Abstract
Objective
The best course of treatment for recurrent ovarian cancer is uncertain. We sought to determine whether secondary cytoreductive surgery for first recurrence of ovarian cancer improves overall survival compared to other treatments.
Methods
We assessed survival using Surveillance, Epidemiology and End Results-Medicare data for advanced stage ovarian cancer cases diagnosed from January 1, 1997- December 31, 2007 with survival data through 2010 using multinomial propensity weighted finite mixture survival regression models to distinguish true from misclassified recurrences. Of 35,995 women ages 66 years and older with ovarian cancer, 3439 underwent optimal primary debulking surgery with 6 cycles of chemotherapy; 2038 experienced a remission.
Results
1635 out of 2038 (80%) women received treatment for recurrence of whom 72% were treated with chemotherapy only, 16% with surgery and chemotherapy and 12% received hospice care. Median survival of women treated with chemotherapy alone, surgery and chemotherapy, or hospice care was 4.1 years, 5.4 years, and 2.2 years, respectively (p<0.001). Of those receiving no secondary treatments, 75% were likely true nonrecurrences with median survival of 15.9 years and 25% misclassified with 2.4 years survival. Survival among women with recurrence was greater for those treated with surgery and chemotherapy compared to chemotherapy alone (HR=1.67; 95% confidence interval: 1.13–2.47). Women who were older with more comorbidities and high grade cancer had worse survival.
Conclusions
Secondary surgery with chemotherapy to treat recurrent ovarian cancer increases survival by 1.3 years compared to chemotherapy alone and pending ongoing randomized trial results, may be considered a standard of care.
Keywords: recurrent ovarian cancer, SEER-medicare, secondary surgery, survival
Background
The majority of ovarian cancers are diagnosed at late stage and initially respond to platinum-based chemotherapy; unfortunately most recur.[1] For women experiencing recurrent ovarian cancer, the best course of treatment is uncertain. Typically, recurrent cancer is treated with chemotherapy but starting in the early 1990s, secondary cytoreductive surgery (SCS) appeared as a treatment option.[2] By 1996, the National Comprehensive Cancer Network guidelines incorporated SCS as a treatment choice along with chemotherapy despite lack of randomized trial data of its efficacy.[3] A recent Cochrane review of secondary surgery for recurrent epithelial ovarian cancer concludes that existing findings are inconclusive.[4] Most of the studies were single institutional case reports and few reported survival comparing surgery with chemotherapy. No study was population-based. As 46% of ovarian cancer cases occurred in women 65 years and older,[1] SEER Medicare data offers an opportunity to assess survival benefits of secondary surgery over chemotherapy alone. However, such explorations have been hampered by the limitations of administrative data – lack of specific procedure codes for recurrences or SCS, the challenge of identifying recurrent cancer, a condition without a specific diagnostic code. While evidence of a second round of surgery may be used to identify recurrences, this approach has the potential to misclassify untreated patients as nonrecurrence. Additionally, there is the inherent observational bias of healthier individuals receiving more aggressive treatments. Randomized trials comparing chemotherapy with secondary surgery are underway but results are years away. To help inform decision-making for women with recurrent ovarian cancer, we performed a population-based analysis of secondary surgical treatment compared with chemotherapy and with hospice using doubly robust regression methods, a novel multiple treatment propensity score methods to account for selection into treatment, and finite mixture survival models to account for potential misclassification of nonrecurrence.
Methods
Cohort Selection
We used linked SEER-Medicare data for ovarian cancer cases diagnosed from January 1, 1997- December 31, 2007 with claims and mortality data through 2010 (N=35,995). We excluded women with more than one primary cancer (N=5806), nonepithelial ovarian cancer cases (N=871), those diagnosed prior to 1997 (N=956) as well as women who received the diagnosis at autopsy or on death certificate (N=16). We excluded women not enrolled in both Medicare parts A (inpatient) and B (outpatient) and those enrolled in a health maintenance organization (N=4150) as they do not have complete Medicare fee-for-service claims. We also excluded women with early stage cancer (N=10,952) who have a low likelihood of recurrence. Because we aimed to assess the effect of secondary debulking surgery for recurrent cancer, we included only who had undergone primary debulking for their primary ovarian cancer treatment. As we aimed to assess the comparative effectiveness of secondary surgery vs chemotherapy for recurrence, we included those treated optimally for their primary ovarian cancer, meaning those who underwent primary surgery and chemotherapy. Finally, we limited the analyses to women ≥66 years (N=5310) to capture comorbidities during the year prior to cancer diagnosis.
Identifying secondary surgery and first recurrence
We obtained codes used by gynecologic oncology billers from centers in the Northeast, mid-Atlantic, Southeast, Midwest, Southwest, Pacific and Pacific Northwest (list in Appendix 1) for SCS. A chart validation of these codes revealed a sensitivity of 77% and specificity of 92%.[5] To identify recurrent cancer, we created an algorithm based on expert gynecologic oncology opinion and the literature[4,6,7] that incorporates timing and utilization of either secondary surgery or chemotherapy. Recurrent cancer was defined by the presence of claims for secondary surgery, chemotherapy or hospice following a 180-day treatment free window after completion of primary course of therapy to distinguish between recurrent and persistent disease. We used diagnosis (ICD9) and procedure (CPT/Medicare HCPCS) codes for chemotherapy regimens commonly administered for recurrent ovarian cancer culled from gynecologic oncology billers (see Appendix 2). Chart validation of a diagnosis recurrent ovarian cancer using this algorithm yielded a sensitivity of 100% and specificity of 89%.[5]
Treatment
We identified primary treatment searching Medicare claims from 30 days prior to 120 days after the date of diagnosis. Since SEER reports only month and year of the diagnoses, the diagnosis date was assigned the 15th day of the month. We identified primary surgical treatment in inpatient files using ICD-9 procedure and outpatient and physician files using HCPCS codes, and receipt of chemotherapy using inpatient, outpatient, physician claims, and DME (Durable Medical Equipment) files. Primary chemotherapy was the first chemotherapy that occurred within 180 days prior to and 90 days after surgical treatment. Chemotherapy claims from all sources within three days were grouped into one. To distinguish between persistent and recurrent cancer, there had to be a 6 month treatment-free window following completion of primary treatment.
Treatment Groups
We classified treatment for recurrent cancer into 4 categories: secondary surgery (with or without chemotherapy), chemotherapy only, hospice, no treatment. A small number of women (N=26) appear to have had secondary surgery without chemotherapy. In preliminary analysis we dropped these observations from the analysis sample. As there were no substantive differences between this and our final sample in terms of means of explanatory variables or treatment effects, we included them with secondary surgery and chemotherapy group.
Sociodemographic and clinical variables
Age, race, SEER geographic region, year of diagnosis, tumor grade, histology and stage are from SEER. We used median household income of the woman’s residence zipcode. We assessed comorbidity using Elixhauser8 and Charlson9 indices to take into account both risk of in-hospital and longer term mortality and used comorbidities identified in inpatient and outpatient claims in the year prior to diagnosis of cancer.
