There were 1241 patients in remission after completion of postoperative chemotherapy included in the posttreatment surveillance analysis and 511 patients who had a doubling of the cancer antigen 125 (CA-125) level and reinitiated treatment included in the time to retreatment analysis.
Patients were censored at the earliest sign of recurrence, including doubling of CA-125 level, documentation of recurrence, initiation of retreatment, last follow-up, and death.
Patients were censored at the earliest sign of recurrence, including doubling of cancer antigen 125 (CA-125) level, documentation of recurrence, initiation of retreatment, last follow-up, and death.
Median time to retreatment was 2.8 months before 2009 and 3.5 months after 2009 (log-rank test, P = .40). Patients were censored at last follow up and death.
Two practice patterns were observed: in group 1 (4 institutions), the time to retreatment was similar both before and after the study by Rustin et al5,6 (log-rank test, P = .70); in group 2 (2 institutions), the time to retreatment was longer after the Rustin et al5,6 study (log-rank test, P = .01). Patients were censored at last follow-up and death.
eTable 1. Patient Characteristics and Unadjusted Associations With Posttreatment CA-125
eTable 2. Patient Characteristics and Unadjusted and Adjusted Associations With Posttreatment CT Scans
eTable 3. Patient Characteristics and Unadjusted Associations With Time to Treatment Following CA-125 Doubling From Posttreatment Nadir
Customize your JAMA Network experience by selecting one or more topics from the list below.
Esselen KM, Cronin AM, Bixel K, et al. Use of CA-125 Tests and Computed Tomographic Scans for Surveillance in Ovarian Cancer. JAMA Oncol. 2016;2(11):1427–1433. doi:10.1001/jamaoncol.2016.1842
Did a 2009 randomized clinical trial demonstrating that cancer antigen 125 (CA-125) tests for routine surveillance in ovarian cancer increase use of chemotherapy and decreased patients’ quality of life without improving overall survival change current use of these tests?
In this multi-institutional prospective cohort study of women diagnosed with ovarian cancer from 2004 to 2011, there was no change in the use of CA-125 tests or computed tomographic scans following the 2009 randomized clinical trial.
CA-125 tests and CT scans are still routinely used for surveillance testing in ovarian cancer without proven benefit.
A 2009 randomized clinical trial demonstrated that using cancer antigen 125 (CA-125) tests for routine surveillance in ovarian cancer increases the use of chemotherapy and decreases patients’ quality of life without improving survival, compared with clinical observation. The Society of Gynecologic Oncology guidelines categorize CA-125 testing as optional and discourage the use of radiographic imaging for routine surveillance. To date, few studies have examined the use of CA-125 tests in clinical practice.
To examine the use of CA-125 tests and computed tomographic (CT) scans in clinical practice before and after the 2009 randomized clinical trial and to estimate the economic effect of surveillance testing.
Design, Setting, and Participants
A prospective cohort of 1241 women with ovarian cancer in clinical remission after completion of primary cytoreductive surgery and chemotherapy at 6 National Cancer Institute–designated cancer centers between January 1, 2004, and December 31, 2011, was followed up through December 31, 2012, to study the use of CA-125 tests and CT scans before and after 2009. Data analysis was conducted from April 9, 2014, to March 28, 2016.
Main Outcomes and Measures
The use of CA-125 tests and CT scans before and after 2009. Secondary outcomes included the time from CA-125 markers doubling to retreatment among women who experienced a rise in CA-125 markers before and after 2009, and the costs associated with surveillance testing using 2015 Medicare reimbursement rates.
Among 1241 women (mean [SD] age 59  years; 1112 white [89.6%]), the use of CA-125 testing and CT scans was similar during the study period. During 12 months of surveillance, the cumulative incidence of patients undergoing 3 or more CA-125 tests was 86% in 2004-2009 vs 91% in 2010-2012 (P = .95), and the cumulative incidence of patients undergoing more than 1 CT scan was 81% in 2004-2009 vs 78% in 2010-2012 (P = .50). Among women whose CA-125 markers doubled (n = 511), there was no significant difference in the time to retreatment with chemotherapy before and after 2009 (median, 2.8 vs 3.5 months; P = .40). During a 12-month period, there was a mean of 4.6 CA-125 tests and 1.7 CT scans performed per patient, resulting in a US population surveillance cost estimate of $1 999 029 per year for CA-125 tests alone and $16 194 647 per year with CT scans added.
