Comparison of outcomes in early stage uterine carcinosarcoma and uterine serous carcinoma

Gynecologic Oncology 135 (2014) 49–53

Contents lists available at ScienceDirect

Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno

Comparison of outcomes in early stage uterine carcinosarcoma and uterine serous carcinoma Neil B. Desai a, Marisa A. Kollmeier a, Vicky Makker b, Douglas A. Levine c, Nadeem R. Abu-Rustum c, Kaled M. Alektiar a,⁎ a b c

Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, USA Gynecologic Medical Oncology Service, Memorial Sloan Kettering Cancer Center and Department of Medicine, Weill Cornell Medical College, New York, NY, USA Gynecology Service, Memorial Sloan Kettering Cancer Center, New York, NY, USA

H I G H L I G H T S • 172 patients with stage I-II UPSC or CS treated with surgery +/- adjuvant therapy from 2000–2011. • Patients with early stage CS had worse outcomes than those with UPSC and were highlighted by early relapses. • However, among 111 (65%) patients able to receive IVRT and chemotherapy, outcomes were no longer statistically different.

a r t i c l e

i n f o

Article history: Received 3 June 2014 Accepted 23 July 2014 Available online 30 July 2014 Keywords: Uterine papillary serous carcinoma Uterine serous carcinoma Carcinosarcoma Intravaginal radiotherapy IVRT High-risk uterine cancer

a b s t r a c t Objective. To assess whether contemporary adjuvant management of early stage uterine carcinosarcoma (CS) produces equal outcomes as in uterine serous carcinoma (USC). Methods. We reviewed 172 women treated from 2000 to 2011 for stage I–II USC (n = 112, 65%) or CS (n = 60, 35%). Adjuvant therapy was initiated in 154 (90%) patients, with 111 patients receiving intravaginal radiotherapy (IVRT)/chemotherapy. Median follow up was 4.6 years for surviving patients. Results. Characteristics for USC vs. CS did not differ significantly by age ≥ 60, pelvic or para-aortic node sampling, stage, lymphovascular invasion, chemotherapy use, RT use or omission of adjuvant therapy. Outcomes were better for USC vs. CS in 5-year actuarial rates of recurrence [17% (C.I. 10–25%) vs. 45% (C.I. 31–59%), p b 0.001],disease-related mortality (DRM) [11% (5–17%) vs. 30% (16–44%), p = 0.016], and all-cause mortality [12% (C.I. 6–18%) vs. 34% (C.I. 20–48%), p = 0.007]. In multivariable analysis, CS histology remained a significant predictor of risk for recurrence [HR 3.1 (C.I. 1.7–5.7), p b 0.001], DRM [HR 2.4 (C.I. 1.1–5.1), p = 0.024], and all-cause mortality [HR 2.4 (C.I. 1.2–4.8), p = 0.012]. On sub-group analysis of 111 patients (77 USC, 34 CS) able to receive IVRT/chemotherapy, CS no longer was associated significantly with increased recurrence (29% vs. 15%, p = 0.18), DRM (22% vs. 10%, p = 0.39), or all-cause mortality (22% vs. 10%, p = 0.45). Conclusions. CS was associated with worse outcomes than USC. However, that difference was not maintained in patients able to receive IVRT and chemotherapy. While intriguing, this result may be due in part to selection against rapid early relapsing CS patients in this group. © 2014 Elsevier Inc. All rights reserved.

Introduction Uterine carcinosarcoma (CS) is an uncommon biphasic tumor believed to represent a de-differentiated state of endometrial carcinoma [1,2]. Studies have long suggested worse outcomes in CS, leading to its exclusion from investigations of high-risk early stage uterine cancer, such as the recently completed GOG 249 [3,4]. However, it is not clear ⁎ Corresponding author at: Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. Fax: +1 212 639 2417. E-mail address: [email protected] (K.M. Alektiar).

http://dx.doi.org/10.1016/j.ygyno.2014.07.097 0090-8258/© 2014 Elsevier Inc. All rights reserved.

whether its poor outcomes are related to intrinsic biology or to worse response to treatment. Comparison of CS to a suitable control with modern treatment is needed to answer these questions. Based on a high TP53 alteration rate, CS is most closely linked to uterine serous carcinoma (USC), which shares a high risk of locoregional and distant relapse [5]. Several investigations in early stage USC have reported excellent outcomes with adjuvant strategies combining chemotherapy (CT) and RT [6–15]. Recent trends favor intravaginal radiotherapy (IVRT) with carboplatin/taxane CT for its adaptation of treatment to recurrence pattern and good tolerability [14,15].

