Association Between Overall Survival and the Tendency for Cancer Programs to Administer Neoadjuvant Chemotherapy for Patients With Advanced Ovarian Cancer

Alexander Melamed; J Alejandro Rauh-Hain; Allison A Gockley; Roni Nitecki; Pedro T Ramirez; Dawn L Hershman; Nancy Keating; Jason D Wright

doi:10.1001/jamaoncol.2021.4252

. 2021 Sep 30;7(12):1–9. doi: 10.1001/jamaoncol.2021.4252

Association Between Overall Survival and the Tendency for Cancer Programs to Administer Neoadjuvant Chemotherapy for Patients With Advanced Ovarian Cancer

Alexander Melamed ^1,^2,^3,^✉, J Alejandro Rauh-Hain ⁴, Allison A Gockley ^1,², Roni Nitecki ⁴, Pedro T Ramirez ⁴, Dawn L Hershman ^2,^3,^5,⁶, Nancy Keating ^7,⁸, Jason D Wright ^1,^2,³

¹Department of Obstetrics and Gynecology, Columbia University Vagelos College of Physicians and Surgeons, New York, New York

²NewYork-Presbyterian/Columbia University Irving Medical Center, New York, New York

³Herbert Irving Comprehensive Cancer Center, New York, New York

⁴Department of Gynecologic Oncology and Reproductive Medicine, The University of Texas MD Anderson Cancer Center, Houston

⁵Department of Medicine, Columbia University Vagelos College of Physicians and Surgeons, New York, New York

⁶Department of Epidemiology, Columbia University Mailman School of Public Health, New York, New York

⁷Department of Health Policy, Harvard Medical School, Boston, Massachusetts

⁸Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts

Accepted for Publication: July 7, 2021.

Published Online: September 30, 2021. doi:10.1001/jamaoncol.2021.4252

^✉

Corresponding Author: Alexander Melamed, MD, MPH, Department of Obstetrics and Gynecology, Columbia University Vagelos College of Physicians and Surgeons, 161 Fort Washington Ave, Room 456, New York, NY 10032 (am5195@cumc.columbia.edu).

Author Contributions: Dr Melamed had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis that he conducted. Drs Melamed and Rauh-Hain contributed equally.

Concept and design: Melamed, Rauh-Hain, Nitecki, Ramirez, Hershman, Keating.

Acquisition, analysis, or interpretation of data: Melamed, Rauh-Hain, Gockley, Keating, Wright.

Drafting of the manuscript: Melamed, Rauh-Hain, Ramirez.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Melamed, Rauh-Hain, Nitecki, Keating.

Obtained funding: Melamed.

Administrative, technical, or material support: Melamed, Gockley, Wright.

Supervision: Rauh-Hain, Ramirez.

Conflict of Interest Disclosures: Dr Keating reported grants from National Cancer Institute and a contract from the US Centers for Medicare & Medicaid Services outside the submitted work. Dr Wright reported personal fees from Clovis Oncology and UpToDate and grants from Merck outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by the National Center for Advancing Translational Science (Dr Melamed; grant KL2TR001874) and the National Cancer Institute (Dr Rauh-Hain; grant K08 CA234333; Drs Rauh-Hain, Nitecki, and Ramirez; grant P30 CA016672; Drs Melamed, Hershman, and Wright; grant P30CA013696).

Role of the Funder/Sponsor: The funding organizations had no role in 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: We thank Laura L. Russell, ELS, MD Anderson Cancer Center’s Research Medical Library, for editorial assistance with this article. She was not compensated for her contributions.

^✉

Corresponding author.

PMCID: PMC8485210 PMID: 34591081

Key Points

Question

Was the differential adoption of neoadjuvant chemotherapy by US cancer centers for advanced-stage epithelial ovarian cancer associated with differences in overall survival?

Findings

In this difference-in-differences comparative effectiveness research study that included 39 299 patients treated in 664 cancer programs, patients treated in programs that markedly increased administration of neoadjuvant chemotherapy achieved greater improvements in short-term mortality and equivalent gains in median overall survival compared with patients who were treated in programs that continued to use the treatment infrequently.

Meaning

The study findings suggest that neoadjuvant chemotherapy may be an appropriate first-line treatment strategy for many patients with advanced-stage ovarian cancer.

Abstract

Importance

Randomized clinical trials have found that, in patients with advanced-stage epithelial ovarian cancer, neoadjuvant chemotherapy has similar long-term survival and improved perioperative outcomes compared with primary cytoreductive surgery. Despite this, considerable controversy remains about the appropriate use of neoadjuvant chemotherapy, and the proportion of patients who receive this treatment varies considerably among cancer programs in the US.

Objective

To evaluate the association between high levels of neoadjuvant chemotherapy administration and overall survival in patients with advanced ovarian cancer.

Design, Setting, and Participants

This difference-in-differences comparative effectiveness analysis leveraged differential adoption of neoadjuvant chemotherapy in Commission on Cancer–accredited cancer programs in the US and included women with a diagnosis of stage IIIC and IV epithelial ovarian cancer between January 2004 and December 2015 who were followed up through the end of 2018. The data were analyzed between September 2020 and January 2021.

Exposures

Treatment in a cancer program with high levels of neoadjuvant chemotherapy administration (more often than expected based on case mix) or in a program that continued to restrict its use after the 2010 publication of a clinical trial demonstrating the noninferiority of neoadjuvant chemotherapy compared with primary surgery for the treatment of patients with advanced ovarian cancer.