Conclusions and Relevance
CA-125 tests and CT scans are still routinely used for surveillance testing in patients with ovarian cancer, although their benefit has not been proven and their use may have significant implications for patients’ quality of life as well as costs.
The cost of cancer care is rapidly rising,1 prompting physicians and payors to identify low-value practices that increase costs without improving outcomes.2 In 2012, the American Board of Internal Medicine’s Choosing Wisely campaign3 challenged specialists to identify tests and treatments that are overused and do not provide meaningful clinical benefit. In response, the American Society of Clinical Oncology and the Society for Gynecologic Oncology recommended stopping surveillance testing in some asymptomatic patients.4
In the 2009 American Society of Clinical Oncology Annual Meeting plenary session, Rustin et al5,6 presented results from a clinical trial that randomized patients with ovarian cancer in clinical remission to undergo active surveillance with cancer antigen 125 (CA-125) testing or monitoring for symptoms of recurrent disease. There was no difference in survival between the groups, but women who underwent CA-125 testing received more chemotherapy and had poorer quality of life. The authors concluded that early detection and treatment of recurrent ovarian cancer did not improve clinical outcomes and recommended stopping surveillance testing.
Despite this recommendation, the 2015 National Comprehensive Cancer Center Network’s (NCCN) clinical guidelines recommend routine CA-125 testing for patients whose CA-125 levels were elevated at diagnosis and computed tomographic (CT) scans as clinically indicated.7 Society for Gynecologic Oncology recommendations categorize CA-125 testing as optional, and advise against routine radiographic surveillance.8 More recently, as part of the Choosing Wisely campaign, the Society for Gynecologic Oncology recommended avoidance of routine imaging for surveillance of ovarian cancer.4 To date, surveillance testing remains controversial and it is unclear how frequently such testing is performed in patients with ovarian cancer.
The aim of this study was to examine the use of CA-125 surveillance testing at 6 National Cancer Institute–designated cancer centers over time, and to determine whether the rates decreased after the data from the Rustin et al5,6 study were presented in 2009. We hypothesized that rates of CA-125 testing would not decrease, but the time to retreatment would increase, as health care professionals delayed treatment to minimize reductions in patients’ quality of life while monitoring them closely for disease complications. We also examined the use of CT scans, as few multi-institutional studies have described use of such imaging in this population previously, to our knowledge. Finally, we sought to explore the economic effect of current surveillance testing in the treatment of ovarian cancer.
The NCCN Ovarian Cancer Outcomes Database contains data collected prospectively on all patients diagnosed with ovarian, fallopian, or primary peritoneal cancers and treated at 6 institutions between January 1, 2004, and December 31, 2011, including City of Hope Comprehensive Cancer Center, Dana-Farber/Brigham and Women’s Cancer Center, Fox Chase Cancer Center, The Ohio State University Comprehensive Cancer Center, The University of Texas MD Anderson Cancer Center, and the University of Michigan Comprehensive Cancer Center. Patients who received all or part of their treatment at these centers were included; those seen for a single consultation were excluded. Abstraction of medical records was performed to obtain patient characteristics (eg, demographics, comorbidities, and performance status), tumor characteristics (histologic features, grade, and stage), treatments (eg, surgical procedures and chemotherapy), surveillance studies (CA-125 and CT scans), and vital status. Data on race/ethnicity were collected at initial presentation to the NCCN institution and analyzed because of documented disparities in the quality of ovarian cancer care by race.9 Additional audits were undertaken at all 6 institutions before analysis for any patients with missing data on key variables (CA-125 values and CT scans) to further verify the accuracy of the database. The Ohio State University and City of Hope Institutional Review Boards at each center approved the overall project, while the Dana-Farber Cancer Institute, MD Anderson Cancer Center, and University of Michigan Institutional Review Boards waived review. Patients did not provide consent for the study as it involved medical record reviews at 5 centers (Dana-Farber Cancer Institute, The Ohio State University, MD Anderson Cancer Center, University of Michigan, and Fox Chase Cancer Center) that were exempt from patient consent requirements. At City of Hope, patients provided written consent.