Outcome
Overall survival was measured from the 15th day of the month of diagnosis to the date of death from any cause as reported in the Medicare database. Observations were censored at 12/31/2010, the last recorded date of follow-up in Medicare records.
Statistical Analysis
We use propensity score methods for multiple treatments[10] to minimize bias from selection into treatment. We estimate a multinomial logit model of treatments controlling for relevant socioeconomic and clinical characteristics such as age, race, income, stage, histology, comorbidity, diagnosis year and SEER region to predict type of treatment (surgery, chemotherapy, hospice, none) received. We use the estimated parameters to create inverse propensity score weights.
We use finite mixture log logistic survival models to assess the association of type of therapy with survival while controlling for propensity scores using inverse probability weighting.
The survival models also control for socioeconomic and clinical characteristics, thus the treatment effects obtained are doubly robust.[11] Finite mixture models allow for estimation of survival differences between treatment groups while differentiating the truly nonrecurrent cases from those women who did experience a recurrence but were misclassified because they did not receive SCS, chemotherapy or hospice care. Finite mixture models have been used to model cancer survival,[12,13] and have many applications in other areas of health services, biology, marketing, and engineering.[14–17] This approach distinguishes between individuals with a greater probability to be in one group as compared to another, such as those with longer vs. shorter survival.
Parametric survival analyses were used to identify distributions that provided the best fit to the data. Commonly used distributions including exponential, Weibull, log-normal and log-logistic curves were tested, and the fit of each was assessed by using the Akaike information criterion (AIC) criterion. These analyses showed that a log-logistic survival model fit the data best. The estimated parametric survival curves were also very close to observed Kaplan–Meier curves. Therefore, we used log-logistic distributions to construct the finite mixture model of survival allowing for two classes (types) of observations. Both the class probability and determinants of survival for each class (including effect of treatment) were estimated simultaneously using inverse propensity score weighted maximum likelihood.
Results
Of the 7934 women, 3439 underwent primary cytoreductive surgery and chemotherapy, standard optimal treatment for a new ovarian cancer. Of these, 1401 had persistent cancer, defined as receiving treatment for ovarian cancer within 6 months of completing their primary treatment, and were excluded. The final sample included 2038 women (See CONSORT Figure 1).
Figure 1.
CONSORT diagram
Treatments
Table 1 shows the characteristics of women by treatment status in the sample and also weighted by the inverse propensity score obtained from multinomial logit regressions. Propensity weighted analyses resulted in similarly matched groups. Of the 1635 women who experience recurrence, 16% were treated with SCS and chemotherapy, 72% with chemotherapy alone and 12% received hospice care. Younger women and those with fewer comorbidities were more likely to undergo surgery. There was no racial difference in treatment receipt. There was geographic variation with higher rates of SCS in the West. On average, women enrolled in hospice were older, had more comorbidities and were more likely to have metastatic disease. Women in the no treatment/no recurrence group tended to be younger, less likely to have stage 4 and have a slightly higher income. Year of diagnosis was not associated with treatment type and rates of SCS did not increase during the 1997 to 2007 time period (p=0.33) (see Table 2). Although there are numerous statistically significant differences in baseline characteristics across treatments in the sample, these differences become much smaller and always statistically insignificant once the sample is weighted using inverse propensity score weights (Table 1). While there was a statistically significant difference in median time from diagnosis to recurrence between women treated for recurrence with chemotherapy vs surgery (643d vs 694d; p=0.007), there remains a significant difference in time between recurrence and death for those treated with chemotherapy vs surgery (660d vs 1097d; p<0.0001).
Table 1.
Sample means by treatment status before and after weighting by propensity score
| Prior to weighting by propensity score | After weighting by propensity score | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||
| Chemotherapy N= 1171 |
Chemotherapy and Surgery N= 265 |
Hospice N= 199 |
No treatment N= 403 |
p value |
Chemotherapy N= 1171 |
Chemotherapy and Surgery N= 265 |
Hospice N= 199 |
No treatment N= 403 |
p value |
|
| Age 65–69 | 29.4 | 40.8 | 17.1 | 30.5 | <0.001 | 29.9 | 31.2 | 25.0 | 30.9 | 0.61 |
| Age 70–74 | 30.0 | 31.3 | 18.6 | 26.8 | 0.002 | 28.1 | 29.3 | 28.8 | 27.7 | 0.98 |
| Age 75–79 | 27.2 | 20.0 | 29.6 | 25.1 | 0.04 | 26.2 | 24.6 | 28.3 | 25.3 | 0.89 |
| Age 80-up | 13.5 | 7.9 | 34.7 | 17.6 | <0.001 | 15.8 | 14.8 | 18.0 | 16.1 | 0.86 |
| White | 91.9 | 92.1 | 93.5 | 87.8 | 0.08 | 91.0 | 90.1 | 93.2 | 90.8 | 0.80 |
| Black | ≤5.0* | 4.2 | <5.5* | 5.7 | 0.37 | ≤5.0* | 6.2 | <5.5* | 4.3 | 0.85 |
| Other minority | ≤5.0 | 3.8 | <5.5* | 6.5 | 0.005 | ≤4.6 | 3.7 | <5.5* | 4.9 | 0.80 |
| Elixhauser comorbidity score | 5.4 | 4.2 | 6.5 | 5.9 | <0.001 | 5.5 | 5.8 | 5.6 | 5.6 | 0.97 |
| Charlson comorbidity score | 0.5 | 0.4 | 0.6 | 0.5 | 0.06 | 0.5 | 0.6 | 0.5 | 0.5 | 0.73 |
| Charlson score 0 | 68.2 | 73.2 | 59.3 | 64.8 | 0.009 | 67.8 | 65.3 | 63.8 | 65.8 | 0.73 |
| Charlson score 1 | 21.9 | 19.2 | 28.6 | 23.6 | 0.11 | 22.0 | 21.4 | 25.4 | 23.2 | 0.80 |
| Charlson score 2+ | 9.9 | 7.5 | 12.1 | 11.7 | 0.24 | 10.2 | 13.3 | 10.8 | 11.0 | 0.83 |
| Stage 3 | 71.5 | 78.5 | 70.9 | 76.7 | 0.03 | 73.4 | 72.9 | 71.0 | 74.1 | 0.94 |
| Stage 4 | 28.5 | 21.5 | 29.1 | 23.3 | 0.03 | 26.6 | 27.1 | 29.0 | 25.9 | 0.94 |
| Serous adenocarcinoma | 85.4 | 81.9 | 85.4 | 80.1 | 0.08 | 83.7 | 82.9 | 84.3 | 82.8 | 0.97 |
| Grade 1 or 2 | 17.4 | 16.6 | 23.1 | 18.6 | 0.30 | 18.1 | 19.1 | 17.9 | 18.1 | 0.99 |
| Grade 3 or 4 | 67.0 | 69.8 | 58.8 | 68.0 | 0.08 | 67.0 | 65.2 | 67.5 | 67.1 | 0.97 |
| Grade unknown | 15.5 | 13.6 | 18.1 | 13.4 | 0.41 | 14.8 | 15.7 | 14.6 | 14.8 | 0.99 |
| Northeast | 20.9 | 17.7 | 12.6 | 22.8 | 0.004 | 20.1 | 16.9 | 15.2 | 20.0 | 0.40 |
| South | 15.4 | 12.5 | 18.6 | 15.1 | 0.34 | 15.4 | 15.2 | 17.4 | 15.9 | 0.94 |
| Midwest | 17.0 | 15.5 | 26.1 | 10.9 | <0.001 | 16.5 | 17.6 | 17.9 | 16.8 | 0.96 |
| West | 46.3 | 52.8 | 42.2 | 50.9 | 0.05 | 47.6 | 48.5 | 47.6 | 47.3 | 1.00 |
| Urban | 54.1 | 56.2 | 44.7 | 59.3 | 0.008 | 54.3 | 55.3 | 55.9 | 53.4 | 0.96 |
| Suburban | 36.1 | 34.7 | 40.7 | 34.5 | 0.49 | 36.3 | 35.5 | 35.6 | 36.0 | 1.00 |
| Rural | 9.8 | 9.1 | 14.6 | 6.2 | 0.01 | 9.4 | 9.2 | 8.5 | 10.6 | 0.91 |
| Zip code median income in $10,000 (Census 2000) | 5.1 | 5.2 | 4.7 | 5.3 | 0.01 | 5.2 | 5.2 | 5.0 | 5.1 | 0.94 |
Cell size <11
Table 2.
