence with a 6-month cumulative recurrent VTE incidence of 8.5% (95% CI, 6.6-10.8) (Figure 3). The area under the ROC curve was 0.56 (95% CI, 0.51-0.62). The sensitivity, specificity, and negative and positive predictive values were 72.8%, 41.9%, 8.6%, and 95.3%, respectively (Table 3).
Figure 3.
Cumulative incidence of recurrent venous thromboembolism according to the risk scores in each study and in the overall meta-analysis population.
Table 3.
Performance of the Ottawa score, enhanced score, and alternative score.
| Score category | Ottawa score (score range, −3 to 3) | Enhanced score (score range, −4 to 5) | Alternative score (score range, −3 to 2) |
|---|---|---|---|
| Low risk of recurrence (score) | |||
| Cutoff value | <1 | ≤1 | ≤0 |
| No. of patients | 442 | 673 | 665 |
| Cumulative incidence (%) (95% CI)a | 5.0 (3.2-7.8) | 4.8 (3.3-6.9) | 4.3 (3.0-6.2) |
| High risk of recurrence (score) | |||
| Cutoff value | ≥1 | ≥2 | 1 |
| No. of patients | 646 | 375 | 430 |
| Cumulative incidence (%) (95% CI)a | 8.5 (6.6-10.8) | 11.6 (8.4-16.0) | 9.3 (6.8-12.8) |
| Very high risk of recurrence (score) | |||
| Cutoff value | — | — | 2 |
| No. of patients | — | — | 25 |
| Cumulative incidence (%) (95% CI)a | — | — | 27.4 (14.9-50.6) |
| Performance of the score | |||
| AUC (95% CI) | 0.56 (0.51-0.62) | 0.59 (0.53-0.65) | 0.59 (0.52-0.65) |
| Sensitivity (95% CI) | 72.8 (62.6-83.0) | 59.1 (47.8-70.4) | 9.6 (2.8-16.4) |
| Specificity (95% CI) | 41.9 (37.8-45.9) | 61.2 (57.2-65.2) | 97.7 (96.5-98.9) |
| Positive predictive value (95% CI) | 8.6 (6.4-10.8) | 10.3 (7.3-13.3) | 22.7 (6.8-38.7) |
| Negative predictive value (95% CI) | 95.3 (93.3-97.4) | 95.2 (93.5-96.9) | 93.9 (92.4-95.3) |
AUC, area under the curve.
Estimated by the Kalbfleisch and Prentice method to take into account the competing risk of death.
3.4. Attempts to enhance the Ottawa score
In a multivariate analysis adjusted for the variables included in the Ottawa score (Table 4), the following additional factors were independently associated with the risk of recurrent VTE: proximal DVT at inclusion (subHR, 2.01; 95% CI, 1.18-3.43), surgery or trauma in the previous month (subHR, 2.51; 95% CI, 1.22-5.14), and colorectal cancer (subHR, 0.34; 95% CI, 0.13-0.85). Bootstrapping analysis confirmed these results. The c-statistic of this enhanced model adjusted for the variables included in the Ottawa score was 0.64 (95% CI, 0.58-0.71) (Figure 2).
Table 4.
Results of the enhanced analysis.
| Variables included in the enhanced model | Regression coefficients (SD) | subHR (95% CI) | P | Scoring |
|---|---|---|---|---|
| Female gendera | — | — | — | +1 |
| Tumor stage Ia | — | — | — | −2 |
| Lung cancera | — | — | — | +1 |
| Breast cancera | — | — | — | −1 |
| Personal history of VTEa | — | — | — | +1 |
| Proximal DVT at inclusion | 0.70 (0.27) | 2.01 (1.18-3.43) | .01 | +1 |
| Surgery or trauma (in the previous month) | 0.92 (0.37) | 2.51 (1.22-5.14) | .01 | +1 |
| Colorectal cancer | −1.09 (0.47) | 0.34 (0.13-0.85) | .02 | −1 |
DVT, deep vein thrombosis; subHR, subdistribution hazard ratio; VTE, venous thromboembolism.
Variables included in the Ottawa score.
