Background
Use of risk assessment tools, such as the Caprini score or Rogers score, is recommended by national societies to stratify surgical patients by venous thromboembolism risk and guide prophylaxis. However, these tools were not developed in a gynecological oncology patient population, and their utility in this population is unknown.
Objective
The objective of the study was to examine the ability of both the Caprini and Rogers scores to stratify gynecological oncology patients by the risk of venous thromboembolism.
Study Design
Patients undergoing surgery for cervical, ovarian, uterine, vaginal, and vulvar cancers between 2008 and 2013 were identified from the National Surgical Quality Improvement Program Database using International Classification of Diseases , ninth revision, codes. The Caprini and Rogers scores were calculated for each patient based on the recorded demographic and procedure data. Venous thromboembolism events were recorded for 30 days postoperatively. Patients were categorized into risk groups based on the calculated Caprini and Rogers scores and the incidence of venous thromboembolism, and the 95% confidence interval was estimated for each of these groups. The relationship between the risk score and venous thromboembolism incidence was examined with Pearson’s correlation coefficient.
Results
Of 17,713 patients, 1.8% developed a venous thromboembolism. No patients were classified by the Caprini score as low risk, 0.1% were moderate risk, 3.0% were higher risk (score 4), and 96.9% were highest risk (score ≥5). The Caprini score groupings did not correlate with venous thromboembolism. The high-risk group had a paradoxically higher incidence of venous thromboembolism of 2.5% compared with the highest-risk group, 1.7% ( P = .40). However, when the highest-risk group of the Caprini score was substratified, it was highly correlated with venous thromboembolism (R 2 = 0.93). For the Rogers score, only 0.2% of patients were low risk (score <7), 36.9% were medium risk (score 7–10), and 63.0% were high risk (score >10). When the highest risk group of the Rogers score was substratified, it was also highly correlated with venous thromboembolism (R 2 = 0.99).
Conclusion
Gynecological oncology patients score very high on current venous thromboembolism risk assessment models. The Caprini score is limited in its ability to discriminate relative venous thromboembolism risk among gynecological oncology patients because 97% are in the highest-risk category. Substratification of the highest-risk groups allows for relative venous thromboembolism risk stratification among gynecological oncology patients, suggesting that further evaluation of risk stratification is needed in gynecological oncology surgery.
Venous thromboembolism is the second most common complication and third most common cause of excess mortality after surgery. Patients with gynecological cancer are at an increased risk of thromboembolism.
National organizations, such as the American College of Chest Physicians, have published guidelines for the appropriate use of postoperative mechanical and pharmacological prophylaxis. Within these guidelines, risk stratification is used to determine the appropriate degree of prophylaxis for each individual patient. Those patients at the highest risk (∼6% risk of venous thromboembolism) are recommended to receive dual prophylaxis with both mechanical and pharmacological prophylaxis including extended duration pharmacological prophylaxis for those with pelvic cancers. Populations at low risk (∼1.5% risk of venous thromboembolism) are recommended to receive only mechanical prophylaxis.
The risk stratification tools used in the American College of Chest Physicians guidelines are the Caprini score and the Rogers score. Both the Caprini and the Rogers models assign points to various risk factors for venous thromboembolism and use those point totals to place patients into risk strata.
The Caprini score is the most widely used venous thromboembolism risk assessment tool. The Rogers score was developed from vascular and general surgery cases recorded in a large Veterans Administration quality database and is used less frequently.
The authors of the American College of Chest Physicians guidelines note that although risk stratification models have not been validated in gynecological surgery patients, gynecological surgery patients are likely sufficiently similar to other patients undergoing abdominal and pelvic surgery for extrapolation. However, neither score was developed in a gynecological oncology patient population, and the validity of this extrapolation hypothesis is unknown.
Two studies have examined the use of the Caprini score in gynecological oncology patients. Both found that gynecological oncology patients score high using the Caprini risk assessment model with >92% and >96% falling into the highest risk category, suggesting a limited ability to risk stratify as the vast majority of patients are categorized in a single stratum. However, further examination into the possibility of risk stratification within this highest risk category has not been performed, and the Rogers score has yet to be examined in a gynecological oncology patient population.
Our objective was to examine the utility of the Caprini and Rogers scores to risk stratify gynecological oncology patients undergoing surgery by their risk of postoperative venous thromboembolism.