Factors associated with type of secondary treatment for patients with ovarian cancer
| Chemotherapy and surgery N=265 |
Hospice N=199 |
No Treatment N=403 |
|
|---|---|---|---|
| Stage 3 | reference | reference | reference |
| Stage 4 | 0.69 (0.53,0.90) | 1.10 (0.77,1.59) | 0.77 (0.63,0.93) |
| Serous adenocarcinoma | 0.82 (0.55,1.23) | 1.00 (0.57,1.76) | 0.66 (0.52,0.83) |
| Other histology | reference | reference | reference |
| Elixhauser comorbidity score | 0.98 (0.96,0.99) | 1.02 (0.98,1.05) | 1.01 (0.99,1.02) |
| Charlson comorbidity score | 0.96 (0.86,1.07) | 1.07 (0.88,1.29) | 1.03 (0.88,1.21) |
| Grade 1 or 2 | reference | reference | reference |
| Grade 3 or 4 | 1.16 (0.96,1.41) | 0.65 (0.53,0.79) | 0.85 (0.68,1.07) |
| Grade unknown | 1.01 (0.59,1.73) | 0.85 (0.52,1.39) | 0.79 (0.64,0.97) |
| Age 65–69 | reference | reference | reference |
| Age 70–74 | 0.74 (0.60,0.90) | 1.03 (0.70,1.52) | 0.89 (0.63,1.24) |
| Age 75–79 | 0.56 (0.36,0.85) | 1.84 (1.08,3.15) | 0.92 (0.70,1.22) |
| Age 80 up | 0.43 (0.30,0.60) | 4.45 (2.67,7.41) | 1.24 (0.77,2.01) |
| Zip code median income (Census 2000) | 1.01 (0.93,1.10) | 0.97 (0.89,1.06) | 1.00 (0.96,1.05) |
| Urban | reference | reference | reference |
| Suburban | 0.94 (0.75,1.19) | 1.34 (1.00,1.79) | 0.86 (0.73,1.01) |
| Rural | 0.98 (0.61,1.57) | 1.58 (1.05,2.39) | 0.70 (0.50,0.97) |
| Northeast | reference | reference | reference |
| South | 0.83 (0.57,1.21) | 1.90 (1.56,2.32) | 1.02 (0.76,1.37) |
| Midwest | 0.96 (0.67,1.37) | 2.35 (2.06,2.68) | 0.76 (0.69,0.84) |
| West | 1.16 (0.88,1.51) | 1.63 (1.29,2.07) | 1.06 (0.94,1.19) |
| White | reference | reference | reference |
| Black | 1.27 (0.74,2.18) | 1.26 (0.65,2.46) | 1.64 (0.99,2.72) |
| Other minority | 0.74 (0.52,1.03) | 0.38 (0.17,0.83) | 1.55 (1.10,2.17) |
| Diagnoses year 1997 | reference | reference | reference |
| Diagnoses year 1998 | 1.16 (0.54,2.50) | 1.19 (0.55,2.58) | 1.64 (0.99,2.71) |
| Diagnoses year 1999 | 0.56 (0.35,0.89) | 0.63 (0.13,3.04) | 1.66 (0.65,4.29) |
| Diagnoses year 2000 | 0.69 (0.29,1.67) | 0.87 (0.41,1.84) | 1.52 (0.53,4.33) |
| Diagnoses year 2001 | 0.90 (0.48,1.67) | 1.17 (0.51,2.68) | 1.53 (0.78,3.01) |
| Diagnoses year 2002 | 0.93 (0.48,1.81) | 1.05 (0.49,2.23) | 1.68 (0.83,3.40) |
| Diagnoses year 2003 | 0.92 (0.48,1.76) | 0.70 (0.31,1.58) | 1.82 (0.85,3.92) |
| Diagnoses year 2004 | 0.98 (0.52,1.85) | 0.75 (0.41,1.38) | 2.53 (1.20,5.34) |
| Diagnoses year 2005 | 0.61 (0.26,1.46) | 0.96 (0.37,2.49) | 3.08 (1.49,6.39) |
| Diagnoses year 2006 | 0.51 (0.20,1.28) | 1.30 (0.45,3.78) | 4.78 (2.38,9.60) |
| Diagnoses year 2007 | 0.44 (0.19,0.98) | 0.43 (0.13,1.44) | 10.13 (4.89,20.95) |
Reference group: Chemotherapy only (N=1171)
Survival Analyses
Actual survival is shown in Figure 2. Survival for women who underwent secondary debulking with chemotherapy was better compared to those who received chemotherapy only. In both unadjusted for propensity of treatment and propensity weighted analyses, treatment type, and younger age were associated with improved survival (Table 3). Propensity score adjusted log logistic analyses showed that women undergoing surgery with chemotherapy had significantly greater survival compared to those receiving chemotherapy alone (hazard ratio [HR]: 1.33, 95% confidence interval [CI]:1.20,1.47. Women receiving no secondary treatment, classified as nonrecurrent, had significantly better survival (HR: 2.33, 95% CI:1.97,2.76), while women in hospice had significantly lower survival than those who underwent chemotherapy alone (HR=0.55, 95% CI: 0.51,0.59). The estimated median survival of women treated with chemotherapy was 4.1 years from time of diagnosis; 77% of this group died within the study period. Those treated with secondary surgery and chemotherapy survived a median of 5.4 years; 67% died. Those receiving hospice survived an average of 2.2 years; 99% died. The 403 women who received no secondary treatments were classified as nonrecurrent and had a median survival of 9.3 years; 19.6% died (see Figure 3).