We identified 2 cohorts of patients. In cohort 1, we examined the cumulative incidence of CA-125 testing and CT scans for all 1241 patients who presented to an NCCN center between January 1, 2004, and December 31, 2011, with newly diagnosed epithelial ovarian cancer, underwent cytoreductive surgery and chemotherapy, and met the criteria from the study by Rustin et al5,6 for remission, defined as normal CA-125 concentration and no additional chemotherapy or hormonal treatments within 4 months after completion of chemotherapy. We excluded 272 patients enrolled in clinical trials because their surveillance testing would be dictated by the trial protocol. In cohort 2, which included a subset of patients from cohort 1, we examined the time to retreatment with chemotherapy among 511 patients whose CA-125 levels doubled relative to the posttreatment nadir during the surveillance period. Patients were excluded from cohort 2 if they initiated secondary therapy before their CA-125 level doubled (suggesting clinical signs of recurrence; n = 60) or had radiographic evidence of recurrence before their CA-125 level doubled (n = 31). All patients were divided into the following 2 groups: “pre-Rustin,” diagnosed between 2004 and 2009, and “post-Rustin,” diagnosed between 2010 and 2011. Data were collected on all patients through December 31, 2012.
The primary outcome was the use of CA-125 testing and CT scans in the first year after remission (cohort 1). We also examined patient factors associated with testing. Secondary outcomes included time to retreatment with chemotherapy or radiation after a documented doubling of CA-125 levels (cohort 2) and the costs associated with surveillance testing using 2015 Medicare reimbursement rates.
The primary independent variable of interest was era of diagnosis (pre-Rustin vs post-Rustin). Additional covariates included age, race/ethnicity, Hispanic ethnicity, Gynecologic Oncology Group performance status, Charlson comorbidity score,10 stage, grade, histologic features, CA-125 level at diagnosis, and institution.
Data analysis was conducted from April 9, 2014, to March 28, 2016. Posttreatment CA-125 tests and CT scans were quantified using the cumulative incidence of patients with 1, 2, and 3 or more surveillance tests plotted over time for patients diagnosed in the pre-Rustin and post-Rustin periods (cohort 1). Patients were categorized as in remission after documentation of the first normal CA-125 value within 4 months of completion of primary chemotherapy. Patients were censored at the earliest occurrence of initiation of second-line therapy, doubling of CA-125 level, recurrence, or death or unavailability for follow-up. The time to retreatment after doubling of the CA-125 level was also quantified using the cumulative incidence of retreatment plotted over time (cohort 2).
Unadjusted associations between patient and disease characteristics and the time to posttreatment surveillance and retreatment were assessed using the log-rank test. Multivariable associations were assessed with a Cox proportional hazards regression model including all covariates, except for Gynecologic Oncology Group performance status and race/ethnicity, owing to collinearity with the institution variable. Owing to limited variability in posttreatment surveillance testing with CA-125 tests (95% of patients were tested within 6 months), multivariable associations were not assessed for that outcome.
We performed several sensitivity analyses. First, we examined the cumulative incidence of CA-125 testing, the use of CT scans, and time to retreatment before and after the publication of the study by Rustin et al6 in The Lancet (October 2, 2010), instead of on presentation of the data (June 1, 2009). Second, we examined whether individual institutions responded differently to the Rustin trial by plotting the cumulative incidence curves within institutions by era of diagnosis. Although no substantive difference by era of diagnosis was observed within institutions for surveillance, we observed 2 general practice patterns for time to retreatment. Specifically, there was no evidence of a change in practice after 2009 for 4 institutions, but there was evidence of a change in practice after 2009 for 2 institutions. We then combined institutions with similar empirical practice patterns and performed an interaction analysis to test formally for a change in time to retreatment in one group but not the other.