Materials and Methods
This was a secondary analysis cohort study of prospectively collected surgical quality data. The primary outcome was the association between risk assessment score in 2 different risk assessment models and the incidence of venous thromboembolism. The study population was patients who underwent surgery for cervical, ovarian, uterine, vaginal, and vulvar cancers between 2008 and 2013 who were identified from the American College of Surgeons National Surgical Quality Improvement Program Database using International Classification of Diseases , ninth revision, codes. The Institutional Review Board at the University of North Carolina at Chapel Hill reviewed this study and declared it exempt from formal review because it does not constitute human subjects research.
The American College of Surgeons National Surgical Quality Improvement Program database is a national surgical quality improvement project. Participation is voluntary, and participating institutions are able to track their own risk-adjusted outcomes after surgery. Trained clinical reviewers prospectively collect variables such as patient demographics, operative variables, and postoperative outcomes for each individual procedure for 30 days following surgery. The data are then deidentified of patient-, hospital-, and location-specific information and placed into the National Surgical Quality Improvement Program database. Periodic auditing ensures high-quality data, including that for data points occurring after hospital discharge. Details of the methods of data collection and reliability have been previously reported.
Our primary outcome was venous thromboembolism, which was defined as either a pulmonary embolism or a deep vein thrombosis diagnosed within 30 days postoperatively. Both were defined as per the National Surgical Quality Improvement Program participant use file. Demographic, operative, and disease characteristics were also recorded.
Site of malignancy was defined by postoperative International Classification of Diseases , ninth revision, code. Procedure type was defined as laparotomy, minimally invasive, and external. Patients were placed into these categories based on primary procedure Current Procedural Terminology codes. Minimally invasive procedures were defined as laparoscopic or vaginal approaches with abdominal cavity entry, such as a vaginal hysterectomy or operative laparoscopy. External procedures were defined as procedures in which the abdominal cavity was not entered, such as vulvar/vaginal resections, or cervical excisional procedures. The Charlson comorbidity score was calculated for each patient as previously described.
Surgical complexity was defined by the work-relative value unit, which is an estimate of the amount of physician work per Current Procedural Terminology code. The work-relative value unit for each procedure is the sum of the assigned value to each Current Procedural Terminology code for the procedure; thus, the higher work-relative value unit is associated with increasing surgical complexity.
A Caprini score and a Rogers score were calculated for each patient by assigning points to each risk factor present for a given patient. The Caprini score model was used to calculate a Caprini risk score for each patient based on the variables available in the National Surgical Quality Improvement Program database ( Supplemental Table 1 ).
Risk factors that are assigned points in the Caprini score, but are unavailable in National Surgical Quality Improvement Program, include swollen legs, varicose veins, a history of unexplained abortions (>3), the use of hormonal contraceptives or replacement, a history of inflammatory bowel disease, central venous access, a history of venous thromboembolism or a family history venous thromboembolism, and congenital or acquired thrombophilias. Data were missing for less than 5% of patients for all available risk factors in the Caprini model.
The Rogers score model was used to calculate a Rogers risk score for each patient based on the variables available in the National Surgical Quality Improvement Program database ( Supplemental Table 2 ). All risk factors included in the Rogers score are available in National Surgical Quality Improvement Program. For the Rogers model, data were missing for less than 5% for all risk factors with the exception of serum albumin and bilirubin, which were missing for ∼40% of patients. Patients missing these laboratory values were given zero points for that risk factor. Because there are 18 measured risk factors in the Rogers score worth 1–4 points and both hyperbilirubinemia and hypoalbuminemia are worth only 1 point and are relatively rare in the population (17% and 5%, respectively), having the additional data for these 2 variables should not alter the total Rogers score sufficiently to change our conclusions.
Data regarding mechanical or pharmacological prophylaxis are not available in the National Surgical Quality Improvement Program database. However, Surgical Care Improvement Project guidelines that were implemented in 2008 require the use of venous thromboembolism prophylaxis or documentation of why prophylaxis was not provided. We assume that the compliance with these guidelines was very high (>95%) because of the penalties imposed on hospitals, and thus, the majority of the patients in this cohort likely received venous thromboembolism prophylaxis.
After risk scores were calculated for each patient, patients were categorized into the American College of Chest Physicians risk stratification groups based on their Caprini or Rogers score individually. Patients were also categorized into risk categories by their individual scores. The incidence of venous thromboembolism for each of these risk groups was estimated along with a 95% confidence interval.