Figure 2.
Actual Survival by Treatment Type
Table 3.
Factors Associated with Survival
| Log-Logistic Model HR (95%CI) | FMM True Non- Recurrence HR (95%CI) | FMM Untreated Recurrence HR (95%CI) | p-value of Difference | |
|---|---|---|---|---|
| Chemotherapy only | reference | reference | reference | 0.253 |
| Chemotherapy & surgery | 1.33 (1.20,1.46) | 1.24 (1.05,1.46) | 1.67 (1.13,2.47) | |
| Hospice | 0.55 (0.51,0.59) | 0.54 (0.49,0.60) | 0.54 (0.36,0.79) | 0.952 |
| No Treatment | 2.33 (1.97,2.75) | 3.91 (2.88,5.31) | 0.57 (0.42,0.78) | 0.000 |
| Age 65–69 | reference | reference | reference | |
| Age 70–74 | 0.91 (0.82,1.00) | 1.01 (0.90,1.13) | 0.62 (0.40,0.96) | |
| Age 75–79 | 0.83 (0.71,0.95) | 0.91 (0.82,1.01) | 0.70 (0.62,0.80) | |
| Age 80+ | 0.80 (0.68,0.94) | 0.85 (0.69,1.03) | 0.94 (0.38,2.36) | |
| White | reference | reference | reference | |
| Black | 0.87 (0.75,1.02) | 0.95 (0.79,1.15) | 0.66 (0.46,0.93) | |
| Other minority | 0.93 (0.83,1.05) | 1.00 (0.85,1.18) | 0.89 (0.52,1.52) | |
| Stage 3 | reference | reference | reference | |
| Stage 4 | 0.98 (0.91,1.05) | 0.96 (0.86,1.07) | 1.12 (0.89,1.42) | |
| Serous adenocarcinoma | 0.88 (0.81,0.95) | 0.84 (0.73,0.97) | 1.19 (0.80,1.77) | |
| Other histology | reference | reference | reference | |
| Elixhauser comorbidity score | 0.99 (0.99,1.00) | 1.00 (0.99,1.00) | 1.00 (0.99,1.02) | |
| Charlson comorbidity score | 0.91 (0.87,0.96) | 0.91 (0.85,0.97) | 1.04 (0.81,1.33) | |
| Grade 1 or 2 | reference | reference | reference | |
| Grade 3 or 4 | 1.06 (0.96,1.17) | 1.00 (0.92,1.09) | 1.32 (0.92,1.89) | |
| Grade unknown | 0.99 (0.90,1.10) | 1.08 (0.94,1.24) | 0.73 (0.50,1.05) | |
| Zip code median income (Census 2000) | 1.00 (0.98,1.02) | 0.99 (0.97,1.02) | 1.04 (1.00,1.08) | |
| Urban | reference | reference | reference | |
| Suburban | 0.93 (0.80,1.10) | 0.91 (0.80,1.03) | 1.07 (0.62,1.83) | |
| Rural | 0.86 (0.74,1.00) | 0.85 (0.74,0.97) | 1.24 (0.82,1.89) | |
| Northeast | reference | reference | reference | |
| South | 1.01 (0.94,1.09) | 1.05 (0.89,1.23) | 1.12 (0.75,1.68) | |
| Midwest | 0.95 (0.87,1.04) | 1.04 (0.94,1.14) | 0.95 (0.74,1.22) | |
| West | 1.07 (0.96,1.19) | 1.18 (1.04,1.35) | 0.76 (0.44,1.34) | |
| Diagnoses year 1997 | reference | reference | reference | |
| 1998 | 0.90 (0.74,1.11) | 0.88 (0.75,1.03) | 1.37 (0.91,2.07) | |
| 1999 | 0.93 (0.78,1.11) | 0.99 (0.79,1.23) | 1.66 (0.88,3.15) | |
| 2000 | 0.94 (0.76,1.16) | 1.07 (0.91,1.26) | 1.21 (0.90,1.62) | |
| 2001 | 0.95 (0.85,1.05) | 0.97 (0.83,1.13) | 1.76 (1.22,2.54) | |
| 2002 | 0.93 (0.84,1.02) | 0.85 (0.71,1.02) | 2.22 (1.76,2.80) | |
| 2003 | 0.98 (0.88,1.09) | 0.94 (0.84,1.07) | 2.48 (1.61,3.83) | |
| 2004 | 1.02 (0.82,1.27) | 1.20 (0.95,1.51) | 1.27 (0.94,1.72) | |
| 2005 | 0.91 (0.74,1.12) | 0.97 (0.78,1.22) | 1.37 (1.13,1.67) | |
| 2006 | 0.92 (0.78,1.08) | 0.94 (0.80,1.12) | 1.30 (0.81,2.08) | |
| 2007 | 0.84 (0.72,0.98) | 0.78 (0.64,0.95) | 2.16 (0.74,6.26) | |
| Class probability | 0.77 (0.68,0.85) | 0.23 (0.15,0.32) |
FMM: Finite Mixture Model
Figure 3.
Predicted Survival by Treatment Type
However, the group that received no secondary treatments and was classified as nonrecurrent, likely included women who were truly nonrecurrent and those with recurrence who went untreated and were misclassified. Estimation of a 2-class finite mixture log-logistic survival model, including the factors listed in Table 1, identified two distinct classes of women with different survival (Table 3) with associated class probabilities of 75% (95% CI: 62, 85%) and 25% (95% CI: 18, 35%), respectively. The survival ratios, relative to chemotherapy, for observations in the two classes are 3.91 (95% CI: 2.88,5.31) for the truly nonrecurrent and 0.57 (95% CI: 0.42,0.78) for those misclassified as nonrecurrent who went untreated. Survival ratios for chemotherapy and surgery and for hospice are not significantly different across the two classes and are qualitatively very similar to those obtained from the standard log-logistic survival model described above. However, survival ratios for the no-treatment group is significantly different between the log logistic and finite mixture models (HR=2.3 vs 3.9 vs 0.57; p<0.000). Median survival was 15.9 years for the truly nonrecurrent with longer survival as compared to 2.4 years in the group misclassified as nonrecurrent, those with shorter survival (Figure 3). Of the untreated women who were likely misclassified as nonrecurrent cancer, 55% listed cancer as their cause of death.
Most of the control covariates were not statistically significant determinants of survival. However, women with more comorbidities, higher grade cancer or older had worse survival (Table 3).