Costs of routine surveillance tests were estimated from 2015 Medicare reimbursement rates using the following Current Procedural Terminology codes: CA-125 test (code 86304; $28.35), CT scan of the chest with contrast (code 71260; $230.69), and CT scan of the abdomen and pelvis with contrast (code 74177; $314.06).11 To calculate the cost of surveillance testing during a 12-month period, the mean and median number of tests conducted were calculated and 2015 Medicare reimbursement rates applied. A population estimate of women undergoing surveillance testing annually was calculated: 21 290 women were diagnosed with ovarian cancer in 2015, of which approximately 90% were epithelial ovarian cancers; 80% of the patients were expected to achieve clinical remission, resulting in approximately 15 329 women undergoing surveillance.12
P < .05 (2-sided) was considered significant. Statistical analysis was performed using Stata, version 13.1 (StataCorp).
There were 2861 patients diagnosed with ovarian cancer at NCCN institutions between 2004 and 2011. Among these patients, 2082 women underwent cytoreductive surgery and chemotherapy, excluding clinical trial participants, and 1241 women (905 pre-Rustin and 336 post-Rustin; mean [SD] age 59  years; 1112 white [89.6%]) met criteria for clinical remission (Figure 1).
Figure 2 and Figure 3 show the cumulative incidence curves of CA-125 testing and CT scans before and after the presentation of the study by Rustin et al.5 There was no significant difference in surveillance testing in the periods before and after the study by Rustin et al5; 95% vs 96% of patients, respectively, underwent 1 or more CA-125 test within 6 months of remission, and 86% vs 91%, respectively, underwent 3 or more within 12 months. Similarly, 68% vs 64% of patients, respectively, underwent 1 or more CT scan within 6 months, while 30% vs 29%, respectively, underwent 3 or more CT scans within 12 months.
Patient characteristics associated with CA-125 testing are shown in eTable 1 in the Supplement. There was no difference in testing rates among patients who had a CA-125 level of less than 35 U/mL at diagnosis, compared with those who had a CA-125 level of 35 U/mL or more (cumulative incidence, 94% vs 95%, respectively; P = .86) (to convert CA-125 level to kilounits per liter, multiply by 1.0). Institutions differed in their rates of testing at 6 months, although the absolute differences were small (range, 93%-98%; P < .001). CA-125 testing did not vary significantly by diagnosis year (range, 94%-96%; P = .95).
Patient characteristics associated with CT scans after completion of treatment are shown in eTable 2 in the Supplement. Patients diagnosed with advanced-stage disease were more likely to undergo CT scans, compared with patients diagnosed with early-stage disease. There were also significant differences in the use of CT scans among institutions (range at 6 months, 49%-81%; P < .001). However, the use of CT scans did not differ significantly by year (P = .29). We found no evidence of change in the use of CT scans before vs after the study by Rustin et al5,6 within institutions.
Figure 4 shows the cumulative incidence of the time to retreatment after doubling of the CA-125 level before and after 2009. The median time to retreatment was 2.8 months before 2009 and 3.5 months after 2009 (P = .40); 71% of patients restarted treatment within 6 months during both periods. In adjusted analyses, patients diagnosed with advanced-stage disease had shorter times to retreatment compared with those with stage I disease (adjusted odds ratio, 6.54; 95% CI, 3.31-12.95), although there was no significant difference in the time to retreatment by era overall.