Main Outcomes and Measures

Case mix–standardized median overall survival time and 1-year all-cause mortality assessed with a flexible parametric survival model.

Results

We identified 19 562 patients (mean [SD] age, 63.9 [12.6] years; 3.2% Asian, 8.0% Black, 4.8% Hispanic, 82.5% White individuals) who were treated in 332 cancer programs that increased use of neoadjuvant chemotherapy from 21.7% in 2004 to 2009 to 42.2% in 2010 to 2015 and 19 737 patients (mean [SD] age, 63.5 [12.6] years; 3.1% Asian, 7.7% Black, 6.5% Hispanic, 81.8% White individuals) who were treated in 332 programs that marginally increased use of neoadjuvant chemotherapy (20.1% to 22.5%) over these periods. The standardized median overall survival times improved by similar magnitudes in programs with high (from 31.6 [IQR, 12.3-70.1] to 37.9 [IQR, 17.0-84.9] months; 6.3-month difference; 95% CI, 4.2-8.3) and low (from 31.4 [IQR, 12.1-67.2] to 36.8 [IQR, 15.0-80.3] months; 5.4-month difference, 95% CI, 3.5-7.3) use of neoadjuvant chemotherapy after 2010 (difference-in-differences, 0.9 months; 95% CI, −1.9 to 3.7). One-year mortality declined more in programs with high (from 25.6% to 19.3%; risk difference, −5.2%; 95% CI, −6.4 to −4.1) than with low (from 24.9% to 21.8%; risk difference, −3.2%, 95% CI, −4.3 to −2.0) use of neoadjuvant chemotherapy (difference-in-differences, −2.1%; 95% CI, −3.7 to −0.5).

Conclusions and Relevance

In this comparative effectiveness research study, compared with cancer programs with low use of neoadjuvant chemotherapy, those with high use had similar improvements in median overall survival and larger declines in short-term mortality.

This comparative effectiveness research study examines the association between high levels of neoadjuvant chemotherapy administration and overall survival in patients with advanced ovarian cancer.

Introduction

In 2010, Vergote and colleagues¹ reported that patients with advanced ovarian cancer randomized to receive neoadjuvant chemotherapy followed by surgery had less perioperative morbidity and similar long-term survival rates compared with patients randomized to undergo primary surgery. Although these findings have been supported by 3 subsequent randomized clinical trials,^2,3,4 national guidelines^5,6 and some experts^7,8 continue to recommend primary surgery as the preferred treatment for most patients with advanced ovarian cancer.

This discrepancy between trial findings^1,2,3,4 and expert recommendations^5,6,7,8 may be associated with perceived limitations of the published randomized clinical trials, including the low rates of optimal cytoreduction,^1,2 selective enrollment of patients with a poor prognosis,^1,2,3,8 and shorter-than-expected survival durations.^1,2 Furthermore, some observational studies have found that patients who undergo primary cytoreductive surgery live longer than those who receive neoadjuvant chemotherapy.^9,10 Accordingly, there are ongoing concerns that the results of randomized clinical trials may not apply to many patients with advanced ovarian cancer and that too frequent use of neoadjuvant chemotherapy could harm patients who might have otherwise undergone primary surgery.

Despite these apprehensions, use of neoadjuvant chemotherapy in patients with advanced ovarian cancer increased substantially in the US following the publication of the randomized clinical trials.^11,12 Some cancer programs now administer neoadjuvant chemotherapy to most patients with advanced ovarian cancer.¹³ To estimate the association between high use of neoadjuvant chemotherapy and survival, we used data from the National Cancer Database to document the varied uptake of neoadjuvant chemotherapy in US cancer programs following the publication of the first randomized clinical trial in 2010.¹ While some programs markedly increased the administration of neoadjuvant chemotherapy, practice patterns in others remained largely unchanged. This divergence in practice allowed us to implement a difference-in-differences¹⁴ study that evaluated the hypothesis that increased use of neoadjuvant chemotherapy by some cancer programs was not associated with decreased median overall survival.

Methods

Data Source

This study used data collected by the National Cancer Database, a multi-institutional cancer registry administered by the American College of Surgeons and the American Cancer Society (eAppendix in the Supplement).^15,16 Because these data are publicly available and deidentified, the study was exempted from review and informed consent by the Columbia University institutional review board.

Sample Selection

We included patients with clinical stage IIIC or IV epithelial ovarian cancer, including serous, mucinous, endometrioid, clear cell, or other adenocarcinoma histologies, who received a diagnosis from January 2004 through December 2015. We excluded patients who lacked pathologic confirmation or follow-up data, those who received no treatment at the reporting facility, and those treated in a program in which less than 1 case of advanced ovarian cancer was treated per year (Figure 1).

Figure 1. — Exclusion criteria are not mutually exclusive. The year 2010 served as the cutoff year for analysis owing to the 2010 publication of a clinical trial demonstrating the noninferiority of neoadjuvant chemotherapy compared with primary surgery for the treatment of patients with advanced ovarian cancer.¹

Periods

We relied on the 2010 publication of the first known randomized clinical trial that demonstrated the noninferiority of neoadjuvant chemotherapy compared with primary debulking surgery¹ to define the 2 periods used in this analysis. We categorized 2004 to 2009 as the prepublication period that preceded widespread use of neoadjuvant chemotherapy and 2010 to 2015 as the postpublication period, during which the use of this treatment approach became more frequent.^11,12

Case-Mix Adjustments

Multivariable models were used to estimate associations and predict case-mix adjusted rates. These included the following covariates: age (modeled as a restricted cubic spline), race and ethnicity (as recorded by the cancer registrar), Charlson-Deyo comorbidity score (0, 1, 2, or ≥3), histologic type (serous, mucinous, clear cell, endometrioid, or other adenocarcinoma), stage (IIIC or IV), tumor grade (1, 2, 3, or unknown), insurance status (uninsured, private, Medicare, Medicaid, other government program, or unknown), year of diagnosis (modeled discretely), cancer program type (community cancer, comprehensive community cancer, academic, or integrated network program), census region, and zip code–level median income.