Discussion
Ovarian cancer patients are living longer than they did 30 years ago with 5 year survival rates increasing from 20% to 40%.[19,20] While some of the improved survival is likely due to platinum-based therapies, the role of secondary surgery in improving survival has been uncertain. Many case studies and a recent Cochrane review suggest its effectiveness[4] but there has not been a population based study that takes into account the potential observational biases inherent in cohort studies.[20] Our population-based findings suggest that treating first recurrence of ovarian cancer with secondary surgery and chemotherapy increases survival by >1 year compared to chemotherapy alone. While randomized controlled trial data are the gold standard for assessing treatment efficacy, eligibility limitations often make it difficult to apply findings to broader populations, particularly the elderly. Comparing effectiveness of treatments using population-based data can help inform physician and patient decision-making pending results of ongoing randomized phase 3 trials.[18] Our findings pertain to women 66 years and older who represent a substantial portion of ovarian cancer in the US. Yet, we expect these findings may be useful for younger women who tend to have less comorbidity and may be better surgical candidates.
Our finding of less aggressive treatments among older and sicker patients is consistent with the literature of primary treatment of ovarian cancer.[21–25] While others have described a racial disparity in primary treatment of ovarian cancer,[21,26,27] we did not see this pattern for secondary treatments, perhaps because we limited this analysis to those patients who received optimal primary therapy.
A key challenge to evaluating the effectiveness of secondary treatments lies in the difficulty of accurately identifying secondary surgery and recurrent cancer. We approached and overcame these challenges with 2 methods – the first, a chart validation study that yielded excellent sensitivity and specificity of the coding algorithm. Although the algorithm was based on data billing codes from across the country and appears accurate, it was validated using a single institution’s charts and may not work as well in other settings. Secondary surgery codes are further confounded by second look procedures which, despite improvements in imaging techniques, continue to be done to ensure that no lesions remain. This surgery can include removal of tissue for pathologic review which can be for diagnostic and treatment purposes. Because the vast majority of secondary surgeries in our cohort were accompanied by chemotherapy, it is unlikely that there was much misclassification. Similarly, we may have misclassified women with persistent cancer as they may have been receiving ongoing maintenance therapy. We believe such misclassification is unlikely since this group had a shortened median survival time (1.6 years), which is consistent with persistent disease. In addition, to further maximize our ability to assess effectiveness of secondary surgery for recurrent disease, we excluded women with persistent disease – those who continued to receive chemotherapy beyond the usual 6 months post-primary surgery. That our sample included only 265 women undergoing secondary surgery yet finds significant survival differences suggests the potential strength of the association found. Rates of performance may differ in a younger population. As with all administrative claims based analyses, our findings are limited by the inherent unmeasured biases of observational data. We used standard inverse propensity weighted analytic methods to reduce the impact of such biases, but may not eliminate them.
We recognize the inherent difficulty of classifying recurrent cancer by the presence of claims for treatment. Particularly in an elderly population, many women will go untreated when cancer recurs resulting in distorted survival estimates. To counter this challenge, we used finite mixture models, a novel approach to separate out the truly nonrecurrent from those with recurrence who forego active or hospice treatment. By utilizing this approach with persistent cancer, we obtain more accurate measures of survival among women who truly did not experience a recurrence, a count that is not obscured by the inclusion of those with recurrence who went untreated.
In summary, for women with advanced ovarian cancer initially treated with surgery and chemotherapy who experience a recurrence, secondary cytoreductive surgery with chemotherapy appears to provide a significant survival advantage over chemotherapy alone and may be considered a standard of care pending results of ongoing randomized trials.
Acknowledgments
Dedicated to and in loving memory of Erica Bickell whose question sparked this work. This work was supported by NCI R01CA157176.
Funded by: NCI 5R01CA157176
Appendix 1. ICD9, HCPCS, CPT codes for ovarian cancer surgery
Surgery
ICD 9 Procedure Codes with or without a diagnosis of ovarian cancer
65.31- Lap Unilat Oophorectomy
65.39- Oth Unilat Oophorectomy
65.41- Lap Uni Salpingo-Oophor
65.49- Oth Uni Salpingo-Oophor
65.51- Oth Remove Both Ovaries
65.52-Oth Remove Remain Ovary
65.53- Lap Remove Both Ovaries
65.54- Lap Remove Remain Ovary
65.61- Oth Remove Ovaries/Tubes
65.62-Oth Remove Rem Ova/Tube
65.63- Lap Remove Ovaries/Tubes
65.64- Lap Remove Rem Ova/Tube
ICD9 procedure codes only with a diagnoses of ovarian cancer (183.XX):
40.3X- Regional lymph node excision
40.5X- Radical excision of other lymph nodes
54.3X- Excision/destruction abdominal wall lesion
54.4X- Omentectomy, excision, destruction of the peritoneal tissue
68.3X- Subtotal Abdominal Hysterectomy
68.4X- Total Abdominal Hysterectomy
68.5X- Vaginal Hysterectomy
68.6X- Radical Abdominal Hysterectomy
68.7X- Radical Vaginal Hysterectomy
68.8X- Pelvic Evisceration
68.9X- Other and Unspecified Hysterectomy
70.32- Excision/destruction cul de sac lesion
HCPCS codes with or without a diagnosis of ovarian cancer:
58180- Supracervical abdominal hysterectomy (subtotal hysterectomy), with or without removal of tube(s), with or without removal of ovary
58200- Total abdominal hysterectomy, including partial vaginectomy, with para-arotic and pelvic lymph node sampling, with or without removal of tube(s), with or without removal of ovary(s)
58210- Radical abdominal hysterectomy, with bilateral total pelvic lymphadenectomy and para-arotic lymph node sampling (biopsy) with or without removal of tune(s), with or without removal of ovary(s)