The score developed from this model ranged from -4 to 5 points and was dichotomized as follows: patients with a score of ≤1 were considered at low to moderate risk of VTE recurrence, and patients with a score of ≥2 were considered at high to very high risk of VTE recurrence. The 6-month cumulative incidence of recurrent VTE was 4.8% (95% CI, 3.3-6.9) in the low- to moderate-risk group and 11.6% (95% CI, 8.4-16.0) in the high- to very high–risk group (Figure 3). The predictive performance of this enhanced score was no better than that of the original Ottawa score (Table 3).
3.5. Attempts to develop an alternative predictive model
It was assumed that data on patients from all 4 studies could contribute to the development of an alternative predictive model. In the multivariate analysis (Table 5), the following factors were independently associated with the risk of recurrent VTE: surgery or trauma during the previous month (subHR, 2.55; 95% CI, 1.27-5.14), proximal DVT at inclusion (subHR, 1.90; 95% CI, 1.13-3.20), lower gastrointestinal cancer (mainly colorectal cancer; subHR, 0.36; 95% CI, 0.15-0.83), and breast cancer (subHR, 0.10; 95% CI, 0.01-0.72). Bootstrapping analysis confirmed these results. The c-statistic of this model was 0.61 (95% CI, 0.55-0.68) (Figure 2).
Table 5.
Results of the alternative predictive model.
| Variables included in the alternative model | Regression coefficient (SD) | subHR (95% CI) | P | Scoring |
|---|---|---|---|---|
| Proximal DVT at inclusion | 0.64 (0.27) | 1.90 (1.13-3.20) | .02 | +1 |
| Surgery or trauma (in the previous month) | 0.94 (0.35) | 2.55 (1.27-5.14) | .009 | +1 |
| Breast cancer | −2.30 (1.01) | 0.10 (0.01-0.72) | .02 | −2 |
| Lower GI cancera | −1.03 (0.43) | 0.36 (0.15-0.83) | .02 | −1 |
DVT, deep vein thrombosis; subHR, subdistribution hazard ratio.
Excluding esophagus, stomach, and pancreas.
The alternative score developed from this model ranged from −3 to 2 points and comprised 3 categories, as follows: patients with a score of ≤0 were considered at low to moderate risk of recurrence of VTE, those with a score of 1 were considered at high risk of recurrence of VTE, and those with a score of 2 were considered at very high risk of recurrence of VTE. The 6-month cumulative incidence of recurrent VTE was 4.3% (95% CI, 3.0-6.2) in the low- to moderate-risk group, 9.3% (95% CI, 6.8-12.8) in the high-risk group, and 27.4% (95% CI, 14.9-50.6) in the very high–risk group (Figure 3). The predictive performance of this alternative score was no better than that of the original Ottawa score (Table 3).
4. Discussion
In this individual patient–data meta-analysis including 1413 patients with CAT treated with tinzaparin, the Ottawa score failed to accurately predict recurrent VTE. Although this score was able to stratify patients by risk of recurrent VTE, showing a lower cumulative incidence in low-risk patients—that is, with an Ottawa score of <1—than that in high-risk patients (5.0% vs 8.5%), its overall predictive performance was poor, with an area under the ROC curve of only 0.56. This study also failed to identify alternative significant clinical factors that could help better predict the risk of recurrent VTE in patients with CAT.
After this study was registered in PROSPERO, 2 meta-analyses aiming to validate the Ottawa score were published. An unregistered meta-analysis involving patients with CAT treated with various low-molecular-weight heparins failed to validate the Ottawa score [13]. The second meta-analysis concluded that the original Ottawa score effectively stratified the risk of recurrent VTE, concurring with findings from the present meta-analysis [14]. However, neither meta-analysis assessed the predictive performance of this score. Importantly, substantial heterogeneity was observed both between and within the individual studies incorporated, with regard to inclusion criteria, types of anticoagulant and anticancer treatments, study design, and outcomes. The performance of the Ottawa score in a more homogeneous and contemporary patient population therefore remained unexplored, justifying the present analysis. This meta-analysis, focusing exclusively on prospective studies using a standardized tinzaparin regimen and independent central outcome adjudication, aimed to address the key limitations of previous validation studies, particularly the heterogeneity in treatment protocols and diagnostic criteria. However, even under these more controlled and homogeneous conditions, the Ottawa score demonstrated only modest discriminatory ability. This suggests that its limited performance is not merely attributable to methodological variability but rather reflects an intrinsic limitation of the score itself in reliably identifying patients truly at high risk of recurrence.