Categorizing Cancer Programs

The exposure of interest was the cancer program-level tendency to administer neoadjuvant chemotherapy to patients with advanced ovarian cancer during the postpublication period. We categorized the tendency to administer neoadjuvant chemotherapy using the approach described by Barber et al.¹⁷ We used a multivariable logistic regression model to estimate the probabilities of patients receiving neoadjuvant chemotherapy on the basis of the aforementioned set of covariates (eFigure 1 in the Supplement). The predicted probabilities were summed for the patients treated in each cancer program to estimate the expected number of cases of neoadjuvant chemotherapy based on case mix. Programs were categorized as high users of neoadjuvant chemotherapy when the observed number of patients who received neoadjuvant chemotherapy exceeded the expected number and low users when the expected cases outnumbered the observed cases (eAppendix in the Supplement).

Cancer Program Matching

We used coarsened exact matching,¹⁸ a nonparametric matching algorithm, to identify programs with similar patterns of neoadjuvant chemotherapy use in the prepublication period that diverged after publication of the randomized clinical trial in 2010.¹ Specifically, we matched each cancer program with high use of neoadjuvant chemotherapy in the postpublication period to one with low use during that period based on their prepublication rates of case-mix–adjusted neoadjuvant chemotherapy and case volumes (eAppendix in the Supplement).

Outcomes

The primary outcome of interest was the median overall survival time. Vital status and the interval between diagnosis and last follow-up were ascertained through the end of 2018. Secondary outcomes included the 6- and 12-month mortality rates, deaths within 30 and 90 days of cytoreductive surgery, and treatment with frontline chemotherapy (adjuvant or neoadjuvant) and cytoreductive surgery.

Analytic Approach

We used a difference-in-differences design for this study. Difference-in-differences studies can estimate unbiased associations between group-level exposures and outcomes when it can be assumed that exposed and unexposed groups would have had identical trends in their outcomes had both groups remained unexposed.¹⁴ The differential responses of cancer programs to emerging evidence of the benefits of neoadjuvant chemotherapy in patients with advanced ovarian cancer provided a natural experiment in which it was possible to use this study design.

To evaluate the appropriateness of the difference-in-differences design, we compared prepublication trends in standardized median overall survival times between programs categorized as high or low users of neoadjuvant chemotherapy (eAppendix in the Supplement). Furthermore, we investigated whether patient characteristics evolved differentially among programs categorized as high vs low users of neoadjuvant chemotherapy over the study period because evidence of differential changes in patient characteristics would challenge the plausibility of nondifferential trends in outcomes. We used linear and logistic regression models to quantify differential changes in demographic characteristics, tumor features, and socioeconomic variables between programs with high vs low use of neoadjuvant chemotherapy. These models included indicator variables for period (prepublication vs postpublication) and program-level neoadjuvant chemotherapy use during the postpublication period (high vs low) as well as the interaction between these variables (eAppendix in the Supplement). Similar models were used to estimate the case-mix–adjusted rates of chemotherapy use, frontline surgery, and 30-day and 90-day surgical mortality, except that these models also included the covariates used for case-mix adjustment (eAppendix in the Supplement).

To evaluate the association between high vs low use of neoadjuvant chemotherapy and median overall survival time, as well as 6-month and 12-month all-cause mortality, we used flexible parametric survival models.¹⁹ These models were convenient for estimating associations between program-level treatment tendencies and changes in case-mix–adjusted survival outcomes, generating and contrasting standardized survival curves, and modeling time-varying treatment associations (eAppendix in the Supplement). We used a model that included indicator variables for the period of diagnosis (prepublication vs postpublication) and cancer program treatment tendency (high vs low use of neoadjuvant chemotherapy) and a period-by-treatment tendency interaction. We also included the set of covariates used for case-mix adjustment. The cancer program treatment tendency, period, and interaction between these variables were modeled as time-varying covariates to allow for time-varying associations between treatment tendency and survival, such as the possibility of an early benefit but impaired long-term survival.²⁰ Marginal standardization was used to estimate and contrast standardized median survival times and mortality rates (eAppendix in the Supplement).²¹ Analyses were conducted in Stata, version 16 (StataCorp). Statistical significance was inferred when 95% CIs excluded a null association.

Sensitivity Analyses

We undertook several sensitivity analyses. Because matching on prepublication covariates may reduce bias in some circumstances²² but introduce bias in others,²³ we repeated the main analysis without matching. In addition, we repeated the primary analysis without case-mix standardization because there was little evidence of differential change in covariates between programs with high vs low use of neoadjuvant chemotherapy. We varied the definitions of high-use vs low-use programs and prepublication and postpublication periods to evaluate the sensitivity of our analysis to these choices.