58940- Oophhorectomy, partial or total, unilateral or bilateral
58943- Oophhorectomy, partial or total, unilateral or bilateral; for ovarian, tubal or primary peritoneal malignancy, with para-aortic and pelvic lymph node biopsies, peritoneal washings, peritoneal biopsies, diaphragmatic assessments, with or without salpingectomy(s), with or without omentectomy
58950- Resection (initial) of ovarian, tubal or primary peritoneal malignancy with bilateral salpingo-oophorectomy and omentectomy
58951- Resection of ovarian, tubal or primary peritoneal malignancy with bilateral salpingo-oophorectomy and omentectomy; with total abdominal hysterectomy, pelvic and limited para-aortic lymphademectomy
58952- Resection of ovarian, tubal or primary peritoneal malignancy with bilateral salpingo-oophorectomy and omentectomy; with radical dissection for debulking (ie, radical excision or destruction, intra-abdominal or retroperitoneal tumors)
58953- Bilateral salpingo-oophorectomy with omentectomy, total abdominal hysterectomy and radical dissection for debulking
58954- Bilateral salpingo-oophorectomy with omentectomy, total abdominal hysterectomy and radical dissection for debulking; with pelvic lymphadenctomy and limited para-aortic lymphadenctomy
58956- Total abdominal hysterectomy, BSO with malignancy
HCPCS codes with a diagnosis of ovarian cancer (183.XX):
38562- Limited lymphadenectomy for staging (separate procedure)
38570- Laparoscopy, surgical; with retroperitoneal lymph node sampling (biopsy) single or multiple
38571- Laparoscopy, surgical; with bilateral total pelvic lymphadenectomy
38572-Laparoscopy, surgical; with bilateral total pelvic lymphadenectomy and peri-aortic lymph node sampling, single or multiple
44140- Colectomy
44200- Laparoscopy, surgical; enterolysis (freeing of intestinal adhesion) (separate procedure)
44950- Appendectomy
44970- Laparoscopy, surgical appendectomy
49000- Exploratory laparotomy, exploratory ceilotomy with or without biopsy (s) (separate procedure)
49002- Reopening of recent laparotomy
49010- Exploration, retroperitoneal area with or without biopsy (separate procedure)
49320- Laparoscopy, abdomen, peritoneum, and omentum, diagnostic, with or without collection of specimen (s) by brushing or washing (separate procedure)
49321- Laparoscopy, surgical; with biopsy (single or multiple)
49322- Laparoscopy, surgical; with aspiration of cavity or cyst (eg. Ovarian cyst) (single or multiple)
58150- Total abdominal hysterectomy (corpus and cervix) with or without removal of tube(s), ovary(s)
58550- Laparoscopy, surgical, with vaginal hysterectomy, for uterus 250 grams or less
58552- Laparoscopy, surgical, with vaginal hysterectomy, for uterus 250 grams or less, with removal of tubes and/or ovary
58553- Laparoscopy, surgical, with vaginal hysterectomy, for uterus greater than 250 grams
58554- Laparoscopy, surgical, with vaginal hysterectomy, for uterus greater than 250 grams, with removal of tubes and/or ovary
58570- Laparoscopy, surgical, with total hysterectomy, for uterus 250 grams or less
58571- Laparoscopy, surgical, with total hysterectomy, for uterus 250 grams or less, with removal of tubes and/or ovary
58572- Laparoscopy, surgical, with total hysterectomy, for uterus greater than 250 grams
58573- Laparoscopy, surgical, with total hysterectomy, for uterus greater than 250 grams, with removal of tubes and/or ovary
58660- Laparoscopy, surgical; with lysis of adhesions (salpingolysis, ovariolysis) (separate procedure)
58661- Laparoscopy, surgical, with retroperitoneal lymph node sampling (biopsy) single or multiple
58662- Laparoscopy, surgical, with retroperitoneal lymph node sampling (biopsy) single or multiple
58700- Salpingectomy, complete or partial
58720- Salpingo-oophorectomy, complete or partial (separate procedure)
58925- Closure of vesicouterine fistula; with hysterectomy
58960- Laparoscopy for staging or restaging or ovarian, tubal or primary peritoneal malignancy (second look) with or without omentectomy, peritoneal washing, biopsy of abdominal and pelvic peritoneum, diaphragmatic assessment with pelvic and limited para-aorti
Appendix 2: Chemotherapy Codes
ICD9 procedure codes:
99.25- Infusion of therapeutic substance into intraperitoneal cavity
ICD9 diagnoses codes:
V58.1X- Encounter for chemotherapy and immunotherapy for neoplastic conditions
V66.2X- Convalescence and palliative care following chemotherapy
V67.2X- Follow-up examination following chemotherapy
E93.07 – body measurement
DRG code:
410-
HCPCS codes:
964XX-Administration
965XX- Administration
Q0083- Chemotherapy
Q0084- Administration
G0356- Chemo injection, sq or intramuscular, hormonal agent
G0357- Chemo IV push, single drug
G0358- Administration of each additional pushed chemo drug
G0359- Chemo IV infusion, single/initial drug, initial hour
G0360- Each additional hour of infusion (up to 8 hrs)
G0361- Initiation of prolonged chemo (>8 hrs)
G0362- Administration of each additional infused chemo drug, up to 1 hr
G9021- Chemotherapy assessment for nausea and/or vomiting, patient reported, performed at the time of chemotherapy administration; assessment level one: not at all
G9022- Chemotherapy assessment for nausea and/or vomiting, patient reported, performed at the time of chemotherapy administration; assessment level two: a little
G9023- Chemotherapy assessment for nausea and/or vomiting, patient reported, performed at the time of chemotherapy administration; assessment level three: quite a bit
G9024- Chemotherapy assessment for nausea and/or vomiting, patient reported, performed at the time of chemotherapy administration; assessment level four: very much
G9025- Chemotherapy assessment for pain, patient reported, performed at the time of chemotherapy administration, assessment level one: not at all
G9026- Chemotherapy assessment for pain, patient reported, performed at the time of chemotherapy administration, assessment level two: a little
G9027- Chemotherapy assessment for pain, patient reported, performed at the time of chemotherapy administration, assessment level three: quite a bit
G9028- Chemotherapy assessment for pain, patient reported, performed at the time of chemotherapy administration, assessment level four: very much
G9029- Chemotherapy assessment for lack of energy (fatigue), patient reported, performed at the time of chemotherapy administration, assessment level one: not at all
G9030- Chemotherapy assessment for lack of energy (fatigue), patient reported, performed at the time of chemotherapy administration, assessment level two: a little bit