Nevertheless, the negative predictive value in our cohort was 95.3%, indicating that patients classified as low risk (Ottawa score < 1) had a very low likelihood of recurrent VTE within the first 6 months. Although the sensitivity and specificity of the score were limited, its high negative predictive value, partly due to the relatively low (<10%) incidence of recurrent VTE even among high-risk patients, may still provide reassurance in identifying patients at lower risk of recurrence while on anticoagulation. From a clinical perspective, the Ottawa score is not intended to guide the discontinuation of anticoagulation, which remains recommended for all patients with CAT for at least 6 months. Nevertheless, it may still provide useful information for risk assessment and follow-up strategies during ongoing therapy.
To enhance the predictive accuracy of this score, we tested additional clinical variables associated with the risk of thrombosis. Although factors such as recent surgery or trauma and proximal DVT increased the risk of VTE recurrence, whereas colorectal or breast cancer were linked to a lower risk, integrating these factors into the enhanced and alternative scores did not substantially improve performance (c-statistics < 0.65). The alternative model stratified risk modestly, identifying a very high–risk group, but its overall clinical utility remains limited. Notably, recent surgery or trauma was independently associated with recurrence, challenging the common view of surgery as a short-term risk factor. In cancer patients, surgery may lead to prolonged prothrombotic conditions due to factors such as extended immobility, central venous access, treatment delays, and systemic inflammation, possibly justifying longer anticoagulation in this population [25,26].
These findings highlight the complexity of predicting recurrent VTE in cancer patients. The heterogeneity of cancer types, evolving anticancer treatments, and dynamic clinical states likely limit the utility of static baseline risk scores. Our results support the need for individualized assessments rather than reliance on current predictive tools. Future models may benefit from incorporating dynamic or time-varying variables, such as treatment response, biomarkers, or imaging findings, possibly using machine learning techniques on large prospective data sets to account for complex interactions. Pending the development of more accurate tools, caution is warranted in using the Ottawa score to guide management decisions in patients with CAT.
Our study has several strengths, including a large, homogeneous population of prospectively recruited patients, rendered possible by effective data sharing and collaborations. The inclusion of all eligible studies underscores the invaluable contribution of data sharing in contemporary clinical research. However, several limitations should be noted. First, data collection was not standardized across the different cohorts. Variability in definitions, data quality, and completeness may have introduced misclassification or residual confounding, particularly for key clinical variables. Notably, race and ethnicity data were available in only 1 of the included data sets and could not be included in the pooled analyses. This heterogeneity, despite harmonization efforts, may have impacted the reliability of predictive modelling. Second, the score itself has limitations: while it is applicable to breast and lung cancer, it does not apply to the full spectrum of tumor sites, despite the known differences in the risk of recurrent VTE by cancer type [27]. Third, only 3 studies contributed to validation of the enhanced and alternative scores as data on recent surgery or trauma were not collected in the TiCAT study. However, this was not necessarily a power issue, as increased power would increase precision rather than performance. Importantly, TiCAT patients contributed substantially to the analysis of other predictors of recurrence. Finally, although the original literature search was conducted in December 2023, an updated search covering the period from December 2023 to October 2025 did not yield any additional eligible studies.
5. Conclusion
Despite the use of individual data from 4 large, recent, prospective cohorts of patients with CAT treated with tinzaparin, this meta-analysis failed to demonstrate the clinical relevance of the Ottawa score in predicting VTE recurrence over the first 6 months of anticoagulant therapy. As high VTE recurrence rates persist despite anticoagulation in patients with CAT, further research is needed to identify more accurate and personalized predictors of recurrence, ideally in even more homogeneous patient populations, taking into account cancer type and treatment.
Acknowledgments
The authors thank the sponsors of all studies included in the meta-analyses for agreeing to share their data. The authors are grateful to Paula Harry for her valuable assistance with the English language revision of this manuscript, which was supported by the University Hospital of Saint-Étienne.
Funding
This study was funded by LEO Pharma. The funding source had no role in the study design; data collection, analysis, or interpretation; writing of the report; or the decision to submit this for publication. The corresponding author had full access to all data in the study and had final responsibility to submit the manuscript for publication.