50

N.B. Desai et al. / Gynecologic Oncology 135 (2014) 49–53

In contrast, contemporary studies of early stage CS are few [16–18], and not all evaluate adjuvant IVRT + CT. At our institution, early stage CS has been treated under a similar paradigm to USC since 2000. This study assessed whether this similar contemporary management of CS has achieved equal success as in USC. Methods An institutional database was screened for patients with surgically treated FIGO 2009 stage I–II USC or CS between 01/2000 and 12/2011. We excluded those with no disease present at surgery (n = 8), positive pelvic washings (n = 19), prior abdominopelvic malignancy (n = 7), or adjuvant treatment outside our institution (n = 10). This avoided confounding by those not representing high-risk disease [USC stage IA without disease at surgery [8]] or those who may do worse than other early stage [positive washings [18]]. Specimens were reviewed by specialists in gynecologic pathology. The term mixed USC was used when USC represented at least 10% of the tumor. Standard clinical follow-up consisted of vaginal exam/Pap smear every 3 months and computed tomography of chest/abdomen/ pelvis every 6 months for the first 2 years, followed by graded decreases in visit and exam frequency. Comparisons of patient and treatment characteristics were performed with the chi-square statistic. Follow-up was measured from the date of surgery to the last known visit or death. The Kaplan– Meier method was used to estimate actuarial rates of relapse and death. Site-specific recurrences were calculated regardless of prior relapse(s) in other sites. The univariate 2-sided log-rank test and multivariable Cox-proportional hazards analysis were used for evaluation of prognostic factors. Results Patient and surgical characteristics The overall cohort of 172 women treated from 2000 to 2011 is described in Table 1. Median age was 67 (range 27–96). Diagnosis was

Table 1 Cohort characteristics by histology and as a whole.

Age b60 ≥60 Race Caucasian African-American Others LVI Yes No Stage⁎⁎ IA IB II PLN sampling Median nodes PALN sampling Median nodes Adjuvant therapy given Chemotherapy RT

USC (n = 112)⁎

CS (n = 60)

N

N

%

%

Total (n = 172) p

N

%

39 135

22% 78%

124 43 5

72% 25% 3%

57 117

33% 67%

126 26 20 155 15 121 6 154 127 142

73% 15% 12% 90% – 70% – 90% 74% 83%

0.77 24 88

21% 79%

14 46

23% 77%

78 32 2

70% 28% 2%

46 11 3

77% 18% 5%

33 79

30% 70%

24 36

40% 60%

82 15 15 100 15 76 6 100 84 96

73% 13% 13% 89% – 68% – 89% 75% 86%

44 11 5 55 16 45 6 54 43 46

74% 18% 8% 92% – 75% – 90% 72% 77%

0.19

0.16

0.47

0.62 0.33 0.85 0.64 0.14

USC — uterine serous carcinoma; CS — carcinosarcoma; LVI — lymphovascular invasion; PLN — pelvic lymph node; PALN — para-aortic lymph node; IVRT — intravaginal RT. ⁎ Mixed USC includes any other histology except CS. ⁎⁎ FIGO 2009; only patients with residual disease at surgery included.