G9031- Chemotherapy assessment for lack of energy (fatigue), patient reported, performed at the time of chemotherapy administration, assessment level three: quite a bit
G9032- Chemotherapy assessment for lack of energy (fatigue), patient reported, performed at the time of chemotherapy administration, assessment level four: very much
- Neulasta 6 mg-
- HCPCS: J2505
- NDC: 54868522900, 55513019001
- Topotecan-
- HCPCS: J8705, J9350
- NDC: 00007420101, 00007420105
- Cyclophosphomide (Cytoxan) 25 mg-
- HCPCS: J9070, J9080, J9090, J9091, J9092, J9093, J9094, J9095, J9096, J9097
- NDC: 00013560693, 00015053941, 00013561693, 00015054641, 00013562693, 00015054712, 00015054741, 10019095501, 00013563670, 00015050541, 00015054812, 00015054841, 10019095601, 38779050603, 38779050604, 38779050605, 00013564670, 00015050641, 00015054912, 00015054941, 10019095701
- Etoposide (Toposar)/Etopophos/VePesid-
- HCPCS: J8560, J9181
- NDC: 00013733691, 00013734694, 00013735688, 00015306120, 00015306220, 00015308420, 00015309520, 00015340420, 00703565301, 00703565601, 00703565701, 10019093001, 10019093002, 55390029101, 55390029201, 55390029301, 55390049101, 55390049201, 55390049301, 63323010405, 63323010425, 63323010450
- Adriamycin 10 mg-
- HCPCS: J9000
- NDC: 00013108691, 00013109691, 00013110679, 00013111683, 00013113691, 00013114691, 00013115679, 00013116683, 00015335222, 00015335322, 00703504001, 00703504303, 00703504601, 10019092001, 10019092102, 38779065206, 38779065209, 49452243701, 55390023110, 55390023210, 55390023301, 55390023510, 55390023610, 55390023701, 55390023801, 55390024110, 55390024210, 55390024301, 55390024510, 55390024610, 55390024701, 55390024801, 63323010161, 63323088305, 63323088310, 63323088330
- Doxil 10 mg-
- HCPCS: J9001
- NDC: 17314960001, 17314960002
- Bevacizumab (Avastin) 10 mg/Bevacizumab-
- HCPCS: J9035, J9999
- NDC: 50242006001, 50242006101, 50242006002
- Carboplatin 50 mg/Paraplatin-
- HCPCS: J9045
- NDC: 00015321030, 00015321076, 00015321130, 00015321130, 00015321176, 00015321230, 00015321276, 00015321330, 00015321430, 00015321530, 00015321630, 00015323011, 00015323111, 00015323211, 00409112911, 00409112912, 00591333626, 0591333712, 00591333889, 00591345460, 00703324411, 00703324611, 00703326401, 00703326601, 00703326801, 00703326871, 00703327401, 00703327601, 00703327801, 00703424401, 00703424601, 00703424801, 10019091201, 10019091202, 10019091203, 10019091501, 10019091601, 10019091701, 10139006005, 10139006015, 10139006045, 15210006112, 15210006312, 15210006612, 15210006712, 50111096576, 50111096676, 50111096776, 55390015001, 55390015101, 55390015201, 55390015301, 55390015401, 55390015501, 55390015601, 55390022001, 55390022101, 55390022201, 61703033922, 61703033950, 61703033956, 61703036018, 61703036022, 61703036050, 63323016610, 63323016720, 63323016721, 63323016800, 63323017245, 63323017260, 67817006112, 67817006312, 67817006612, 67817006712,
- Cisplatin/Platinol-
- HCPCS: J9060, J9062
- NDC: 00015322022, 00015322122, 00703574711, 00703574811, 10019091001, 10019091002, 49452207801, 49452207802, 49452207803, 51552107609, 55390009901, 55390011250, 55390011299, 55390018701, 55390041450, 55390041499, 63323010351, 63323010364, 63323010365, 63323010391, 63323010395
- Taxotere 20 mg/Docetaxel-
- HCPCS: J9170
- NDC: 00075800120, 00075800180
- Gemzar/Gemcitabine-
- HCPCS: J9201
- NDC: 00002750101, 00002750101, 00002750201
- Taxol/Paclitaxel-
- HCPCS: J9264, J9265
- NDC: 68817013450, 00015347530, 00015347630, 00015347911, 00074433501, 00074433502, 00074433504, 00172375377, 00172375396, 00172375473, 00172375494, 00172375531, 00172375675, 00172375695, 00555198414, 00555198514, 10518010207, 10518010208, 10518010209, 51079096101, 51079096201, 51079096301, 55390011405, 55390011420, 55390011450, 55390030405, 55390030420, 55390030450, 55390031405, 55390031420, 55390031450, 61703034209, 61703034222, 61703034250, 66758004301, 66758004302, 66758004303, 68817013450
- Toptecan/Aromazin/Megace/Ceenu-
- HCPCS: J8999
- NDC: 00007420511, 00007420711, 00054224725, 0054458111(missing a zero. Code should be 00054458111), 00054458127, 00054483121, 00054483126, 00054483413, 00054483422, 00054824725, 00054883125, 00054860425, 00093078201, 00093078205, 00093078210, 00093078256, 00093078405, 00093078406, 00093078410, 00093078486, 00093551006, 00172565658, 00172565649, 00172565670, 00172565680, 00172565746, 00172565760, 00172565770, 00172565780, 00310060018, 00310060060, 00310060075, 00310060412, 00310060430, 00310060490, 00310073060, 00310073130, 00378014405, 00378014491, 00378027401, 00378027493, 00378354725, 00378354752, 00555044605, 00555044609, 00555044663, 00555088202, 00555090401, 00555090405, 00555090414, 00603394621, 00677168001, 13632012301, 49884072401, 49884092202, 49884092204, 54569037800, 54569376500, 54569376501, 54569571500, 54569571600, 54569585700, 54868136700, 54868300401, 54868300402, 54868300403, 54868300404, 54868300405, 54868428700, 54868428701, 54868428702, 54868428703, 54868428704, 54868477300, 54868477301, 54868477302, 54868477303, 54868528200, 54868528201, 55289058530, 57866443601, 57866661501, 57866661801, 58016065760, 63739026910, 63739026915, 66105083201, 66105083203, 66105083206, 66105083209, 66105083210
- Hydoxyurea, Hydrea, Droxia-
- NCD: 00003083050, 00003633517, 00003633617, 00003633717
- Ifosfamide/Ifex/Mesnex Kit-
- NCD: 00015355410, 00015355427, 00015355610, 00015355626, 00015356410, 0001-356415, 00703410048, 00703410058, 00703410068, 00703410948, 00703410958, 00703410968
Footnotes
Conflict of Interest Statement: Dr. Wisnivesky is a member of the research board of EHE International, has received consulting honorarium from Merck, BMS, and Quintiles and research grants from Aventis and Quorum. All other authors do not have a conflict of interest to disclose.
References
- 1.Coleman RL, Monk BJ, Sood AK, et al. Nature Reviews Clinical Oncology SEER Cancer Statistics Factsheets: Ovary Cancer. National Cancer Institute; Bethesda, MD: Latest research and treatment of advanced-stage epithelial ovarian cancer. http://seer.cancer.gov/statfacts/html/ovary.html. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Janicke F, Holscher M, Kuhn W, von Hugo R, Pache L, Siewert JR, et al. Radical surgical procedure improves survival time in patients with recurrent ovarian cancer. Cancer. 1992;70(8):2129–2136. doi: 10.1002/1097-0142(19921015)70:8<2129::aid-cncr2820700820>3.0.co;2-u. [DOI] [PubMed] [Google Scholar]
- 3.The NCCN Ovarian Cancer Clinical Practice Guidelines in Oncology (Version 1.2008). © 2006 National Comprehensive Cancer Network, Inc. Available at: http://www.nccn.org.