Author contributions
All authors contributed substantially to this study. C.C.: Conceptualization, Data curation, Formal analysis, Visualization, Methodology, Project administration, Software, Writing – original draft, Writing – review and editing. P.G.: Conceptualization, Writing – original draft, Writing – review and editing. L.J.-P.: Investigation, Resources, Writing – review and editing. A.Y.Y.L.: Investigation, Resources, Writing – review and editing. O.S.: Investigation, Resources, Writing – review and editing. G.M.: Conceptualization, Methodology, Funding acquisition, Investigation, Resources. G.P.: Writing – review and editing. P.M.: Conceptualization, Writing – review and editing. I.M.: Conceptualization, Writing – original draft, Writing – review and editing. S.L.: Conceptualization, Data curation, Funding acquisition, Methodology, Project administration, Supervision, Validation, Writing – original draft, Writing – review and editing. All authors approved the manuscript prior to its submission for publication.
Relationship disclosure
L.J-P. reports speaking fees or honoraria, outside this work from Janssen, Sanofi Aventis, Viatris Pharmaceuticals SL, Boston Scientific Ibérica SA, Laboratorios Farmacéuticos Rovi SA, Chiesi España S.A.U., Bristol Myers Squibb SA, Len Mayo International SA, Daiichi Sankyo España SA, and Merck Sharp & Dohme De España SA and grant received to the hospital from ROVI and Len Mayo outside of this project. O.S. received research grant support from Bristol Myers Squib, Pfizer, Merck Sharp & Dohme, Boehringer Ingelheim, Boston Scientifics, Inari, Sanofi Aventis, and LEO Pharma and fees or nonfinancial support for consultancy activities from Bristol Myers Squib, Pfizer, Merck Sharp & Dohme, Boston Scientifics, Inari, Sanofi Aventis, GlaxoSmithKline, Boehringer Ingelheim, Viatris, and Janssen. G.P. received fees or funding for participation as a speaker at conferences from BMS, Pfizer, Bayer, LEO Pharma, and Viatris and for clinical trials as a co-investigator sponsored by Anthos, BMS, Pfizer, Bayer, and LEO Pharma. P.M. is a consultant to LEO Pharma AS, Pfizer, Sanofi, and Viatris. I.M. reports participation in boards of Bayer, Bristol Myers Squibb, Pfizer, LEO Pharma, and Sanofi; is a speaker for Bristol Myers Squibb/Pfizer alliance; reports grants from APHP (clinical trial about extended treatment for VTE in patients with CAT). S.L. reports education for employees of Eli Lilly and Boehringer Ingelheim; is a consultant for Ferring; and is a speaker for Pfizer. The other authors report no competing interests to disclose.
Footnotes
Handling editor: Dr Lara Roberts
The online version contains supplementary material available at https://doi.org/10.1016/j.rpth.2025.103278.
Supplementary material
References
- 1.Key N.S., Khorana A.A., Kuderer N.M., Bohlke K., Lee A.Y.Y., Arcelus J.I., et al. Venous thromboembolism prophylaxis and treatment in patients with cancer: ASCO clinical practice guideline update. J Clin Oncol Off. 2020;38:496–520. doi: 10.1200/JCO.19.01461. [DOI] [PubMed] [Google Scholar]
- 2.Lee A.Y.Y., Kamphuisen P.W., Meyer G., Bauersachs R., Janas M.S., Jarner M.F., et al. Tinzaparin vs warfarin for treatment of acute venous thromboembolism in patients with active cancer: a randomized clinical trial. JAMA. 2015;314:677–686. doi: 10.1001/jama.2015.9243. [DOI] [PubMed] [Google Scholar]
- 3.Lee A.Y.Y., Levine M.N., Baker R.I., Bowden C., Kakkar A.K., Prins M., et al. Low-molecular-weight heparin versus a coumarin for the prevention of recurrent venous thromboembolism in patients with cancer. N Engl J Med. 2003;349:146–153. doi: 10.1056/NEJMoa025313. [DOI] [PubMed] [Google Scholar]