made by endometrial biopsy (n = 110, 64%), curettage (n = 57, 33%), or at surgery 5 (3%). Median body mass index was 30 (18–60) kg/m2. The hysterectomy/salpingo-oophorectomy type was open in 99 (58%) patients and minimally invasive in 73 patients (42%). Peritoneal cytology was obtained in 168 (98%) patients and omentum sampling in 135 (78%) patients. Pelvic node sampling was done in 155 (90%) of patients and para-aortic node sampling in 121 (70%) of patients. Obesity was more prevalent in the 30% of patients not undergoing para-aortic sampling vs. those who did (65% vs. 44%, p = 0.004). Stage was IA in 126 (73%), IB in 26 (15%), and II in 20 (12%) of patients. Lymphovascular invasion (LVI) was present in 57 (33%) patients. Histology was USC (n = 112, 65%) or CS (n = 60, 25%). Cohort characteristics, summarized in Table 1, were not significantly different between individual histologies, with regard to age, race, LVI, stage, nodal sampling, and receipt of adjuvant therapy. Further, similar proportions of women for USC vs. CS reported prior oral contraceptive use (30% vs. 24%, p = 0.47), anti-estrogen treatment (9% vs. 12%, p = 0.54), diabetes (19% vs. 20%, p = 0.84), hypertension (54% vs. 40%, p = 0.09), tobacco use (38% vs. 30%, p = 0.23), and nonabdominopelvic malignancy history (22% s. 13%, p = 0.18). While disease confined to the endometrium was more common in USC vs. CS (46% vs 20%, p b 0.001), overall FIGO 2009 stage distribution was not significantly different (p = 0.48). Specifically, there was no difference for USC vs. CS in the proportion of ≥ 50% myometrial involvement (16% vs. 25%, p = 0.16), cervical stromal involvement (13% vs. 8%, p = 0.32), or any cervical involvement (17% vs. 16%, p = 0.53). All patients had disease present at time of surgery. Adjuvant therapy characteristics Adjuvant therapy was initiated in 154 (90%) of women as follows: IVRT with CT (n = 111, 64%), IVRT alone (n = 21, 12%), CT alone (n = 12, 7%), pelvic RT alone (n = 6, 4%), or pelvic RT with CT (n = 4, 2%). The time to treatment initiation was 1.4 months (range 0.4–3.8 months). Of the 4 (3%) patients starting N 3 months after surgery, delay was related to post-operative wound complication [2], medical complications [1], and patient preference [1]. The median time to completion of adjuvant therapy was longest for CT containing regimens (5 months, range 1.6–9.2 months). Of the 18 (10%) patients not receiving any adjuvant therapy, 12 were USC, 6 were CS, and stage was as follows: IA (n = 16), IB (n = 2), II (n = 1). The reasons for not receiving adjuvant therapy were patient refusal (n = 10), MD preference (n = 7), or rapid disease relapse before the start of adjuvant (n = 1). The reasons by histology for omitting adjuvant for USC vs. CS were patient refusal (n = 8 vs. 2), physician preference (n = 4 vs. 3), and disease progression before adjuvant (n = 0 vs. 1). When given, IVRT was prescribed to a median of 21 Gy in 3 fractions at a depth of 0.5 cm to the proximal two-thirds of the vagina and delivered by high-dose rate Ir-192 remote afterloader with pretreatment planning and dosimetry, as previously described [19,20]. Of the 132 (77%) patients receiving IVRT, two did not complete treatment due to early relapse or cuff dehiscence, while the remainder received 18–21 Gy. Median time to IVRT was 7 weeks. When given with CT, treatments were concurrent but interspaced such that CT and IVRT were given on separate weeks. Pelvic RT was given to 10 (6%) patients, either alone (n = 8) or with CT (n = 2), to a median dose of 50.4 Gy, as previously described [21]. Of the 30 patients not receiving any RT, 12 received CT alone. In these 12 patients, the reasons for omission of RT were physician or patient preference (n = 7) or co-morbidities, such as post-operative complication or inflammatory bowel disease (n = 5). CT was given to 128 (74%) patients. CT consisted of carboplatin/ taxane (paclitaxel or docetaxel; n = 120, 94%), carboplatin alone (n = 3, 2%), or ifosfamide doublets (n = 5, 4%) with carboplatin, cisplatin, or paclitaxel. Starting doses for CT were as follows: carboplatin AUC 5–6,

N.B. Desai et al. / Gynecologic Oncology 135 (2014) 49–53

paclitaxel 175 mg/m2, ifosfamide 1200–2000 mg/m2, or cisplatin 20 mg/m2. For those specifically not receiving CT, the reasons were polyp-confined disease (n = 4) and physician or patient preference (n = 29). The pattern of RT and/or CT use did not differ significantly for USC vs. CS: adjuvant therapy initiated (89% vs. 90%, p = 0.88), CT received (75% vs. 72%, p = 0.64), or RT received (86% vs. 77%, p = 0.14). There was no significant difference in distribution in the number of cycles of CT received for USC vs. CS (p = 0.23): 6 cycles (80% vs. 74%), 5 cycles (4% vs. 5%), 4 cycles (7% vs. 15%), 3 cycles (10% vs. 5%), 1 cycle (0% vs. 6%). There were no significant differences in the proportion of USC vs. CS patients requiring CT dose reduction (30% vs. 23%, p = 0.44) or regimen change (20% vs. 26%, p = 0.49). However, while CT for USC patients was exclusively carboplatin/taxane (paclitaxel 98%, docetaxel 2%), CS patients received carboplatin/paclitaxel (81%), ifosfamide doublets (12%), or carboplatin alone (7%).