Results

We identified 86 701 women with a diagnosis of stage IIIC and IV epithelial ovarian cancer who were treated at Commission on Cancer–accredited programs in the US from January 2004 through December 2015. The flow diagram for the cohort selection is shown in Figure 1. After excluding 11 216 patients, we identified 35 222 patients who were treated in 522 programs with high use of neoadjuvant chemotherapy during the postpublication period (2010-2015) and 40 263 patients treated in 616 programs that maintained low use of neoadjuvant chemotherapy. After matching each program with high use of neoadjuvant chemotherapy during the postpublication period to one that maintained low use, based on case volume and use of neoadjuvant chemotherapy during the prepublication period (2004-2009), we identified 19 562 patients who were treated in 332 programs with high use of neoadjuvant chemotherapy and 19 737 patients treated in 332 programs with low use of neoadjuvant chemotherapy.

The cancer programs with high and low use of neoadjuvant chemotherapy administered it at similar rates during the prepublication period but diverged following the publication of a randomized clinical trial that demonstrated the noninferiority of neoadjuvant chemotherapy (Figure 2; eFigure 2 in the Supplement). Programs with low use of neoadjuvant chemotherapy administered neoadjuvant chemotherapy to 20.1% of patients with advanced ovarian cancer in the prepublication period (95% CI, 18.9%-21.4%) and to 22.5% of patients in the postpublication period (95% CI, 21.2%-23.4%). Programs with high use of neoadjuvant chemotherapy administered neoadjuvant chemotherapy to 21.7% of patients in the prepublication period (95% CI, 20.3%-23.1%) and to 42.2% of patients in the postpublication period (95% CI, 40.8%-43.7%; difference-in-differences, 18.1%; 95% CI, 15.8%-20.4%).

Figure 2. — Frequency of neoadjuvant chemotherapy administration among patients treated in cancer programs with high (gray) or low (orange) use of neoadjuvant chemotherapy before and after the publication of a randomized clinical trial. The year 2010 served as the cutoff year for analysis owing to the 2010 publication of a clinical trial demonstrating the noninferiority of neoadjuvant chemotherapy compared with primary surgery for the treatment of patients with advanced ovarian cancer.¹

Patient characteristics from the prepublication and postpublication periods are shown in Table 1. Changes in the distributions of patient ages, races and ethnicities, cancer stages, histologies, tumor grades, comorbidity indices, insurance statuses, or residential zip code median incomes from the prepublication to postpublication periods did not differ significantly between cancer programs with high vs low use of neoadjuvant chemotherapy. Additionally, trends in median overall survival time did not differ significantly between the high-use and low-use programs during the prepublication period, suggesting support for parallel trends (eFigure 3 in the Supplement).

Table 1. Characteristics of Patients Treated for Advanced Ovarian Cancer Before and After the 2010 Publication of a Clinical Trial Demonstrating the Noninferiority of Neoadjuvant Chemotherapy Compared With Primary Surgery.

Characteristic	High-use programs^a		Low-use programs^b		Difference-in-differences (95% CI)^c
Characteristic	Prepublication, 2004-2009 (n = 10 099)	Postpublication, 2010-2015 (n = 9463)	Prepublication, 2004-2009 (n = 10 247)	Postpublication, 2010-2015 (n = 9490)	Difference-in-differences (95% CI)^c
Mean (SD) age, y	63.8 (12.9)	63.9 (12.3)	63.4 (12.7)	63.6 (12.4)	−0.4 (−0.61 to 0.52)
Race and ethnicity, %
Asian	2.7	3.7	2.6	3.6	0 (−0.8 to 0.8)
Black	7.7	8.5	6.9	8.5	−0.9 (−2.1 to 0.4)
Hispanic	4.3	5.3	5.4	7.6	−1.1 (−2.4 to 0.1)
White	83.7	81.2	84.0	79.5	1.8 (−0.1 to 3.8)
Other/unknown	1.6	1.3	1.2	0.8	0.1 (−0.5 to 0.8)
Stage IV, %	38.1	38.0	35.1	36.4	−1.5 (−4.6 to 1.7)
Histology, %
Serous	67.5	73.7	69.2	74.1	1.2 (−1.3 to 3.8)
Endometrioid	4.1	2.8	5.2	3.6	0.3 (−0.6 to 1.3)
Clear cell	3.0	3.0	2.9	3.3	−0.3 (−1.0 to 0.5)
Mucinous	2.7	2.0	2.5	2.3	−0.5 (−1.1 to 0.1)
Other adenocarcinoma	22.7	18.5	20.2	16.8	−0.8 (−3.2 to 1.5)
Tumor grade, %
1	3.0	2.4	2.7	2.7	−0.6 (−1.3 to 0.1)
2	12.1	6.8	11.7	6.4	0 (−1.6 to 1.6)
3	61.2	63.8	64.4	66.7	0.3 (−2.7 to 3.3)
Unknown	23.8	27.1	21.2	24.2	0.3 (−2.6 to 3.2)
Charlson-Deyo score, %
0	80.1	78.5	79.2	79.3	−1.8 (−3.8 to 0.2)
1	15.6	16.5	16.3	16.2	0.9 (−0.8 to 2.6)
2	3.2	3.8	3.3	3.3	0.6 (−0.1 to 1.4)
≥3	1.0	1.3	1.3	1.3	0.2 (−0.2 to 0.7)
Insurance type, %
Medicare	44.7	46.3	44.4	44.6	1.4 (−1.0 to 3.8)
Private	44.5	40.8	43.0	40.3	−1.0 (−3.6 to 1.6)
Medicaid/other	5.8	8.1	6.4	8.8	−0.1 (−1.4 to 1.1)
None	3.4	3.7	4.6	4.8	0.1 (−1.2 to 1.5)
Unknown	1.7	1.1	1.7	1.5	−0.4 (−1.2 to 0.5)
Zip code median income, $, %
<38 000	15.3	14.8	17.1	16.3	0.3 (−1.8 to 2.3)
38 000-47 999	21.9	20.6	23.6	21.5	0.8 (−1.2 to 2.8)
48 000-62 999	26.7	25.2	26.5	25.9	−1.0 (−3.0 to 1.1)
≥63 000	34.2	32.6	30.1	29.7	−1.2 (−3.7 to 1.4)
Unknown	1.9	6.9	2.8	6.7	1.1 (−1.3 to 3.5)

Open in a new tab

^{^a}

High-use programs were cancer programs that administered neoadjuvant chemotherapy more frequently than expected based on case mix during the period following publication of Vergote et al¹ (2010-2015).