- 4.Young A.M., Marshall A., Thirlwall J., Chapman O., Lokare A., Hill C., et al. Comparison of an oral factor Xa inhibitor with low molecular weight heparin in patients with cancer with venous thromboembolism: results of a randomized trial (SELECT-D) J Clin Oncol. 2018;36:2017–2023. doi: 10.1200/JCO.2018.78.8034. [DOI] [PubMed] [Google Scholar]
- 5.Raskob G.E., van Es N., Verhamme P., Carrier M., Di Nisio M., Garcia D., et al. Edoxaban for the treatment of cancer-associated venous thromboembolism. N Engl J Med. 2018;378:615–624. doi: 10.1056/NEJMoa1711948. [DOI] [PubMed] [Google Scholar]
- 6.Agnelli G., Becattini C., Meyer G., Muñoz A., Huisman M.V., Connors J.M., et al. Apixaban for the treatment of venous thromboembolism associated with cancer. N Engl J Med. 2020;382:1599–1607. doi: 10.1056/NEJMoa1915103. [DOI] [PubMed] [Google Scholar]
- 7.McBane R.D., Wysokinski W.E., Le-Rademacher J.G., Zemla T., Ashrani A., Tafur A., et al. Apixaban and dalteparin in active malignancy-associated venous thromboembolism: the ADAM VTE trial. J Thromb Haemost. 2020;18:411–421. doi: 10.1111/jth.14662. [DOI] [PubMed] [Google Scholar]
- 8.Louzada M.L., Carrier M., Lazo-Langner A., Dao V., Kovacs M.J., Ramsay T.O., et al. Development of a clinical prediction rule for risk stratification of recurrent venous thromboembolism in patients with cancer-associated venous thromboembolism. Circulation. 2012;126:448–454. doi: 10.1161/CIRCULATIONAHA.111.051920. [DOI] [PubMed] [Google Scholar]
- 9.Bosch F.T.M., van Es N., Di Nisio M., Carrier M., Segers A., Grosso M.A., et al. The Ottawa score does not predict recurrent venous thromboembolism in cancer patients: results from the Hokusai-VTE cancer study [abstract] Res Pract Thromb Haemost. 2019;3(Suppl 1):717–718. [Google Scholar]
- 10.Louzada M.L., Bose G., Cheung A., Chin-Yee B.H., Wells S., Wells P.S. Predicting venous thromboembolism recurrence risk in patients with cancer: a validation study. Blood. 2012;120:394. [Google Scholar]
- 11.Ahn S., Lim K.S., Lee Y.S., Lee J.L. Validation of the clinical prediction rule for recurrent venous thromboembolism in cancer patients: the Ottawa score. Support Care Cancer. 2013;21:2309–2313. doi: 10.1007/s00520-013-1792-9. [DOI] [PubMed] [Google Scholar]
- 12.van Es N., Louzada M., Carrier M., Tagalakis V., Gross P.L., Shivakumar S., et al. Predicting the risk of recurrent venous thromboembolism in patients with cancer: a prospective cohort study. Thromb Res. 2018;163:41–46. doi: 10.1016/j.thromres.2018.01.009. [DOI] [PubMed] [Google Scholar]
- 13.Frere C., Crichi B., Wahl C., Lesteven E., Connault J., Durant C., et al. The Ottawa score performs poorly to identify cancer patients at high risk of recurrent venous thromboembolism: insights from the TROPIQUE study and updated meta-analysis. J Clin Med. 2022;11:3729. doi: 10.3390/jcm11133729. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Delluc A., Miranda S., den Exter P., Louzada M., Alatri A., Ahn S., et al. Accuracy of the Ottawa score in risk stratification of recurrent venous thromboembolism in patients with cancer-associated venous thromboembolism: a systematic review and meta-analysis. Haematologica. 2020;105:1436–1442. doi: 10.3324/haematol.2019.222828. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Girard P., Laporte S., Chapelle C., Falvo N., Falchero L., Cloarec N., et al. Failure of the Ottawa score to predict the risk of recurrent venous thromboembolism in cancer patients: the prospective PREDICARE cohort study. Thromb Haemost. 2022;122:151–157. doi: 10.1055/a-1486-7497. [DOI] [PubMed] [Google Scholar]