A sensitivity analysis was performed to compare the use of CA-125 and CT scans as well as the time to retreatment before and after the date of study publication (October 2010) instead of the American Society of Clinical Oncology plenary presentation (June 2009), and there were no significant differences in our findings. In a second sensitivity analysis, we examined whether individual institutions responded differently to the results of the study by Rustin et al5,6 and found substantive differences in the time to retreatment by institution (Figure 5). Specifically, while there was no change in the time to retreatment within 4 institutions (institution group 1; median time to treatment before and after 2009, 3.9 months; log-rank test, P = .70), there was evidence of a delayed time to retreatment within 2 institutions (institution group 2; median time to retreatment before and after 2009, 1.4 months vs 3.2 months; log-rank test, P = .01). As shown in eTable 3 in the Supplement (see model 2), an interaction analysis confirmed that patients treated at institutions in group 2 were more likely to be retreated earlier than patients treated within institution group 1 before 2009 (hazard ratio, 1.69; P < .001), but not after 2009 (hazard ratio, 1.06; P = .69).
The mean and median number of CA-125 tests during the 12-month surveillance period was 4.6 and 4.0, respectively, while the mean and median number of CT scans was 1.7 and 1.0, respectively. The mean cost of testing during a 1-year period using 2015 Medicare national reimbursement rates was $1056.49 per patient ($130.41 for CA-125 tests and $926.08 for CT scans). Assuming that 15 329 women achieve a clinical remission annually, the total cost of annual surveillance testing for this group is approximately $16 194 647 per year ($1 999 029 per year for CA-125 tests and $14 195 618 for CT scans). If the number of surveillance CT scans is increased to 3 (ie, for one-third of patients in this study), the total cost of annual surveillance testing increases to $1747.65 per patient, with an estimated population cost of $26 789 377.
In this study, we found similar rates of surveillance testing at 6 academic centers before and after the presentation of a randomized clinical trial that demonstrated that the use of surveillance CA-125 testing for early detection of recurrent disease decreased patients’ quality of life without improving their overall survival.6 After 2009, more than 90% of patients had undergone 3 or more CA-125 tests during a 12-month period, suggesting that the study by Rustin et al5,6 did not change practice at 6 NCCN academic centers despite widespread discussion.13-16 These results extend those from a prior study that used hypothetical scenarios to survey US gynecologic oncologists about their surveillance practice patterns17 and several single-institution retrospective studies.18-20
There are several reasons why the results of the study by Rustin et al5,6 may not have been widely embraced. National guidelines did not consistently advise against surveillance testing.7,8 Many patients received single-agent platinum for treatment of recurrent disease instead of doublet therapy; thus, survival outcomes may not reflect the US standard of care.15 CA-125 testing may also offer other benefits not examined in the study by Rustin et al.6 Specifically, patients and physicians may be reluctant to stop CA-125 testing because it may offer a relatively inexpensive, patient-centered approach to maximize treatment options by identifying recurrent disease before irreversible complications ensue that might limit opportunities for eligibility for clinical trials or secondary cytoreductive surgery.21-24 To that end, continued testing may be a rational, values-based decision to do everything possible to guard against the worst-case scenario. Alternatively, patients and physicians may share a bias toward therapeutic action. Earlier studies have documented that data from randomized clinical trials supporting the adoption of new treatments and/or technologies change practice more rapidly than do those that support stopping existing practices.25,26
We hypothesized that physicians might delay the time to retreatment with chemotherapy as a strategy to minimize harm while continuing to monitor patients closely for disease complications. Although our overall results demonstrated no differences in the time to retreatment in the periods before and after the study by Rustin et al,5,6 we observed differences by institution. Specifically, while 4 institutions had similar—and somewhat delayed—times to retreatment throughout, 2 institutions demonstrated an increase in time to retreatment in the period after the study by Rustin et al.5,6 This increase was the only evidence of a change in practice patterns we observed. Notably, this delay after the study by Rustin et al5,6 brought these 2 institutions in line with the other 4.