^{^b}

Low-use programs were cancer programs that administered neoadjuvant chemotherapy less frequently than expected based on case mix during the period following publication of Vergote et al¹ (2010-2015).

^{^c}

The difference-in-differences compares the change from the prepublication period with the postpublication period between programs with high use of neoadjuvant chemotherapy and those with low use. Positive quantities reflect a larger increase among high-use programs than low-use programs. There were no statistically significant differential changes.

The standardized median overall survival time improved significantly from the prepublication to postpublication period in programs with low and high use of neoadjuvant chemotherapy (Figure 3, A and B). In programs that maintained low use, the standardized median survival time improved from 31.4 months (IQR, 12.1-67.2 months) during 2004 to 2009 to 36.8 months (IQR, 15.0-80.3 months) during 2010 to 2015 (absolute difference, 5.4 months; 95% CI, 3.5-7.3 months). The standardized median overall survival time improved from 31.6 months (IQR, 12.3-70.1 months) to 37.9 months (IQR, 17.0-84.9 months) (absolute difference, 6.3 months; 95% CI, 4.2-8.3 months) over the same period in programs with high use of neoadjuvant chemotherapy from 2010 to 2015. The change in the standardized median survival time did not differ significantly in programs with high compared with low use of neoadjuvant chemotherapy (difference-in-differences, 0.9 months; 95% CI, −1.9 to 3.7 months). However, patients who were treated in programs with high use of neoadjuvant chemotherapy experienced greater reductions in early mortality (Figure 3C). In programs with high use of neoadjuvant chemotherapy, the standardized 6-month mortality rate declined from 16.4% to 12.0% (risk difference, −4.4%; 95% CI, −5.3% to −3.5%), whereas in programs that maintained low use of neoadjuvant chemotherapy, the rate declined from 16.1% to 14.4% (risk difference, −1.6; 95% CI, −2.6 to 0.8; difference-in-differences, −2.3%; 95% CI, −3.2% to −1.3%). Similarly, the standardized 1-year mortality rate declined by 5.2% (95% CI, 4.1%-6.4%), from 24.6% to 19.3%, in programs with high use of neoadjuvant chemotherapy compared with a decline of 3.2% (95% CI, 2.0%-4.3%), from 24.9% to 21.8%, in low-use programs (difference-in-differences, −2.1%; 95% CI, −3.7% to −0.5%). The differential changes in short-term mortality were mirrored by declines in 30-day and 90-day postoperative mortality, which were larger in high-use than low-use programs (Table 2).

Figure 3. — Standardized survival curves for patients with a diagnosis from 2004 to 2009 (orange) compared with 2010 to 2015 gray) and confidence bands are plotted for cancer programs with high (A) and low (B) use of neoadjuvant chemotherapy during the latter period. Similar gains in median survival, indicated by the dashed lines, were observed in programs with high and low use. The change in the OS probability (and associated confidence bands[C]) is plotted for programs with high (gray) and low (orange) use of neoadjuvant chemotherapy. The year 2010 served as the cutoff year for analysis owing to the 2010 publication of a clinical trial demonstrating the noninferiority of neoadjuvant chemotherapy compared with primary surgery for the treatment of patients with advanced ovarian cancer.¹

Table 2. Secondary Outcomes During the Prepublication and Postpublication Periods for Programs With High vs Low Use of Neoadjuvant Chemotherapy.

Outcome	Percentage (95% CI)				Difference-in-differences (95% CI)^c
	High-use programs^a		Low-use programs^b
	Prepublication, 2004-2009 (n = 10 099)	Postpublication, 2010-2015 (n = 9463)	Prepublication, 2004-2009 (n = 10 247)	Postpublication, 2010-2015 (n = 9490)
6-mo Mortality	16.4 (15.5 to 17.3)^d	12.0 (11.3 to 12.8)	16.1 (15.2 to 17.0)	14.4 (13.5 to 15.4)	−2.3 (−3.2 to −1.3)
1-y Mortality	24.6 (23.5 to 25.6)	19.3 (18.4 to 20.2)	24.9 (24.0 to 25.9)	21.8 (20.7 to 22.8)	−2.1 (−3.7 to −0.5)
30-d Postoperative mortality^b	3.3 (2.9 to 3.7)	1.6 (1.3 to 1.9)	3.2 (2.8 to 3.5)	2.4 (2.0 to 2.7)	−1.0 (−1.6 to −0.3)
90-d Postoperative mortality^b	7.3 (6.6 to 7.9)	4.3 (3.8 to 4.9)	7.8 (7.1 to 8.4)	6.3 (5.7 to 6.9)	−1.5 (−2.5 to −0.5)
Surgery	83.5 (82.4 to 84.7)	83.0 (81.9 to 84.1)	84.4 (83.5 to 85.3)	86.4 (85.5 to 57.3)	−2.5 (−4.3 to −0.8)
Chemotherapy	82.2 (80.2 to 84.1)	90.6 (89.6 to 91.6)	83.4 (81.9 to 84.8)	84.6 (83.3 to 85.8)	7.2 (5.0 to 9.3)
Surgery and chemotherapy	68.0 (65.8 to 70.1)	75.2 (73.7 to 76.8)	69.9 (68.2 to 71.5)	74.5 (73.0 to 75.9)	2.7 (0.01 to 5.3)

Open in a new tab

^{^a}

^{^b}

^{^c}

The difference-in-differences compares the change, from the prepublication period to the postpublication period, between programs with high use of neoadjuvant chemotherapy and those with low use. Positive quantities reflect a larger increase among high-use programs compared with low-use programs.