The relationship between risk score and venous thromboembolism incidence was examined using a Pearson’s correlation coefficient. For bivariable analysis, 2-tailed Student t tests were used for continuous variables and Pearson’s χ 2 tests for categorical variables. A value of P < .05 was considered significant for all analyses. SPSS version 20.0 (IBM Corp, Armonk, NY) was used for all analyses.
Results
We identified 17,713 patients with the demographic and operative characteristics as listed in Table 1 . The majority of the patients in this cohort had uterine cancer (59.5%) followed by ovarian cancer (27.2%), cervical cancer (8.8%), and vulvar/vaginal (4.5%) cancer. The surgical approach was laparotomy (52.3%), minimally invasive surgery (43.0%), or an external procedure (4.7%). This cohort also represents a group of patients undergoing surgically complex procedures. The median work-relative value unit was 31.5 (interquartile range, 19.3–38.0). For reference, the number of work-relative value units assigned to a total abdominal hysterectomy is 17.3. Therefore, the majority of these procedures were complicated oncological procedures.
| Characteristics | Values |
|---|---|
| Age, y | 62 (53–70) |
| BMI, kg/m 2 | 30.4 (25.1–37.4) |
| Race | |
| White | 13,744 (77.6) |
| Black | 1358 (7.7) |
| Asian | 657 (3.7) |
| American or Alaskan Native | 140 (0.8) |
| Unknown | 1814 (10.2) |
| Site of malignancy | |
| Uterus | 10,543 (59.5) |
| Ovary | 4812 (27.2) |
| Cervix | 1560 (8.8) |
| Vulva/vagina | 793 (4.5) |
| GTN | 5 (0.01) |
| Procedure type | |
| Laparotomy | 9263 (52.3) |
| Minimally invasive | 7624 (43.0) |
| External | 826 (4.7) |
| Charlson comorbidity index score | |
| 0 | 11,923 (67.3) |
| 1 | 2890 (16.3) |
| 2 | 1435 (8.1) |
| 3 or greater | 1465 (8.3) |
| Operating time, min | 159 (112–221) |
| Length of hospital stay, d | 2.5 (1–4) |
| Work-relative value unit | 31.5 (19.3–38.0) |
Of 17,713 patients with gynecological cancer, 1.8% (n = 313) developed venous thromboembolism. One hundred thirty-one patients experienced a pulmonary embolism, 149 experienced a deep-vein thrombosis and 33 experienced both a pulmonary embolism and a deep-vein thrombosis. There were 143 deaths recorded in the 30 day postoperative period, and venous thromboembolism was associated with an increased risk of death. Patients with venous thromboembolism experienced a 4.7% 30 day mortality compared with 0.7% for patients without venous thromboembolism ( P < .001).
On bivariable analysis, the cancer site was associated with venous thromboembolism. Patients with ovarian cancer had the highest venous thromboembolism incidence (3.0%) followed by vulvar/vaginal cancer patients (1.5%), uterine cancer patients (1.3%), and cervical cancer patients (1.2%) ( P < .001).
The surgical approach was also associated with venous thromboembolism. Laparotomy patients had the highest venous thromboembolism incidence (2.7%), followed by patients undergoing external procedures (1.1%) and those undergoing minimally invasive procedures (0.7%) ( P < .001). However, the surgical approach is not factored into the Caprini score; thus, patients undergoing minimally invasive surgery had almost identical mean Caprini scores as patients undergoing laparotomy (6.5 vs 6.4, P = .06).
Rogers scores were significantly lower for patients undergoing minimally invasive surgery as compared with laparotomy, although the magnitude of the difference was small (10.6 vs 11.6, P < .001).
Patients were classified into American College of Chest Physicians Guideline risk groups based first on their Caprini risk score and then by their Rogers risk score ( Table 2 ). These guidelines classify patients undergoing nonorthopedic surgical procedures into 4 groups based on the risk of venous thromboembolism and corresponding Caprini and Rogers scores for each group are given.
| ACCP risk group | Patients categorized by Caprini score | Patients categorized by Rogers score | ||
|---|---|---|---|---|
| Caprini score | Patients in risk group | Rogers score | Patients in risk group | |
| Very low risk | 0 | 0 (0.0) | <7 | 29 (0.2) |
| Low risk | 1–2 | 18 (0.1) | 7-10 | 6532 (36.9) |
| Moderate risk | 3–4 | 527 (3.0) | >10 | 11,152 (63.0) |
| High risk | ≥5 | 17,168 (96.9) | a | a |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