To date, few studies have examined the use of CT scans for surveillance testing in ovarian cancer, to our knowledge. Although 1 small study found that patients who underwent routine CT scans had improved overall survival,27 another documented no difference.20 Although the NCCN and the Society for Gynecologic Oncology guidelines state that CT scans should be used only when clinically indicated,7,8 and the Choosing Wisely campaign explicitly states that health care professionals should avoid routine radiographic surveillance,4 we found that two-thirds of patients had undergone at least 1 CT scan within 6 months, while one-third of patients had undergone 3 or more CT scans within 12 months of stopping treatment. There was significant variation in the use of CT scans between institutions, suggesting that this may be a reasonable target for improving value. Future studies should prospectively examine whether CT scans improve survival, given the significant variation and costs associated with this practice.
There is a growing focus on the costs of cancer care as physicians and payors try to maximize value in cancer care.1,2,4,28 Others have estimated the cost of ovarian cancer surveillance by modeling the NCCN guidelines, including costs of CA-125 tests, associated laboratories, physician visits, and CT scans.18 Our findings extend these studies by quantifying the costs of actual surveillance testing at 6 cancer centers. Our data demonstrate that women undergo CA-125 tests approximately every 3 months and undergo 1 or more CT scan within the first year after entering remission, resulting in an estimated mean population cost for 1 year of ovarian cancer surveillance testing of $16 194 647, predominantly owing to radiographic imaging. If the number of CT scans is increased to 3 (one-third of patients underwent 3 or more CT scans), this calculation increases by $10 million per year. These are conservative estimates, including only the costs of the tests, without consideration of expenditures on laboratory tests, physician visits, diagnostic evaluations, or societal costs.
There are several limitations to our analysis. We were not able to definitively distinguish between CA-125 tests and imaging ordered in response to patients’ symptoms vs surveillance testing in asymptomatic patients, despite careful audits of medical records. We expect that our estimates are conservative, however, because some patients received care at outside facilities, which we could not capture. We also restricted the surveillance window to include the time between the first normal CA-125 concentration and a doubling of the CA-125 level, excluding baseline CT scans obtained after chemotherapy and CT scans performed after a doubling of the CA-125 level. Therefore, our results may underestimate actual surveillance testing. The analysis is also limited to 6 comprehensive cancer centers through 2012 and may not be representative of practices outside of academic institutions or in later time periods. Finally, the cost calculations are rough population estimates, reliant on assumptions of the number of women entering surveillance annually.
We found that CA-125 tests and CT scans are routinely used in patients with ovarian cancer who are in clinical remission. Although a randomized clinical trial demonstrated that surveillance testing results in poorer quality of life without improvements in survival, our results demonstrate that the recommendation to avoid routine surveillance testing has not been adopted into clinical practice in the United States. Similarly, although the routine use of CT scans has been strongly discouraged by guideline committees, CT scans appear to be routinely used, at significant cost. These practices have significant, but poorly understood, psychosocial and cost implications, and no benefit on survival to date. Future studies should examine which patient populations benefit most from surveillance testing to improve the value of cancer care in this patient population.
Accepted for Publication: April 5, 2016.
Corresponding Author: Katharine M. Esselen, MD, MBA, Division of Gynecologic Oncology, Department of Obstetrics/Gynecology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Ave, Boston, MA 02215 (firstname.lastname@example.org).
Published Online: July 21, 2016. doi:10.1001/jamaoncol.2016.1842
Author Contributions: Ms Cronin and Dr Wright had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Esselen, Wright.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Esselen, Cronin, Wright.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Cronin, Wright.
Obtained funding: Wright.
Administrative, technical, or material support: Esselen, Matulonis, Niland, O’Malley, Wright.
Study supervision: Wright.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant K07 CA166210 from the National Cancer Institute (Dr Wright) and grant RP140020 from the Cancer Prevention and Research Institute of Texas (Dr Meyer).
Role of the Funder/Sponsor: The funding sources had no input into the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Jane C. Weeks, MD, MSc, Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, was involved in the conception of this study, design of the database, and obtained funding to create the database, but did not live to see its completion. She was not compensated for her contribution. We acknowledge the following people for their contributions to this project: all the study participants, the Ovarian Cancer Outcomes Consortium for material and administrative support, and the National Comprehensive Cancer Center Network for providing partial support for the outcomes database infrastructure and management.