^{^d}

All outcomes were standardized for age, stage, histologic type, tumor grade, Charlson comorbidity index score, race and ethnicity, zip code–level median income, cancer program type, census region, and insurance status.

Consistent with the main analysis, improvements in median overall survival between the prepublication and postpublication periods were similar among high-use and low-use cancer programs in a sensitivity analysis omitting cancer program matching based on prepublication practice patterns (eFigures 4 and 5 in the Supplement). Restricting the primary analysis to include only those high-use programs that were in the highest quartile of observed vs expected neoadjuvant chemotherapy use identified a cohort of cancer programs that administered neoadjuvant chemotherapy to 49.2% of patients who were treated during the postpublication period. Comparing this subset of high users with low users, after matching on case volumes and the tendency to administer neoadjuvant chemotherapy during the prepublication period, produced similar results to those seen in the primary analysis (eFigures 6 and 7 in the Supplement). The main results were unchanged by omitting covariates from the models used in the primary analysis, nor did the results appear to depend on model-based estimates of median survival time (eFigure 8 in the Supplement). Finally, excluding patients who were treated in 2010 or 2011 to allow a ramp-up period (eFigure 9 in the Supplement) or defining the cutoff year as 2008 or 2009 (eFigure 10 in the Supplement) did not alter the primary findings.

Discussion

In this observational study, we identified substantial variation in the use of neoadjuvant chemotherapy following the publication of a randomized clinical trial that demonstrated the noninferiority of this treatment approach.¹ We used this variation in clinical practice to conduct a difference-in-differences analysis that evaluated the association between receiving care in cancer programs with high use of neoadjuvant chemotherapy compared with programs with low use of this treatment. The study findings suggest that receiving care in a program with high use of neoadjuvant chemotherapy was associated with a reduction in the risk of early mortality without a decrease in median overall survival.

The present results mirror those of 4 randomized clinical trials that compared neoadjuvant chemotherapy with primary surgery for advanced ovarian cancer.^1,2,3,4 Each trial found that neoadjuvant chemotherapy reduced the risk of surgical morbidity and mortality without compromising long-term overall survival.^1,2,3,4 A fifth randomized clinical trial investigating this question has completed patient enrollment, and results are expected by 2023.²⁴

While some observational studies comparing outcomes between patients who received neoadjuvant chemotherapy vs primary cytoreductive surgery have concluded that neoadjuvant chemotherapy is associated with worse survival outcomes,^25,26 these studies failed to consider the effect of important, unmeasured, patient-level confounders, such as disease burden and functional status. One study that evaluated the potential effect of these unmeasured confounders found that the apparent benefit of primary surgery could be explained by such factors.¹⁰ A cross-sectional analysis of National Cancer Database data found that patients who were treated in hospitals with lower-than-expected use of neoadjuvant chemotherapy had worse survival than those treated in hospitals with expected or higher-than-expected use,¹⁷ although this analysis did not account for unmeasured hospital-level differences. A strength of the present study is that, instead of focusing on irredeemably confounded variations in patient-level treatment selection, this analysis quantified the association between cancer program practice patterns and survival and was not susceptible to bias from patient-level factors that affect the selection of an individual’s treatment. Furthermore, the difference-in-differences design adjusted for measured and unmeasured time-invariant differences among cancer programs.¹⁴

In a prior study that examined the associations between regional variation in neoadjuvant chemotherapy use and survival,²⁷ our group found that US regions that rapidly adopted neoadjuvant chemotherapy for ovarian cancer between 2011 and 2012 had greater improvements in 3-year overall survival than did regions that did not adopt neoadjuvant therapy. The apparent discrepancy between those results and those of this study may be because of the longer follow-up period in this study and the time-variable associations between neoadjuvant chemotherapy use and overall survival (Figure 3C).

While this study suggests that administering neoadjuvant chemotherapy for up to 49% of patients with advanced ovarian cancer is associated with a decline in early mortality and does not negatively affect median overall survival, it remains unknown if there is an upper limit on the safe use of neoadjuvant chemotherapy in this population. A meta-analysis of 2 randomized clinical trials suggested that patients with low-volume stage IIIC disease may have a modest progression-free survival benefit with primary cytoreductive surgery.²⁸