- 4.Al Rawahi T, Lopes AD, Bristow RE, Bryant A, Elattar A, Chattopadhyay S, Galaal K. Surgical cytoreduction for recurrent epithelial ovarian cancer. Cochrane Database of Systematic Reviews. 2013;(2) doi: 10.1002/14651858.CD008765.pub3. Art. No.: CD008765. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Livaudais JC, Egorova N, Hayes MP, Howell EA, Wisnivesky JP, Franco R, Bickell NA. Using hospital claims data to identify ovarian cancer recurrence: A validation study. J Clin Oncol. 2012;30(15S):e15513. [Google Scholar]
- 6.Chi DS, McCaughty K, Diaz JP, Huh J, Schwabenbauer S, Hummer AJ, Venkatraman ES, Aghajanian C, Sonoda Y, Abu-Rustum NR, Barakat RR. Guidelines and selection criteria for secondary cytoreductive surgery in patients with recurrent, platinum-sensitive epithelial ovarian carcinoma. Cancer. 2006 May 1;106(9):1933–9. doi: 10.1002/cncr.21845. [DOI] [PubMed] [Google Scholar]
- 7.Eisenkop SM, Friedman RL, Spirtos NM. The role of secondary cytoreductive surgery in the treatment of patients with recurrent epithelial ovarian carcinoma. Cancer. 2000;88:144–153. doi: 10.1002/(sici)1097-0142(20000101)88:1<144::aid-cncr20>3.3.co;2-o. [DOI] [PubMed] [Google Scholar]
- 8.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998 Jan;36(1):8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 9.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–83. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 10.Imbens GW. The role of the propensity score in estimating dose-response functions. Biometrika. 2000;87(3):706–710. doi: 10.1093/biomet/87.3.706. [DOI] [Google Scholar]
- 11.Lunceford JK, Davidian M. Stratification and weighting via the propensity score in estimation of causal treatment effects: A comparative study. Statistics in Medicine. 2004;23:2937–2960. doi: 10.1002/sim.1903. [DOI] [PubMed] [Google Scholar]
- 12.Huang L, Johnson KA, Mariotto AB, Dignam JJ, Feuer EJ. Population-based survival-cure analysis of ER-negative breast cancer. Breast Cancer Research and Treatment. 2010;123(1):257–264. doi: 10.1007/s10549-010-0752-z. [DOI] [PubMed] [Google Scholar]
- 13.Joulain F, Proskorovsky I, Allegra C, Tabernero J, Hoyle M, Iqbal SU, Van Cutsem E. Mean overall survival gain with aflibercept plus FOLFIRI vs placebo plus FOLFIRI in patients with previously treated metastatic colorectal cancer. British Journal of Cancer. 2013;109(7):1735–1743. doi: 10.1038/bjc.2013.523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Conway KS, Deb P. Is prenatal care really ineffective? Or, is the ‘devil’ in the distribution? Journal of Health Economics. 2005;24(3):489–513. doi: 10.1016/j.jhealeco.2004.09.012. [DOI] [PubMed] [Google Scholar]
- 15.McLachlan G, Peel D. Finite mixture models. John Wiley & Sons; 2004. [Google Scholar]
- 16.Wedel M, Desarbo WS, Bult JR, Ramaswamy V. A latent class poisson regression model for heterogeneous count data. Journal of Applied Econometrics. 1993;8(4):397–411. [Google Scholar]
- 17.Deb P, Trivedi PK. Demand for medical care by the elderly: a finite mixture approach. Journal of Applied Econometrics. 1997;12(3):313–336. [Google Scholar]
- 18.van de Laar Rafli, Zusterzeel Petra LM, Van Gorp Toon, Buist Marrije R, van Driel Willemien J, Gaarenstroom Katja N, Arts Henriette JG, van Huisseling Johannes CM, Hermans Ralph HM, Pijnenborg Johanna MA, Schutter Eltjo MJ, Pelikan Harold MP, Vollebergh Jos HA, Engelen Mirjam JA, IntHout Joanna, Kruitwagen Roy FPM, Massuger Leon FAG. Cytoreductive surgery followed by chemotherapy versus chemotherapy alone for recurrent platinum-sensitive epithelial ovarian cancer (SOCceR trial): a multicenter randomised controlled study. BMC Cancer. 2014;14:22. doi: 10.1186/1471-2407-14-22. GOG 0213 ( clinicaltrials.gov NCT No: NCT00565851) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Britain) NIfHaCEG, Britain) NCCfCG. Ovarian cancer: the recognition and initial management of ovarian cancer [electronic resource]/developed for NICE by the National Collaborating Centre for Cancer. Cardiff, Wales: National Collaborating Centre for Cancer; 2011. [PubMed] [Google Scholar]
- 20.Welch HG, Schwartz LM, Woloshin S. Are increasing 5-year survival rates evidence of success against cancer? JAMA. 2000;283(22):2975–8. doi: 10.1001/jama.283.22.2975. [DOI] [PubMed] [Google Scholar]
- 21.Howell EA, Egorova N, Hayes MP, Wisnivesky J, Franco R, Bickell N. Racial disparities in the treatment of advanced epithelial ovarian cancer. Obstet Gynecol. 2013 Nov;122(5):1025–32. doi: 10.1097/AOG.0b013e3182a92011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Chan JK, Zhang M, Hu JM, Shin JY, Osann K, Kapp DS. Racial disparities in surgical treatment and survival of epithelial ovarian cancer in United States. J Surg Oncol. 2008 Feb 1;97(2):103–7. doi: 10.1002/jso.20932. [DOI] [PubMed] [Google Scholar]
- 23.Bristow RE, Zahurak ML, Ibeanu OA. Racial disparities in ovarian cancer surgical care: a population-based analysis. Gynecol Oncol. 2011 May 1;121(2):364–8. doi: 10.1016/j.ygyno.2010.12.347. Epub 2011 Feb 1. [DOI] [PubMed] [Google Scholar]
- 24.Thrall MM, Gray HJ, Symons RG, Weiss NS, Flum DR, Goff BA. Trends in treatment of advanced epithelial ovarian cancer in the Medicare population. Gynecol Oncol. 2011 Jul;122(1):100–6. doi: 10.1016/j.ygyno.2011.03.022. Epub 2011 Apr 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.van Altena AM, Karim-Kos HE, de Vries E, Kruitwagen RF, Massuger LF, Kiemeney LA. Trends in therapy and survival of advanced stage epithelial ovarian cancer patients in the Netherlands. Gynecol Oncol. 2012 Jun;125(3):649–54. doi: 10.1016/j.ygyno.2012.02.033. [DOI] [PubMed] [Google Scholar]
- 26.Liu FW, Randall LM, Tewari KS, Bristow RE. Racial disparities and patterns of ovarian cancer surgical care in California. Gynecol Oncol. 2014 Jan;132(1):221–6. doi: 10.1016/j.ygyno.2013.08.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Usach I, Blansit K, Chen LM, Ueda S, Brooks R, Kapp DS, Chan JK. Survival differences in women with serous tubal, ovarian, peritoneal, and uterine carcinomas. Am J Obstet Gynecol. 2014 Aug 18; doi: 10.1016/j.ajog.2014.08.016. pii: S0002-9378(14)00847-3. [DOI] [PubMed] [Google Scholar]