- 16.Stewart L.A., Clarke M., Rovers M., Riley R.D., Simmonds M., Stewart G., et al. Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data: the PRISMA-IPD Statement. JAMA. 2015;313:1657–1665. doi: 10.1001/jama.2015.3656. [DOI] [PubMed] [Google Scholar]
- 17.Wells G.A., Wells G., Shea B., Peterson J.E., Welch V., Losos M., et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2014. https://api.semanticscholar.org/CorpusID:79550924
- 18.Wolbers M., Koller M.T., Witteman J.C.M., Steyerberg E.W. Prognostic models with competing risks: methods and application to coronary risk prediction. Epidemiol Camb Mass. 2009;20:555–561. doi: 10.1097/EDE.0b013e3181a39056. [DOI] [PubMed] [Google Scholar]
- 19.Frere C., Crichi B., Rueda-Camino J.A., Cajfinger F., Spiess N., Janus N., et al. Long-term use of tinzaparin for the treatment of cancer-associated thrombosis in clinical practice: insights from the prospective TROPIQUE study. J Med Vasc. 2022;47:56–64. doi: 10.1016/j.jdmv.2022.04.004. [DOI] [PubMed] [Google Scholar]
- 20.Papakotoulas P., Tsoukalas N., Christopoulou A., Ardavanis A., Koumakis G., Papandreou C., et al. Management of cancer-associated thrombosis (CAT): symptomatic or incidental. Anticancer Res. 2020;40:305–313. doi: 10.21873/anticanres.13954. [DOI] [PubMed] [Google Scholar]
- 21.Debourdeau P., Arvers P., Hij A., Bennani H., Desauw C., Falvo N., et al. Factors influencing patient’s perception of long-term treatment with low-molecular-weight heparins for cancer-associated thrombosis: an updated analysis of TROPIQUE, a prospective observational study. Support Care Cancer. 2020;28:287–293. doi: 10.1007/s00520-019-04815-z. [DOI] [PubMed] [Google Scholar]
- 22.Piatek C.I., Tagawa S.T., Wei-Tsai D., Hanna D., Weitz I.C., O'Connell C., et al. OC-14—baseline D-dimer levels are predictive of recurrent venous thromboembolism (VTE) at 6 months in cancer patients with VTE treated with tinzaparin. Thromb Res. 2016;140(Suppl 1) doi: 10.1016/S0049-3848(16)30131-1. [DOI] [PubMed] [Google Scholar]
- 23.Hull R.D., Pineo G.F., Brant R.F., Mah A.F., Burke N., Dear R., et al. Long-term low-molecular-weight heparin versus usual care in proximal-vein thrombosis patients with cancer. Am J Med. 2006;119:1062–1072. doi: 10.1016/j.amjmed.2006.02.022. [DOI] [PubMed] [Google Scholar]
- 24.Jara-Palomares L., Solier-Lopez A., Elias-Hernandez T., Asensio-Cruz M., Blasco-Esquivias I., et al. Tinzaparin in cancer associated thrombosis beyond 6months: TiCAT study. Thromb Res. 2017;157:90–96. doi: 10.1016/j.thromres.2017.07.004. [DOI] [PubMed] [Google Scholar]
- 25.Schmidt R.A., Al Zaki A., Desilet N., Szefer E., Ratnaweera N., Peterson E., et al. Patient characteristics and long-term outcomes beyond the first 6 months after a diagnosis of cancer-associated venous thromboembolism. Thromb Res. 2020;188:106–114. doi: 10.1016/j.thromres.2020.02.005. [DOI] [PubMed] [Google Scholar]
- 26.Lapébie F.X., Bura-Rivière A., Espitia O., Bongard V., Ciammaichella M.M., Martínez J.G., et al. Predictors of recurrence of cancer-associated venous thromboembolism after discontinuation of anticoagulant therapy: a multicenter cohort study. J Thromb Haemost. 2023;21:2189–2201. doi: 10.1016/j.jtha.2023.04.010. [DOI] [PubMed] [Google Scholar]
- 27.Mahe I., Chapelle C., Jara-Palomares L., Sanchez O., Girard P., Meyer G., et al. Efficacy and safety outcomes in patients with cancer-associated thromboembolism treated with tinzaparin according to the tumor site: meta-analysis of prospective studies involving 1413 individual patients’ data. Blood. 2023;142(Suppl 1):2646. doi: 10.1182/blood-2023-177979. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.