Limitations

This study must be interpreted while considering limitations. The validity of the difference-in-differences design depends on the assumption that, if the programs included in the high-use and low-use groups had not developed differential use of neoadjuvant chemotherapy, improvements in survival would have been the same in these cancer programs. It is possible that unmeasured patient-level or clinician-level factors affected the survival of the patients treated in these programs differentially. For example, the National Cancer Database includes no information on maintenance therapy, dosing schedules, routes of primary chemotherapy, secondary surgery, or systemic therapy administered for recurrence, and substantial differences in these treatments could have biased our results. However, such a bias could only obscure a deleterious effect of high levels of neoadjuvant chemotherapy administration if high users of neoadjuvant chemotherapy also adopted another beneficial treatment to a larger extent than did programs with low use of neoadjuvant chemotherapy. Furthermore, confounding bias could have affected our results if the population of patients who received care in programs with high use of neoadjuvant chemotherapy evolved dissimilarly over the study period compared with programs with low use of neoadjuvant chemotherapy. Reassuringly, examination of observed patient characteristics revealed no evidence of such a selection bias. Finally, we categorized cancer programs as high or low users of neoadjuvant chemotherapy based on the observed-to-expected ratio of patients who received chemotherapy as their first treatment, even though some patients do not receive interval cytoreductive surgery. We chose not to exclude patients who received only chemotherapy or surgery because treatment for such patients is often initiated with the intention of administering surgery and chemotherapy and because understanding these patients’ outcomes is vital to understanding associations between program-level neoadjuvant chemotherapy use and survival.

Conclusions

This comparative effectiveness research study suggests that programs with high use of neoadjuvant chemotherapy after 2010 achieved greater improvements in short-term mortality and equivalent gains in median overall survival compared with programs that continued to use the treatment infrequently. These results may reassure those concerned that frequent administration of neoadjuvant chemotherapy has harmed patients with advanced ovarian cancer.

Supplement.

eMethods.

eFigure 1. Receiver operating characteristic curve

eFigure 2. Observed-to-expected ratios in high-use vs. low-use programs

eFigure 3. Graphic evaluation of nondifferential trends

eFigure 4. Differential adoption of neoadjuvant chemotherapy in unmatched cancer programs

eFigure 5. Difference-in-differences analysis in unmatched programs.

eFigure 6. Differential adoption of neoadjuvant chemotherapy in very high- vs. low-use programs

eFigure 7. Difference-in-differences analysis comparing very high- vs. low-use programs

eTable 1. Comparison of crude median survival time and 95% CIs estimated using flexible parametric models and the Kaplan-Meier method

eFigure 8. Model-based vs. Kaplan-Meier estimates of changes in the crude survival probability

eFigure 9. Difference-in-differences analyses with excluded ramp-up period

eFigure10. Difference-in-differences analyses with early adoption of neoadjuvant chemotherapy

eFigure 11. Differential adoption in very high-use vs. very low-use programs

eFigure 12. Difference-in-differences analysis comparing very high-use vs. very low-use programs