Outcomes The median follow-up time was 48 months (range 3–139) for all patients and 55 months (range 3–139) for surviving patients. The overall 5-year actuarial rates of relapse, DRM, and all-cause mortality were 27% (C.I. 20–34%), 17% (C.I. 11–23%), and 19% (C.I. 13–26%), respectively. By histology, the surviving patient median follow-up for USC vs. CS was 58 vs. 48 months. On log-rank univariate comparison, there was a significantly lower rate of 5-year actuarial relapse in USC vs. CS [17% (C.I. 10–25%) vs. 45% (C.I. 31–59%), p b 0.001; Fig. 1A]. There was also a lower 5-year actuarial risk for USC vs. CS of DRM [11% (5–17%) vs. 30% (16–44%), p = 0.016; Fig. 1B] and all-cause mortality [12% (C.I. 6–18%) vs. 34% (C.I. 20–48%), p = 0.007]. In Cox proportional hazards multivariable analysis (Table 2), CS histology remained significantly associated with poor outcome: recurrence hazard ratio (HR) 3.1 (C.I. 1.7–5.7, p b 0.001); DRM HR 2.4 (C.I. 1.1–5.1, p = 0.02); all-cause mortality HR 2.4 (1.2–4.8, p = 0.01). This was also the case for the association of stage IB–II with poor outcome: recurrence HR 2.9 (C.I. 1.6–5.3, p = 0.001); DRM HR 2.6 (C.I. 1.2–5.7), p = 0.02; all-cause mortality HR 2.6 (1.3–5.3, p = 0.01). While LVI was associated with recurrence risk (HR 1.9, p = 0.04), there was no significant difference in receipt of CT (p = 0.41) or RT (p = 0.63) based on LVI, arguing against influence on treatment selection. Use of minimally invasive vs. open surgery did not influence DRM (p = 0.81). There was no influence of CS heterologous vs. homologous differentiation on outcome (recurrence p = 0.3, DRM p = 0.8, all-cause mortality p = 0.9).

51

Patterns of relapse The 5-year actuarial site-specific relapse rates for all patients were: vaginal 6% (C.I. 2–10%), pelvic 11% (C.I. 6–16%), para-aortic LN 9% (C.I. 4–13%), peritoneal 12% (C.I. 7–17%), distant non-peritoneal 19% (C.I. 13–25%), and any distant 22% (C.I. 15–29%). Table 3 compares the sitespecific relapse rates by histology. On univariate comparison, there was a significant increase in relapse for CS in the para-aortic nodes (p = 0.037) and at distant non-peritoneal sites (p = 0.001). There was no significant difference for vaginal, pelvic, or peritoneal relapse. Of the 12/172 patients who did not receive adjuvant therapy, 3/6 in the CS group relapsed vs. only 1/12 in the USC group. Of 9 vaginal relapses, 7 had received adjuvant IVRT. There were 5 isolated relapses (1 USC, 4 CS). Of these 5, 1 patient with CS did not receive adjuvant RT, and 1 other patient with CS received only one fraction before detection of vaginal relapse. Of the 17 pelvic relapses, 14 patients had undergone pelvic node sampling (median # nodes 12), 2 received pelvic RT, and all received adjuvant CT. Pelvic relapse was isolated in 4 patients (3 USC, 1 CS), all of whom had pelvic node sampling and none of whom had received adjuvant pelvic RT. The rate of isolated pelvic relapse was 2.5% (4/162) in those not receiving pelvic RT vs. 0% (0/ 10) in those receiving pelvic RT. There were 14 para-aortic LN relapses (para-aortic node sampling 12/14, median # nodes 6), and adjuvant CT was received by 10/14 of these patients. Three patients had isolated para-aortic LN recurrence, all of whom had para-aortic node sampling (median # nodes 8). Distant relapse occurred in 36 patients involving the following sites: peritoneum (n = 19), lung (n = 21), liver (n = 10), distant lymph nodes (n = 10), bone (n = 7), brain (n = 1), and others (n = 3). In these patients, pelvic nodes were sampled in 32 cases (89%, median # nodes 15) and para-aortic nodes in 28 patients (78%, median # nodes 6). Adjuvant CT was received by 23 (64%) such patients. The temporal pattern of relapse was skewed towards early recurrence (Fig. 2), with 31 (70%) recurrences occurring within two years. This was particularly magnified for CS. Of the 5 relapses within 5 months of surgery (the median time to completion of CT containing regimens), all were in CS patients. Furthermore, 2 out of 5 of these recurrences occurred within 6 weeks of surgery (the median time to initiation of adjuvant therapy).

Sub-group analysis of patients receiving combined modality approach A main study goal was to examine the effect of contemporary multimodality management on outcomes for CS and UPSC. As only 4 (2%)

Fig. 1. USC vs. CS Kaplan–Meier functions for A) relapse-free survival and B) disease-related mortality.

52

N.B. Desai et al. / Gynecologic Oncology 135 (2014) 49–53

Table 2 Analysis of prognostic factors for actuarial risk of relapse and disease-related mortality (DRM) by univariate log-rank (left) and Cox-proportional hazards multivariable analysis (right). Univariate log-rank

Stage IB-II IA Histology CS USC LVI Yes No Age ≧60 b60

Multivariable Cox

5-Year relapse (44 events)

5-Year DRM (27 events)

Relapse risk (44 events)

Risk of DRM (27 events)

N

Rate (95% CI)

p

Rate (95% CI)

p

HR (95% CI)

p

HR (95% CI)

p

46 126

48% (32–63%) 19% (12–26%)

b0.001 –

26% (13–40%) 14% (7–21%)

0.005 –

2.9 (1.6–5.3) –

0.001 –

2.6 (1.2–5.7) –

0.02 –

60 110

45% (31–59%) 17% (10–25%)

b0.001 –

30% (16–44%) 11% (5–17%)

0.016 –

3.1 (1.7–5.7) –

b0.001 –

2.4 (1.1–5.1) –

0.02 –

57 115

41% (36–46%) 20% (14–25%)

0.003 –

21% (10–33%) 17% (11–22%)

0.15 –

1.9 (1.0–3.6) –

0.04 –

1.3 (0.6–2.9) –

0.53 –

134 38

27% (19–35%) 26% (11–42%)

0.76 –

17% (0–24%) 16% (3–30%)

0.72 –

1.2 (0.6–2.5) –

0.66 –

1.2 (0.4–3.3) –

0.71 –

LVI — lymphovascular invasion, HR — hazard ratio, CI — confidence interval. Italics - significant p-value.

patients received pelvic RT with CT, we focused on the 111 patients (64% of total) receiving IVRT + CT [77 USC (69% of USC) vs. 34 CS (57% of CS)]. Median follow-up was 57 months for surviving patients. There were 22 recurrences (5-year actuarial rate 17%, C.I. 11–23%) and 13 disease-related deaths (5-year actuarial rate 13%, C.I. 6–20). Sub-group characteristics again were similar for the USC vs. CS patients: age ≥ 60 (77% vs. 79%, p = 0.75), stage distribution (p = 0.64), LVI (27% vs. 38%, p = 0.25), pelvic node sampling (91% vs. 97%, p = 0.25; median # nodes 16 vs 16), and para-aortic node sampling (71% vs. 79%, p = 0.38; median # nodes 6 vs. 6). In contrast to the whole cohort comparisons, this sub-group analysis was unable to detect a statistically significant difference for USC vs. CS in 5-year actuarial rates of recurrence (15% vs. 29%, p = 0.18), DRM (10% vs. 22%, p = 0.39), or all-cause mortality (10% vs. 22%, p = 0.45), respectively. These changes appeared driven by improved performance of CS patients receiving IVRT + CT vs. the overall CS cohort, in terms of 5-year relapse (29% vs. 45%), DRM (22% vs. 30%), and all-cause mortality (22% vs. 34%). Discussion

changes in treatment strategy. Thus, even when accounting for stage, LVI and age in multivariable analysis, CS remained associated with higher risk of relapse (HR 3.1, p b 0.001), DRM (HR 2.4, p = 0.02), and all-cause mortality (HR 2.4, p = 0.01). The fact that CS patients do worse is not a new finding. Notably, in Amant et al.'s retrospective comparison of stage I–II CS (n = 25) and USC/clear cell carcinoma (n = 24) patients, only 11/25 CS patients survived, compared to 18/24 USC/clear cell patients (p = 0.008) [3]. Instead, the current study's main question was whether the difference in outcomes between CS and USC was due to intrinsic biology or worse response to treatment. A few observations can be made regarding the worse performance of CS in our study. First, of the 5 patients who recurred within 5 months of surgery (the median to completion of CT-containing regimens), all were CS patients. Indeed, relapse was so rapid in 2/5 patients that it occurred within 6 weeks of surgery, which is the median time to initiation of adjuvant therapy. Second, in those selected patients who did not receive any adjuvant therapy (n = 18), such omission seems to have had less of an impact on UCS than CS (1/12 vs. 3/6 relapsed, respectively). Third, the pattern of relapse also differed between the two histologies. Specifically, CS had a higher 5-year actuarial risk of relapse at non-

In the current study, outcomes for 60 stage I–II CS patients were significantly worse than for 112 stage I–II USC patients, in terms of relapse (45% vs. 17%, p b 0.001), DRM (30% vs. 11%, p = 0.016), and all-cause mortality (34% vs. 12%, p = 0.007). The USC and CS cohorts were well balanced, without significant differences in demographic (age ≥ 60), co-morbidity, surgery (pelvic or para-aortic node sampling and yield), FIGO 2009 stage, LVI, or adjuvant therapy characteristics. Further, the lack of prognostic influence of surgery type minimized bias from

Table 3 5-Year site-specific actuarial relapse rates. Relapse site

Vaginal Isolated Pelvic Isolated Para-aortic Isolated Peritoneal Distant (non-peritoneal) Any distant

USC

CS

% (95% C.I.)

% (95% C.I.)

4% (0–8%) 1 10% (4–16%) 3 5% (1–9%) 0 9% (3–15%) 11% (5–16%) 14% (7–20%)

9% (1–17%) 4 13% (3–23%) 1 16% (6–27%) 3 18% (7–30%) 36% (22–49%) 38% (24–51%)

p⁎

0.147 0.740 0.037 0.148 0.001 0.001

USC — uterine papillary serous carcinoma, CS — carcinosarcoma. Italics - significant p-value. ⁎ 2-Sided log-rank.

Fig. 2. Histogram of time to recurrence (log base 2.5 scale) by histology. Early events are compared to median time to adjuvant therapy initiation (1.5 months, dashed line) and completion (5 months, dotted line).

N.B. Desai et al. / Gynecologic Oncology 135 (2014) 49–53

peritoneal distant sites (38% CS vs. 14% USC, p = 0.001) and the paraaortic nodes (16% CS vs. 5% USC, p = 0.037). In contrast, vaginal–pelvic and peritoneal relapse risks were similar, with rare isolated vaginal (3%) or pelvic (2%) relapses. To assess whether the poorer results in CS patients were due to a difference in response to treatment, outcomes were evaluated in the 111 patients (77 USC, 34 CS) receiving IVRT and CT. These patients performed better with 5-year actuarial relapse of 17% and DRM of 13%, as compared to 27% relapse and 17% DRM in the parent cohort, likely in part due selection. Nonetheless, there were still 22 recurrences and 13 deaths from endometrial cancer, providing reasonable event numbers for comparison by histology. Patient and treatment characteristics remained well balanced, without significant differences in age ≥ 60, pelvic or para-aortic node sampling and yield, and stage distribution. However, the differences in 5-year actuarial relapse (15 vs. 29%, p = 0.18), DRM (10 vs. 22%, p = 0.39), and all-cause mortality (10% vs. 22%, p = 0.45) were no longer significant, due to improved outcomes in the CS sub-group. Whereas USC patients receiving IVRT + CT had similar relapse rate vs. all USC patients (15% vs. 17%), the CS patients receiving IVRT + CT had a lower rate of relapse vs. all CS patients (29% vs. 45%). The first limitation of this sub-group analysis is the possibility of selection for better performing CS patients. Indeed, of the noted 5 rapid CS relapses (≤ 5 months), only 1 received IVRT + CT. However, similar proportions of USC (69%) and CS (57%) patients received IVRT + CT, and the reasons for omission of adjuvant RT or CT were also similar. Thus, systematic differences in treatment selection between CS and USC were not obvious. It also is possible that the smaller size of the sub-group obscured the statistical significance of the differences. The sub-group analysis was intentionally limited to patients receiving both IVRT and CT because they represented both the largest subset in the whole cohort (111/172, 65%) and our current treatment philosophy. For example, only 2% of patients received pelvic RT and CT. We also excluded from sub-group analysis those who received CT alone, because the decision on omitting RT was based on confounding issues. Of the 8 patients with CS treated with CT alone, 3 had significant comorbidity that prevented them from having any form of RT. Including these patients would have risked worsening selection bias. Nonetheless, our observation that combined IVRT and CT provides reasonably good results in patients with stage I–II CS is somewhat unique. In a multi-institutional report of 111 patients with stage I–II CS, only 15 patients were treated with RT and CT. Perhaps, it could be argued that it is the use of CT rather than combined modality approach that led to improved outcome in CS patients. However, in a study by Gonzalez Bosquet et al. [18], vaginal recurrence was observed in 4/16 patients with stage I–II CS, compared to none in 9 patients treated with RT [17]. Thus, we would more conservatively conclude that patients with early stage CS on the whole do worse than those with USC, perhaps due to early recurrences. However, patients who are able to complete IVRT + CT perform somewhat equally in CS and USC. This improved understanding of early stage CS behavior in the contemporary treatment era may help inform prospective study of the disease. Insights from genomic characterization of uterine cancer, which have begun to bridge the gap in understanding between type I and type II uterine carcinoma histology [22], may be also helpful in clarifying the relationship of CS to other uterine carcinomas. Conclusion In this single institution series, early stage CS confers worse outcomes overall than the most common type II high-risk uterine carcinoma, USC. In contrast, those able to complete a regimen of IVRT + CT appear to fare similarly. While this may reflect the efficacy of the

53

treatment, it also appears that early relapse of CS even before or during adjuvant therapy initiation may be an associated problem.

Conflicts of interest statement The authors declare that there are no conflicts of interest.

References [1] Arend R, Doneza JA, Wright JD. Uterine carcinosarcoma. Curr Opin Oncol 2011; 23(5):531–6. [2] Kempson RL, Hendrickson MR. Smooth muscle, endometrial stromal, and mixed Mullerian tumors of the uterus. Mod Pathol 2000;13(3):328–42. [3] Amant F, Cadron I, Fuso L, Berteloot P, de Jonge E, Jacomen G, et al. Endometrial carcinosarcomas have a different prognosis and pattern of spread compared to high-risk epithelial endometrial cancer. Gynecol Oncol 2005;98(2):274–80. [4] Vaidya AP, Horowitz NS, Oliva E, Halpern EF, Duska LR. Uterine malignant mixed Mullerian tumors should not be included in studies of endometrial carcinoma. Gynecol Oncol 2006;103(2):684–7. [5] Boruta 2nd DM, Gehrig PA, Fader AN, Olawaiye AB. Management of women with uterine papillary serous cancer: a Society of Gynecologic Oncology (SGO) review. Gynecol Oncol 2009;115(1):142–53. [6] Huh WK, Powell M, Leath 3rd CA, Straughn Jr JM, Cohn DE, Gold MA, et al. Uterine papillary serous carcinoma: comparisons of outcomes in surgical Stage I patients with and without adjuvant therapy. Gynecol Oncol 2003;91(3):470–5. [7] Dietrich 3rd CS, Modesitt SC, DePriest PD, Ueland FR, Wilder J, Reedy MB, et al. The efficacy of adjuvant platinum-based chemotherapy in Stage I uterine papillary serous carcinoma (UPSC). Gynecol Oncol 2005;99(3):557–63. [8] Kelly MG, O'Malley DM, Hui P, McAlpine J, Yu H, Rutherford TJ, et al. Improved survival in surgical stage I patients with uterine papillary serous carcinoma (UPSC) treated with adjuvant platinum-based chemotherapy. Gynecol Oncol 2005; 98(3):353–9. [9] Fader AN, Nagel C, Axtell AE, Zanotti KM, Kelley JL, Moore KN, et al. Stage II uterine papillary serous carcinoma: Carboplatin/paclitaxel chemotherapy improves recurrence and survival outcomes. Gynecol Oncol 2009;112(3):558–62. [10] Goldberg H, Miller RC, Abdah-Bortnyak R, Steiner M, Yildiz F, Meirovitz A, et al. Outcome after combined modality treatment for uterine papillary serous carcinoma: a study by the Rare Cancer Network (RCN). Gynecol Oncol 2008;108(2):298–305. [11] Viswanathan AN, Macklin EA, Berkowitz R, Matulonis U. The importance of chemotherapy and radiation in uterine papillary serous carcinoma. Gynecol Oncol 2011; 123(3):542–7. [12] Fader AN, Drake RD, O'Malley DM, Gibbons HE, Huh WK, Havrilesky LJ, et al. Platinum/taxane-based chemotherapy with or without radiation therapy favorably impacts survival outcomes in stage I uterine papillary serous carcinoma. Cancer 2009;115(10):2119–27. [13] Jhingran A, Ramondetta LM, Bodurka DC, Slomovitz BM, Brown J, Levy LB, et al. A prospective phase II study of chemoradiation followed by adjuvant chemotherapy for FIGO stage I–IIIA (1988) uterine papillary serous carcinoma of the endometrium. Gynecol Oncol 2013;129(2):304–9. [14] Kiess AP, Damast S, Makker V, Kollmeier MA, Gardner GJ, Aghajanian C, et al. Fiveyear outcomes of adjuvant carboplatin/paclitaxel chemotherapy and intravaginal radiation for stage I–II papillary serous endometrial cancer. Gynecol Oncol 2012; 127(2):321–5. [15] Desai NB, Kiess AP, Kollmeier MA, Abu-Rustum NR, Makker V, Barakat RR, et al. Patterns of relapse in stage I–II uterine papillary serous carcinoma treated with adjuvant intravaginal radiation (IVRT) with or without chemotherapy. Gynecol Oncol 2013;131(3):604–8. [16] Makker V, Abu-Rustum NR, Alektiar KM, Aghajanian CA, Zhou Q, Iasonos A, et al. A retrospective assessment of outcomes of chemotherapy-based versus radiationonly adjuvant treatment for completely resected stage I–IV uterine carcinosarcoma. Gynecol Oncol 2008;111(2):249–54. [17] Cantrell LA, Havrilesky L, Moore DT, O'Malley D, Liotta M, Secord AA, et al. A multiinstitutional cohort study of adjuvant therapy in stage I–II uterine carcinosarcoma. Gynecol Oncol 2012;127(1):22–6. [18] Gonzalez Bosquet J, Terstriep SA, Cliby WA, Brown-Jones M, Kaur JS, Podratz KC, et al. The impact of multi-modal therapy on survival for uterine carcinosarcomas. Gynecol Oncol 2010;116(3):419–23. [19] Turner BC, Knisely JP, Kacinski BM, Haffty BG, Gumbs AA, Roberts KB, et al. Effective treatment of stage I uterine papillary serous carcinoma with high dose-rate vaginal apex radiation (192Ir) and chemotherapy. Int J Radiat Oncol Biol Phys 1998; 40(1):77–84. [20] Sorbe BG, Smeds AC. Postoperative vaginal irradiation with high dose rate afterloading technique in endometrial carcinoma stage I. Int J Radiat Oncol Biol Phys 1990;18(2):305–14. [21] Shih KK, Milgrom SA, Abu-Rustum NR, Kollmeier MA, Gardner GJ, Tew WP, et al. Postoperative pelvic intensity-modulated radiotherapy in high risk endometrial cancer. Gynecol Oncol 2013;128(3):535–9. [22] Cancer Genome Atlas Research NKandoth C, Schultz N, Cherniack AD, Akbani R, Liu Y, et al. Integrated genomic characterization of endometrial carcinoma. Nature 2013; 497(7447):67–73.