Click here for additional data file.^{(1.9MB, pdf)}

References

1.Vergote I, Tropé CG, Amant F, et al. ; European Organization for Research and Treatment of Cancer-Gynaecological Cancer Group; NCIC Clinical Trials Group . Neoadjuvant chemotherapy or primary surgery in stage IIIC or IV ovarian cancer. N Engl J Med. 2010;363(10):943-953. doi: 10.1056/NEJMoa0908806 [DOI] [PubMed] [Google Scholar]
2.Kehoe S, Hook J, Nankivell M, et al. Primary chemotherapy versus primary surgery for newly diagnosed advanced ovarian cancer (CHORUS): an open-label, randomised, controlled, non-inferiority trial. Lancet. 2015;386(9990):249-257. doi: 10.1016/S0140-6736(14)62223-6 [DOI] [PubMed] [Google Scholar]
3.Onda T, Satoh T, Ogawa G, et al. ; Japan Clinical Oncology Group . Comparison of survival between primary debulking surgery and neoadjuvant chemotherapy for stage III/IV ovarian, tubal and peritoneal cancers in phase III randomised trial. Eur J Cancer. 2020;130:114-125. doi: 10.1016/j.ejca.2020.02.020 [DOI] [PubMed] [Google Scholar]
4.Fagotti A, Ferrandina MG, Vizzielli G, et al. Randomized trial of primary debulking surgery versus neoadjuvant chemotherapy for advanced epithelial ovarian cancer (SCORPION-NCT01461850). Int J Gynecol Cancer. 2020;30(11):1657-1664. doi: 10.1136/ijgc-2020-001640 [DOI] [PubMed] [Google Scholar]
5.National Comprehensive Cancer Network . NCCN Clinical Practice Guidelines in Oncology: ovarian cancer including fallopian tube cancer and primary peritoneal cancer: version 1.2021. Accessed March 2, 2021. https://www.nccn.org/professionals/physician_gls/pdf/ovarian.pdf
6.Wright AA, Bohlke K, Armstrong DK, et al. Neoadjuvant chemotherapy for newly diagnosed, advanced ovarian cancer: Society of Gynecologic Oncology and American Society of Clinical Oncology clinical practice guideline. J Clin Oncol. 2016;34(28):3460-3473. doi: 10.1200/JCO.2016.68.6907 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Fotopoulou C, Sehouli J, Aletti G, et al. Value of neoadjuvant chemotherapy for newly diagnosed advanced ovarian cancer: a European perspective. J Clin Oncol. 2017;35(6):587-590. doi: 10.1200/JCO.2016.71.0723 [DOI] [PubMed] [Google Scholar]
8.du Bois A, Baert T, Vergote I. Role of neoadjuvant chemotherapy in advanced epithelial ovarian cancer. J Clin Oncol. 2019;37(27):2398-2405. doi: 10.1200/JCO.19.00022 [DOI] [PubMed] [Google Scholar]
9.Meyer LA, Cronin AM, Sun CC, et al. Use and effectiveness of neoadjuvant chemotherapy for treatment of ovarian cancer. J Clin Oncol. 2016;34(32):3854-3863. doi: 10.1200/JCO.2016.68.1239 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Rauh-Hain JA, Melamed A, Wright A, et al. Overall survival following neoadjuvant chemotherapy vs primary cytoreductive surgery in women with epithelial ovarian cancer: analysis of the National Cancer Database. JAMA Oncol. 2017;3(1):76-82. doi: 10.1001/jamaoncol.2016.4411 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Melamed A, Hinchcliff EM, Clemmer JT, et al. Trends in the use of neoadjuvant chemotherapy for advanced ovarian cancer in the United States. Gynecol Oncol. 2016;143(2):236-240. doi: 10.1016/j.ygyno.2016.09.002 [DOI] [PubMed] [Google Scholar]
12.Knisely AT, St Clair CM, Hou JY, et al. Trends in primary treatment and median survival among women with advanced-stage epithelial ovarian cancer in the US from 2004 to 2016. JAMA Netw Open. 2020;3(9):e2017517. doi: 10.1001/jamanetworkopen.2020.17517 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Fleming ND, Nick AM, Coleman RL, et al. Laparoscopic surgical algorithm to triage the timing of tumor reductive surgery in advanced ovarian cancer. Obstet Gynecol. 2018;132(3):545-554. doi: 10.1097/AOG.0000000000002796 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Moss HA, Melamed A, Wright JD. Measuring cause-and-effect relationships without randomized clinical trials: quasi-experimental methods for gynecologic oncology research. Gynecol Oncol. 2019;152(3):533-539. doi: 10.1016/j.ygyno.2018.11.006 [DOI] [PubMed] [Google Scholar]
15.Merkow RP, Rademaker AW, Bilimoria KY. Practical guide to surgical data sets: National Cancer Database (NCDB). JAMA Surg. 2018;153(9):850-851. doi: 10.1001/jamasurg.2018.0492 [DOI] [PubMed] [Google Scholar]
16.Raval MV, Bilimoria KY, Stewart AK, Bentrem DJ, Ko CY. Using the NCDB for cancer care improvement: an introduction to available quality assessment tools. J Surg Oncol. 2009;99(8):488-490. doi: 10.1002/jso.21173 [DOI] [PubMed] [Google Scholar]
17.Barber EL, Dusetzina SB, Stitzenberg KB, et al. Variation in neoadjuvant chemotherapy utilization for epithelial ovarian cancer at high volume hospitals in the United States and associated survival. Gynecol Oncol. 2017;145(3):500-507. doi: 10.1016/j.ygyno.2017.03.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Iacus SM, King G, Porro G. Multivariate matching methods that are monotonic imbalance bounding. J Am Stat Assoc. 2011;106 (493): 345-361. doi: 10.1198/jasa.2011.tm09599 [DOI] [Google Scholar]
19.Royston P, Parmar MK. Flexible parametric proportional-hazards and proportional-odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med. 2002;21(15):2175-2197. doi: 10.1002/sim.1203 [DOI] [PubMed] [Google Scholar]
20.Havrilesky LJ, Yang JC, Lee PS, et al. Patient preferences for attributes of primary surgical debulking versus neoadjuvant chemotherapy for treatment of newly diagnosed ovarian cancer. Cancer. 2019;125(24):4399-4406. doi: 10.1002/cncr.32447 [DOI] [PubMed] [Google Scholar]
21.Royston P, Lambert PC. Flexible Parametric Survival Analysis in Stata: Beyond the Cox Model. Stata Press; 2011. [Google Scholar]
22.Stuart EA, Huskamp HA, Duckworth K, et al. Using propensity scores in difference-in-differences models to estimate the effects of a policy change. Health Serv Outcomes Res Methodol. 2014;14(4):166-182. doi: 10.1007/s10742-014-0123-z [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Daw JR, Hatfield LA. Matching and regression to the mean in difference-in-differences analysis. Health Serv Res. 2018;53(6):4138-4156. doi: 10.1111/1475-6773.12993 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Reuss A, du Bois A, Harter P, et al. TRUST: Trial of Radical Upfront Surgical Therapy in advanced ovarian cancer (ENGOT ov33/AGO-OVAR OP7). Int J Gynecol Cancer. 2019;29(8):1327-1331. doi: 10.1136/ijgc-2019-000682 [DOI] [PubMed] [Google Scholar]
25.Rosen B, Laframboise S, Ferguson S, et al. The impacts of neoadjuvant chemotherapy and of debulking surgery on survival from advanced ovarian cancer. Gynecol Oncol. 2014;134(3):462-467. doi: 10.1016/j.ygyno.2014.07.004 [DOI] [PubMed] [Google Scholar]
26.Lyons YA, Reyes HD, McDonald ME, et al. Interval debulking surgery is not worth the wait: a National Cancer Database study comparing primary cytoreductive surgery versus neoadjuvant chemotherapy. Int J Gynecol Cancer. 2020;30(6):845-852. doi: 10.1136/ijgc-2019-001124 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Melamed A, Fink G, Wright AA, et al. Effect of adoption of neoadjuvant chemotherapy for advanced ovarian cancer on all cause mortality: quasi-experimental study. BMJ. 2018;360:j5463. doi: 10.1136/bmj.j5463 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Vergote I, Coens C, Nankivell M, et al. ; EORTC; MRC CHORUS study investigators . Neoadjuvant chemotherapy versus debulking surgery in advanced tubo-ovarian cancers: pooled analysis of individual patient data from the EORTC 55971 and CHORUS trials. Lancet Oncol. 2018;19(12):1680-1687. doi: 10.1016/S1470-2045(18)30566-7 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials