new england

The
journal of medicine
established in 1812 December 2, 2021 vol. 385  no. 23

Randomized Trial of Cytoreductive Surgery for Relapsed

Ovarian Cancer
P. Harter, J. Sehouli, I. Vergote, G. Ferron, A. Reuss, W. Meier, S. Greggi, B.J. Mosgaard, F. Selle, F. Guyon,
C. Pomel, F. Lécuru, R. Zang, E. Avall‑Lundqvist, J.-W. Kim, J. Ponce, F. Raspagliesi, G. Kristensen, J.-M. Classe,
P. Hillemanns, P. Jensen, A. Hasenburg, S. Ghaem‑Maghami, M.R. Mirza, B. Lund, A. Reinthaller, A. Santaballa,
A. Olaitan, F. Hilpert, and A. du Bois, for the DESKTOP III Investigators*​​

a bs t r ac t

BACKGROUND
Treatment for patients with recurrent ovarian cancer has been mainly based on The authors’ full names, academic de-
systemic therapy. The role of secondary cytoreductive surgery is unclear. grees, and affiliations are listed in the
Appendix. Dr. Harter can be contacted at
METHODS ­p​.­harter@​­kem-med​.­com or at the Depart-
ment of Gynecology and Gynecologic
We randomly assigned patients with recurrent ovarian cancer who had a first re- Oncology, Kliniken Essen-Mitte, Henricistr.
lapse after a platinum-free interval (an interval during which no platinum-based 92, 45136 Essen, Germany.
chemotherapy was used) of 6 months or more to undergo secondary cytoreductive * A list of the investigators in the DESKTOP
surgery and then receive platinum-based chemotherapy or to receive platinum- III trial is provided in the Supplemen-
based chemotherapy alone. Patients were eligible if they presented with a positive tary Appendix, available at NEJM.org.
Arbeitsgemeinschaft Gynäkologische Onkologie (AGO) score, defined as an East- N Engl J Med 2021;385:2123-31.
ern Cooperative Oncology Group performance-status score of 0 (on a 5-point scale, DOI: 10.1056/NEJMoa2103294
Copyright © 2021 Massachusetts Medical Society.
with higher scores indicating greater disability), ascites of less than 500 ml, and
complete resection at initial surgery. A positive AGO score is used to identify pa- CME
tients in whom a complete resection might be achieved. The primary end point at NEJM.org
was overall survival. We also assessed quality of life and prognostic factors for
survival.
RESULTS
A total of 407 patients underwent randomization: 206 were assigned to cytoreduc-
tive surgery and chemotherapy, and 201 to chemotherapy alone. A complete resec-
tion was achieved in 75.5% of the patients in the surgery group who underwent
the procedure. The median overall survival was 53.7 months in the surgery group
and 46.0 months in the no-surgery group (hazard ratio for death, 0.75; 95% con-
fidence interval, 0.59 to 0.96; P = 0.02). Patients with a complete resection had the
most favorable outcome, with a median overall survival of 61.9 months. A benefit
from surgery was seen in all analyses in subgroups according to prognostic fac-
tors. Quality-of-life measures through 1 year of follow-up did not differ between
the two groups, and we observed no perioperative mortality within 30 days after
surgery.
CONCLUSIONS
In women with recurrent ovarian cancer, cytoreductive surgery followed by chemo-
therapy resulted in longer overall survival than chemotherapy alone. (Funded by
the AGO Study Group and others; DESKTOP III ClinicalTrials.gov number,
NCT01166737.)

n engl j med 385;23  nejm.org  December 2, 2021 2123

The New England Journal of Medicine
Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

T
he mainstay of treatment for wom- analysis for complete resection in the DESKTOP I
en with advanced ovarian cancer has been trial were complete resection at primary surgery,
primary surgery with the goal of complete an Eastern Cooperative Oncology Group perfor-
macroscopic resection of all tumor, followed by mance-status score of 0 (on a 5-point scale, with
A Quick Take
carboplatin and paclitaxel combination chemo- higher scores indicating greater disability), and
is available at therapy.1 More recently, additional systemic ther- ascites of 500 ml or less. The AGO score was de-
NEJM.org apy with bevacizumab or a poly(adenosine diphos- fined as positive if all three factors were present.
phate–ribose) polymerase (PARP) inhibitor has In the subsequent multicenter, prospective
been associated with superior progression-free DESKTOP II trial, which consecutively enrolled
survival.2-7 The standard of care in relapsed ovar- 516 patients who had had a platinum-sensitive
ian cancer has mainly been systemic treatment. relapse, complete resection was achieved in 76%
So far, only a few trials have shown level 1 evi- of 129 patients who had a positive AGO score
dence of a significant overall survival benefit and underwent surgery for a first relapse. This
with the use of systemic therapy in relapsed finding confirmed the value of the AGO score in
ovarian cancer.8,9 Two other trials have shown a predicting complete resectability of a tumor.15
clinically relevant survival benefit with systemic Thereafter, we conducted the prospectively ran-
therapy, but statistical significance was shown domized DESKTOP III trial, the results of which
only after adjustment or in subgroup analyses.10,11 are reported here.
The role of surgery in relapsed ovarian cancer
has not been well defined. Me thods
We initiated the Descriptive Evaluation of Pre-
operative Selection Criteria for Operability in Re- Trial Design
current Ovarian Cancer (DESKTOP) series when Patients who met the eligibility criteria were
the evidence consisted only of retrospective trials randomly assigned in a 1:1 ratio to undergo cyto-
in heterogeneous populations that suggested a reductive surgery and then receive the physi-
benefit of surgery in patients with platinum- cian’s choice of platinum-based chemotherapy or
sensitive relapsed disease (i.e., patients with a to receive the physician’s choice of platinum-
durable response to platinum therapy, defined as based chemotherapy alone. The protocol strongly
≥6 months without disease progression after the recommended combination therapy, but a single
end of platinum therapy).12 First, we defined the agent was allowed. Randomization was stratified
surgical aim associated with a potential survival according to center and platinum-free interval
benefit and determined a score that would indi- (the interval during which no platinum-based
cate predictive factors for complete resection.13 chemotherapy was used: no previous chemo-
The DESKTOP I trial confirmed the beneficial therapy, an interval of 6 to 12 months, or an
role of complete resection, which surpassed the interval of >12 months). We applied a covariate-
effect of cytoreduction in upfront surgery.14 Only adaptive randomization procedure according to
complete resection was associated with any long- Rosenberger and Lachin, which combines ele-
term benefit in recurrent ovarian cancer. There- ments of the minimization approach with a
fore, a predictive score (the Arbeitsgemeinschaft biased coin technique.16
Gynäkologische Onkologie [AGO] score) that The trial was performed according to the
would identify patients in whom a complete re- European Network for Gynaecological Oncologi-
section might be achieved was deemed necessary cal Trial Groups model A.17 Trial centers were
to select patients for a prospective trial of cyto- selected on the basis of experience in ovarian
reductive surgery. These selection criteria should cancer studies and participation in previous sur-
fulfill two goals: the proportion of patients who gical trials in this field. Ethical approval was
are exposed to a potentially harmful interven- obtained at each participating center, and all
tion but do not gain any benefit must be mini- patients provided written informed consent. The
mized, and the trial evaluating surgery as the trial protocol and statistical analysis plan are
method of treatment should not be diluted by a available with the full text of this article at
high number of patients in whom complete NEJM.org. The authors vouch for the accuracy
macroscopic tumor clearance is not successful. and completeness of the data and for the fidelity
Independent predictive factors in multivariate of the trial to the protocol. GlaxoSmithKline and

2124 n engl j med 385;23  nejm.org  December 2, 2021

The New England Journal of Medicine

Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Cytoreductive Surgery for Relapsed Ovarian Cancer

Medac (two of the sponsors of the trial) had no in the no-surgery group, and we planned the
role in the conduct of the trial, the handling of trial to show a hazard ratio for death of 0.70. A
data, or the preparation of the manuscript. two-stage group sequential design according to
the O’Brien–Fleming procedure19 was applied,
Eligibility which would have allowed the trial to finish
Patients were eligible if they had relapsed histo- earlier with 80% power if the hazard ratio was
logically diagnosed epithelial ovarian cancer 0.50. The interim analysis was scheduled after
(clinically defined as a lesion that is palpable or 122 deaths had been reported, testing against a
visible or that is visible on ultrasonographic im- one-sided significance level of 0.003, and the
aging) or relapsed disease radiologically diagnosed final analysis was scheduled to be performed
at least 6 months after the last course of initial after 244 deaths had been reported, testing
platinum-based chemotherapy (i.e., platinum- against a one-sided significance level of 0.024.
sensitive disease) and had a positive AGO score. With a planned accrual period of 3 years and
An elevated cancer antigen 125 level alone was a 3-year follow-up period, and accounting for a
not deemed to be an acceptable entry criterion. potential dropout rate of 10%, we determined
that 408 patients would be needed to observe
End Points and Assessments 244 deaths, which would give the trial a power
The primary end point of overall survival was of 80%. At the interim analysis, the between-
defined as the time from randomization to death. group difference in overall survival was not sig-
We also report the following planned secondary nificant at the local 0.003 significance level, so
end points: quality of life at baseline, 6 months, follow-up continued. The final primary efficacy
and 12 months after randomization, as assessed analysis was performed on an intention-to-treat
with the use of the European Organization for basis after 254 deaths had occurred. Analysis
Research and Treatment of Cancer Quality-of- of surgical characteristics was performed in pa-
Life Questionnaire (QLQ-C30) and the National tients assigned to the surgery group who under-
Comprehensive Cancer Network Functional As- went the procedure. Quality-of-life questionnaires
sessment of Cancer Therapy (FACT)–General were scored according to relevant manuals and
and FACT–Ovarian and its corresponding FACT– analyzed with the use of a model-based ap-
Ovarian Symptom Index; progression-free sur- proach, assuming missingness at random. De-
vival (defined as the time from randomization tailed methods and results are provided in Fig.
to investigator-assessed disease progression or S1 in the Supplementary Appendix, available at
death, whichever came first); complete resection NEJM.org. Confidence intervals were not adjusted
as a prognostic factor; complications associated for multiplicity, so inferences drawn may not be
with surgery up to 60 days after surgery; and an reproducible.
exploratory analysis of surgical characteristics
and applied chemotherapy.
R e sult s
Statistical Analysis Patients and Treatment
For the primary efficacy comparison, we used A total of 407 patients (206 in the surgery group
a two-sided Wald test from a Cox regression and 201 in the no-surgery group) were enrolled
model of overall survival stratified according to from September 2010 through March 2015 and
platinum-free interval (no previous platinum che- were included in the analyses (Fig. 1). The base-
motherapy, an interval of 6 to 12 months, or an line characteristics were well balanced between
interval of >12 months). For all between-group the two groups. Nearly all the patients had re-
comparisons of overall survival and progression- ceived previous platinum-based chemotherapy at
free survival, we checked the proportional-hazards first diagnosis, and 75% of the patients in each
assumptions by performing Grambsch–Therneau group had a platinum-free interval of more than
tests, and we omitted hazard ratios if the tests 12 months (Table 1). The median time from
showed evidence of nonproportionality.18 On the randomization to the start of chemotherapy in
basis of data from the DESKTOP I trial, we as- the no-surgery group was 15 days (interquartile
sumed that 2-year overall survival would be ap- range, 8 to 22). The median time from random-
proximately 66% in the surgery group and 55% ization to surgery in the surgery group was 16

n engl j med 385;23  nejm.org  December 2, 2021 2125

The New England Journal of Medicine
Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Overall Survival and Progression-free

407 Patients were enrolled in the trial
Survival
and underwent randomization The median follow-up was 69.8 months (inter-
quartile range, 59.8 to 80.4). The median overall
survival was 53.7 months (95% confidence inter-
206 Were assigned to cytoreductive 201 Were assigned to platinum-
val [CI], 46.8 to 61.6) in the surgery group and
surgery followed by platinum-based based chemotherapy alone 46.0 months (95% CI, 39.5 to 52.6) in the no-
chemotherapy and were included in and were included in the
the intention-to-treat analysis intention-to-treat analysis
surgery group (hazard ratio for death, 0.75; 95%
CI, 0.59 to 0.96; P = 0.02). The median progres-
sion-free survival was 18.4 months (95% CI, 15.7
192 Underwent cytoreductive surgery 201 Received platinum-based chemo- to 20.8) in the surgery group and 14.0 months
145 Had complete resection therapy (95% CI, 12.7 to 15.4) in the no-surgery group
47 Had macroscopic residual disease 187 Completed the trial per protocol
14 Did not undergo surgery 8 Underwent cytoreductive surgery (hazard ratio for progression or death, 0.66;
8 Withdrew or had intercurrent 6 Withdrew consent or declined 95% CI, 0.54 to 0.82) (Fig. 2). Among the pa-
illness diagnosed after further treatment
randomization tients who were assigned to and underwent
4 Had findings after randomization surgery, the median progression-free survival
that suggested unresectability
1 Had erroneous communication was 18.5 months (95% CI, 15.9 to 21.0), and the
1 Was lost to follow-up median overall survival was 55.5 months (95%
CI, 48.2 to 62.0). Analyses of potential prognos-
Figure 1. Enrollment, Randomization, and Treatment. tic baseline factors such as age, International
Federation of Gynecology and Obstetrics stage at
first diagnosis, histologic subtype, treatment his-
days (interquartile range, 9 to 23), and chemo- tory that included previous maintenance therapy,
therapy was started a median of 35 days (inter- and platinum-free interval (6 to 12 months or
quartile range, 25 to 45 days) after surgery in >12 months) did not identify a subgroup of pa-
this group. Of the 206 patients assigned to sur- tients who did not benefit from surgery (Fig. 3).
gery, 192 (93.2%) underwent the procedure.
Reasons for not undergoing the operation were Other Assessments
an intercurrent illness that was diagnosed after The role of complete resection as a prognostic
randomization (8 patients), findings after ran- factor was analyzed as a secondary end point in
domization that suggested unresectability (4 pa- the surgery group. Among patients in the sur-
tients), loss to follow-up shortly after randomiza- gery group who had complete resection, the
tion (1 patient), and erroneous communication median overall survival was 61.9 months (95%
of a randomization result (1 patient) (Fig. 1). A CI, 55.3 to 78.9), as compared with 27.7 months
total of 8 patients in the no-surgery group under- (95% CI, 23.5 to 38.7) among patients who did
went surgery, and another 6 patients withdrew not have complete resection (Fig. S3). This dif-
consent or declined further treatment. A total of ference remained similar in magnitude when 14
32 of 201 patients (15.9%) in the no-surgery patients who were assigned to surgery but did
group crossed over to surgery because of subse- not undergo the procedure were excluded (data
quent relapse. not shown).
Macroscopic complete resection was achieved The majority of patients in both groups re-
in 75.5% (145 of 192) of the patients who were ceived at least five cycles of a platinum-contain-
assigned to surgery and underwent the proce- ing second-line therapy: 76.7% of the patients in
dure and in 70.4% (145 of 206) of all patients in the surgery group and 79.6% in the no-surgery
the surgery group. There were no deaths within group. A total of 94 patients (47 in each group)
30 days after surgery, and reoperation was per- received bevacizumab as part of second-line
formed in 3.7% (7 of 191 patients). The median therapy, and only a few patients received a PARP
operation time was 222 minutes (range, 150 to inhibitor during the trial (8 in the surgery group
300). Additional details regarding surgery and and 12 in the no-surgery group, with some
perioperative complications are provided in patients participating in double-blind, placebo-
Table S1. controlled trials of PARP inhibition). Kaplan–

2126 n engl j med 385;23  nejm.org  December 2, 2021

The New England Journal of Medicine

Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Cytoreductive Surgery for Relapsed Ovarian Cancer

Table 1. Baseline Characteristics of the Patients.*

Cytoreductive Surgery No Surgery

Characteristic (N = 206) (N = 201)
Median age (IQR) — yr 60.8 (54.2–67.3) 62.2 (54.2–69.9)
FIGO stage at first diagnosis — no. (%)†
I   24 (11.7) 19 (9.5)
II 20 (9.7)   26 (12.9)
III 145 (70.4) 143 (71.1)
IV 16 (7.8) 13 (6.5)
Missing data   1 (0.5) 0
Tumor histologic type — no. (%)
High-grade serous 173 (84.0) 155 (77.1)
Low-grade serous   4 (1.9)   6 (3.0)
Clear cell   1 (0.5)   9 (4.5)
Endometrioid 13 (6.3) 16 (8.0)
Mucinous   3 (1.5)   1 (0.5)
Other or mixed 12 (5.8) 14 (7.0)
Previous first-line therapy — no. (%)
Platinum-based 204 (99.0) 199 (99.0)
Taxane-based 193 (93.7) 180 (89.6)
Antiangiogenic drugs‡ 33 (16.0)   31 (15.4)
No previous chemotherapy 3 (1.5)   3 (1.5)
Platinum-free interval§
6–12 mo — no./total no. (%)   48/203 (23.6) 47/198 (23.7)
>12 mo — no./total no. (%) 155/203 (76.4) 151/198 (76.3)
Median (IQR) — mo 21.1 (12.9–33.6) 18.7 (12.8–32.1)
Tumor marker at baseline
Median CA-125 level (IQR) — U/ml 61 (28–116) 73 (33–166)
Site of relapse — no. (%)
Pelvis 125 (60.7) 112 (55.7)
Intraabdominal, above pelvis 103 (50.0) 110 (54.7)
Retroperitoneal   73 (35.4)   73 (36.3)
Parenchymal   26 (12.6)   30 (14.9)
Spleen 13 (6.3) 18 (9.0)
Liver 13 (6.3) 11 (5.5)
Pancreas 0   1 (0.5)
Other   7 (3.4)   5 (2.5)
Abdominal wall   4 (1.9)   3 (1.5)
Thoracic   3 (1.5)   2 (1.0)

* Percentages may not total 100 because of rounding. CA-125 denotes cancer antigen 125, and IQR interquartile range.
† International Federation of Gynecology and Obstetrics (FIGO) stages range from I to IV, with higher stages indicating
more advanced cancer.
‡ Patients had received the following antiangiogenic drugs as part of their care or because of participation in a previous
trial (some participants may have received placebo rather than the drug): bevacizumab (16 patients in the surgery
group and 15 in the no-surgery group), nintedanib or placebo (9 patients in the surgery group and 6 in the no-surgery
group), pazopanib (1 patient in the surgery group and 2 in the no-surgery group), pazopanib or placebo (6 patients in
the surgery group and 3 in the no-surgery group), sorafenib or placebo (1 patient in the no-surgery group), and treba-
nanib or placebo (1 patient in the surgery group and 4 in the no-surgery group). The drugs had been administered with
chemotherapy or as maintenance therapy or both.
§ Data are shown for 203 patients in the surgery group and 198 patients in the no-surgery group who had received previ-
ous platinum-based chemotherapy.

n engl j med 385;23  nejm.org  December 2, 2021 2127

The New England Journal of Medicine
Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

changes from baseline to 6 months were 9.0

A Overall Survival
1.00
points (95% CI, 1.1 to 17.0) on the insomnia
scale and 12.2 points (95% CI, 3.5 to 20.9) on
the constipation scale of the QLQ-C30, favoring
0.75 the no-surgery group (scores on each scale range
from 0 to 100, with higher scores indicating
Probability

0.50 Cytoreductive surgery worse symptoms). However, at the 6-month

evaluation, only 11 of 99 patients (11%) in the
no-surgery group, as compared with 32 of 85
0.25
No surgery
patients (38%) in the surgery group, were still
receiving chemotherapy. We did not observe any
0.00 differences regarding these symptoms after the
0 12 24 36 48 60 72 84
end of chemotherapy treatment at the 12-month
Months
evaluation (Fig. S1).
No. at Risk
Cytoreductive surgery 206 182 156 133 102 70 35 14
No surgery 201 180 154 115 87 50 20 7
Discussion
B Progression-free Survival Cytoreductive surgery in addition to platinum-
1.00
based chemotherapy in patients with relapsed
ovarian cancer resulted in a benefit with respect
0.75 to overall survival. Appropriate selection of pa-
tients and trial centers was crucial for the suc-
Probability

cess of this trial, and the importance of these

0.50
selections is reflected in both the high efficacy
and low morbidity in the trial. The observed in-
0.25 Cytoreductive surgery cidence of perioperative complications was lower
No surgery
than the incidence that has been reported
0.00 among patients with primary ovarian cancer.20
0 12 24 36 48 60 72 84 In the DESKTOP III trial, the number of patients
Months in whom complete resection was achieved was
No. at Risk high; therefore, many patients were not exposed
Cytoreductive surgery 206 140 68 46 36 28 13 5
No surgery 201 118 40 24 14 8 4 3
to a surgical burden unnecessarily without hav-
ing any potential benefit, and the power of the
Figure 2. Kaplan–Meier Estimates of Overall Survival and Progression-free trial was not diluted because of a large propor-
Survival. tion of patients who did not undergo successful
Tick marks indicate censored data. surgery. We anticipated that only complete re-
section could provide any benefit, and, conse-
quently, a surgical trial could be successful only
Meier curves of progression-free and overall with a sufficiently large number of patients in
survival in the subgroup of patients who did not whom complete resection is achieved. To enrich
receive bevacizumab as part of second-line ther- the trial population, we had developed a predic-
apy are shown in Figure S2. The results of this tive selection tool — the AGO score. This trial
exploratory subgroup analysis did not indicate confirmed the usefulness of the score, which
any influence of bevacizumab maintenance ther- predicted complete resection in 76% of the pa-
apy on survival or on the effect of surgery. The tients, a figure much higher than that reported
subgroup of patients who received bevacizumab in unselected patients in multicenter trials of sys-
therapy was too small for meaningful analysis. temic therapy in women with primary advanced
Results of quality-of-life analyses did not ovarian cancer (usually 50% or even lower).21-23
show any between-group differences with re- Four additional randomized trials evaluating
spect to global health status, quality of life, or the role of surgery in recurrent ovarian cancer
any functional subscale at baseline, visit 1 (at are under way worldwide. Unfortunately, two of
6 months), or visit 2 (at 12 months). Model-based them (EORTC 55963; ClinicalTrials.gov number,
estimates of the between-group difference in the NCT00006356; and Surgery for Ovarian Cancer

2128 n engl j med 385;23  nejm.org  December 2, 2021

The New England Journal of Medicine

Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Cytoreductive Surgery for Relapsed Ovarian Cancer

Cytoreductive
Subgroup Surgery No Surgery Hazard Ratio for Death (95% CI)
deaths/no. of patients
Overall 120/206 134/201 0.75 (0.59–0.96)
Age
<65 yr 78/132 81/124 0.84 (0.62–1.16)
≥65 yr 42/74 53/77 0.62 (0.41–0.94)
FIGO stage at first diagnosis
I or II 25/44 28/45 0.87 (0.49–1.53)
III or IV 94/161 106/156 0.72 (0.54–0.96)
Tumor histologic type
High-grade serous 102/173 105/155 0.74 (0.56–0.97)
Other 18/33 29/46 0.80 (0.43–1.50)
Previous antiangiogenic therapy
Yes 25/33 19/31 0.75 (0.39–1.43)
No 95/173 115/170 0.74 (0.56–0.97)
Platinum-free interval
6–12 mo 33/48 39/47 0.59 (0.37–0.94)
>12 mo 86/155 94/151 0.83 (0.62–1.11)
0.40 0.50 0.75 1.00 1.50 2.00

Cytoreductive Surgery Better No Surgery Better

Figure 3. Analyses of Overall Survival According to Prognostic Baseline Factors.

International Federation of Gynecology and Obstetrics (FIGO) stages range from I to IV, with higher stages indicating more advanced
cancer; the FIGO stage was missing for one patient in the surgery group. The dashed line indicates a hazard ratio of 0.75.

Recurrence [SOCCER]; Netherlands Trial Regis- survival. Complete resection was achieved in 77%
ter number, NTR3337) were stopped because of of patients. The results of progression-free sur-
low recruitment. The surgical part of the Gyne- vival, which were reported recently, showed a
cologic Oncology Group (GOG)–0213 trial com- significant benefit of adding cytoreductive sur-
pleted recruitment, and the results have been gery to chemotherapy.25 As in our trial, the effect
published.24 A total of 485 patients in whom was limited to patients in whom a macroscopic
complete resection was deemed feasible by the complete resection was achieved. Mature data re-
investigator underwent randomization in that garding overall survival in this trial are awaited.
trial. Complete resection was achieved in 67% of Do we have any explanation for the inconsis-
the patients assigned to surgery who underwent tent results of the DESKTOP III and SOC-1 trials
the procedure. After a futility analysis indicated on the one hand and the GOG-0213 trial on the
a low chance of showing a positive trial, the data other hand? One observed difference among the
were locked and analyzed and were published trials was that 84% of the patients in the GOG-
with an updated follow-up. The trial failed to 0213 trial received bevacizumab, whereas 23% in
show a survival benefit of adding surgery to sys- the DESKTOP III trial and 1% in the SOC-1 trial
temic treatment. received bevacizumab. Could the activity of beva­
The Surgery or Chemotherapy in Recurrent cizumab have modified the effect of surgery?
Ovarian Cancer (SOC-1) trial also completed re- This question should be discussed cautiously,
cruitment: 357 patients were randomly assigned since DESKTOP III is a pure surgical trial and
to undergo surgery and then receive chemother- GOG-0213 used a mixture of chemotherapy and
apy or to receive chemotherapy alone.25 Patients surgery. In the subgroup of patients in the GOG-
were selected for enrollment if they had poten- 0213 trial who did not receive maintenance
tially resectable disease as predicted with the treatment with bevacizumab, the median overall
use of the international model (iMODEL) or if survival was 32.4 months among patients who
the surgeon determined that resection was pos- underwent surgery and 67.0 months among those
sible on the basis of positron-emission tomogra- who did not undergo surgery, suggesting a detri-
phy and computed tomography. The primary end mental effect of performing surgery before che-
points were progression-free survival and overall motherapy is administered.22 However, this post

n engl j med 385;23  nejm.org  December 2, 2021 2129

The New England Journal of Medicine
Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

hoc observation is based on the results in only and overall survival, with an acceptable inci-
a small subgroup that was not prospectively de- dence of complications and without a detrimen-
fined. Furthermore, this finding contradicts other tal effect on quality of life in patients selected
clinical studies about the effect of successful sur- for inclusion on the basis of the AGO score. All
gery when added to chemotherapy in ovarian can- patients with a first relapse after a platinum-free
cer.26,27 Although the subgroup of patients who interval of at least 6 months may be evaluated to
did not receive bevacizumab in the DESKTOP III assess whether surgery is an option, and the AGO
trial (77% of the trial population) was larger score may be incorporated into this process.
than that in the GOG-0213 trial, our data indi- Eligible patients could receive counseling about
cated that there was a benefit in these patients the options for cytoreductive surgery in centers
that was similar to that in the analysis of the of gynecologic oncology that have experience in
primary end point in the intention-to-treat pop- surgery for relapsed ovarian cancer. In contrast,
ulation. However, such an analysis in a trial that patients who have a high probability of incom-
is not conducted in a blinded manner and not plete resection on the basis of disease or clinical
placebo-controlled has substantial bias. Subse- characteristics should not be exposed to a poten-
quent therapy was not standardized, and the tially harmful surgical treatment.
choice of therapy might have been influenced by The results of this trial cannot be extrapo-
randomization outcome, residual disease after lated to interval debulking after chemotherapy
surgery for relapse, postoperative complications, or to the treatment of relapse after later lines of
the patients’ wishes, or the preferences of the treatment. These scenarios deserve further study
investigators. Therefore, it is not adequate to that should also focus on the potential interac-
conclude that subsequent systemic therapy could tion of surgery with new drugs such as PARP
compensate for incomplete surgery or no sur- inhibitors or further targeted therapies.
gery for relapsed disease or that systemic ther­ Presented in part at the Annual Meeting of the American
apy could make successful surgery unnecessary. Society of Clinical Oncology, Chicago, 2017, and online, 2020.
Supported by the Arbeitsgemeinschaft Gynäkologische
Another relevant difference between the Onkologie (AGO) Study Group, GlaxoSmithKline, and Medac.
DESKTOP III and SOC-1 trials and the GOG- Disclosure forms provided by the authors are available with
0213 trial is the process of selecting patients and the full text of this article at NEJM.org.
A data sharing statement provided by the authors is available
centers. The strict selection process in the cur- with the full text of this article at NEJM.org.
rent trial was intended to identify a subgroup of We thank the members of the independent data monitor-
patients in whom surgical resection would most ing committee (Robert L. Coleman [chair], Jonathan S. Berek,
Dennis S. Chi, and Jim Paul) and the following study groups
likely be successful. Finally, another potential for their participation in this European Network of Gyneco-
factor is the different selection of centers in the logic Oncology Trialists–Gynecologic Cancer Intergroup trial:
three trials. It is not easy to quantify these fac- AGO Study Group, AGO Austria, Belgian Gynecologic Oncol-
ogy Group, Spanish Ovarian Cancer Research Group, Group
tors. A meta-analysis of the DESKTOP III and d’Investigateurs Nationaux pour l’Etude des Cancers Ovariens,
GOG-0213 trials to better understand the differ- Korean Gynecologic Oncology Group, Mario Negri Gynecologic
ences is under way. Oncology, Multicenter Italian Trials in Ovarian Cancer, United
Kingdom National Cancer Research Institute, Nordic Society of
The results of the current trial showed a bene­ Gynecologic Oncology, and Shanghai Gynecologic Oncology
fit of surgery with respect to progression-free Group.

Appendix
The authors’ full names and academic degrees are as follows: Philipp Harter, M.D., Ph.D., Jalid Sehouli, M.D., Ph.D., Ignace Vergote,
M.D., Ph.D., Gwenael Ferron, M.D., Ph.D., Alexander Reuss, M.Sc., Werner Meier, M.D., Ph.D., Stefano Greggi, M.D., Ph.D., Berit J.
Mosgaard, M.D., Ph.D., Frederic Selle, M.D., Ph.D., Frédéric Guyon, M.D., Ph.D., Christophe Pomel, M.D., Ph.D., Fabrice Lécuru, M.D.,
Ph.D., Rongyu Zang, M.D., Ph.D., Elisabeth Avall‑Lundqvist, M.D., Ph.D., Jae‑Weon Kim, M.D., Ph.D., Jordi Ponce, M.D., Ph.D., Fran-
cesco Raspagliesi, M.D., Ph.D., Gunnar Kristensen, M.D., Ph.D., Jean‑Marc Classe, M.D., Ph.D., Peter Hillemanns, M.D., Ph.D., Per-
nille Jensen, M.D., Ph.D., Annette Hasenburg, M.D., Ph.D., Sadaf Ghaem‑Maghami, M.D., Ph.D., Mansoor R. Mirza, M.D., Ph.D.,
Bente Lund, M.D., Ph.D., Alexander Reinthaller, M.D., Ph.D., Ana Santaballa, M.D., Ph.D., Adeola Olaitan, M.D., Ph.D., Felix Hilpert,
M.D., Ph.D., and Andreas du Bois, M.D., Ph.D.
The authors’ affiliations are as follows: the Department of Gynecology and Gynecologic Oncology, Kliniken Essen-Mitte, Essen (P.
Harter, A.B.), the Department of Gynecology with Center for Oncological Surgery, Charité Berlin, Berlin (J.S.), the Coordinating Center
for Clinical Trials (A. Reuss) and Department of Gynecology (P. Harter), Philipps University, Marburg, the Department of Obstetrics
and Gynecology, Heinrich-Heine-University, Duesseldorf (W.M.), the Department of Gynecology and Obstetrics, Hannover Medical
School, Hannover (P. Hillemanns), the University Medical Center Mainz, Department of Gynecology and Obstetrics, Mainz (A.H.), and
Mammazentrum Hamburg at Jerusalem Hospital, Hamburg (F.H.) — all in Germany; the Department of Gynecological Oncology,
University Hospitals Leuven, Leuven Cancer Institute, Leuven, Belgium (I.V.); the Department of Surgical Oncology, Institut Claudius

2130 n engl j med 385;23  nejm.org  December 2, 2021

The New England Journal of Medicine

Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Cytoreductive Surgery for Relapsed Ovarian Cancer

Regaud, Institut Universitaire du Cancer, Toulouse (G.F.), the Department of Medical Oncology, Groupe Hospitalier Diaconesses Croix
Saint-Simon (F.S.), and Institut Curie, Oncologie Gynécologique and Université de Paris (F.L.), Paris, the Department of Surgical Oncol-
ogy, Institut Bergonié, Comprehensive Cancer Center, Bordeaux (F.G.), the Department of Surgical Oncology, Jean Perrin Cancer
Center, Clermont-Ferrand (C.P.), and Institut de Cancerologie de l’Ouest, Université de Médecine, Nantes (J.-M.C.) — all in France; the
Gynecologic Oncology Unit, Istituto Nazionale Tumori di Napoli, Fondazione IRCCS Pascale, Naples (S.G.), and Fondazione IRCCS
Istituto Nazionale Tumori, Milan (F.R.) — both in Italy; the Gyne-Oncology Department of Gynecology (B.J.M.), Copenhagen Univer-
sity Hospital Rigshospitalet (M.R.M.), Copenhagen, the Department of Gynecology, Aarhus University Hospital and Aarhus University,
Institute of Clinical Medicine, Faculty of Health, Aarhus (P.J.), and Aalborg University Hospital, Aalborg (B.L.) — all in Denmark; the
Ovarian Cancer Program, Department of Gynecologic Oncology, Fudan University Zhongshan Hospital, Shanghai, China (R.Z.); the
Department of Oncology–Pathology, Karolinska Institutet, Stockholm (E.A.-L.); the Department of Obstetrics and Gynecology, Seoul
National University College of Medicine, Seoul, South Korea (J.-W.K.); University Hospital of Bellvitge, Barcelona (J.P.), and the Gyne-
cologic Oncology Unit, La Fe University Hospital, Valencia (A.S.) — both in Spain; the Department of Gynecologic Oncology and Insti-
tute for Cancer Genetics and Informatics, Oslo University Hospital, Oslo (G.K.); the Department of Surgery and Cancer, Imperial Col-
lege London (S.G.-M.), and the Department of Gynaecological Oncology, University College London Hospital (A.O.) — both in London;
and the Department of Gynecology and Gynecological Oncology, Medical University of Vienna, Vienna (A. Reinthaller).

References
1. du Bois A, Quinn M, Thigpen T, et al. et al. Final overall survival (OS) results testing procedure for clinical trials. Bio-
2004 consensus statements on the man- from SOLO2/ENGOT-ov21: a phase III metrics 1979;​35:​549-56.
agement of ovarian cancer: final docu- trial assessing maintenance olaparib in 20. Harter P, Sehouli J, Lorusso D, et al.
ment of the 3rd International Gynecologic patients (pts) with platinum-sensitive, re- A randomized trial of lymphadenectomy
Cancer Intergroup Ovarian Cancer Con- lapsed ovarian cancer and a BRCA muta- in patients with advanced ovarian neo-
sensus Conference (GCIG OCCC 2004). tion. J Clin Oncol 2020;​38:​6002. abstract plasms. N Engl J Med 2019;​380:​822-32.
Ann Oncol 2005;​16:​Suppl 8:​viii7-viii12. (https://ascopubs​.­org/​­doi/​­10​.­1200/​­JCO​.­2020​ 21. du Bois A, Kristensen G, Ray-Coquard
2. Burger RA, Brady MF, Bookman MA, .­38​.­15_suppl​.­6002). I, et al. Standard first-line chemotherapy
et al. Incorporation of bevacizumab in 12. Harter P, du Bois A. The role of sur- with or without nintedanib for advanced
the primary treatment of ovarian cancer. gery in ovarian cancer with special em- ovarian cancer (AGO-OVAR 12): a random­
N Engl J Med 2011;​365:​2473-83. phasis on cytoreductive surgery for recur- ised, double-blind, placebo-controlled
3. Perren TJ, Swart AM, Pfisterer J, et al. rence. Curr Opin Oncol 2005;​17:​505-14. phase 3 trial. Lancet Oncol 2016;​17:​78-89.
A phase 3 trial of bevacizumab in ovarian 13. Harter P, du Bois A, Hahmann M, et al. 22. Clamp AR, James EC, McNeish IA,
cancer. N Engl J Med 2011;​365:​2484-96. Surgery in recurrent ovarian cancer: the et al. Weekly dose-dense chemotherapy in
4. Moore K, Colombo N, Scambia G, et al. Arbeitsgemeinschaft Gynaekologische On­ first-line epithelial ovarian, fallopian tube,
Maintenance olaparib in patients with kologie (AGO) DESKTOP OVAR trial. Ann or primary peritoneal carcinoma treat-
newly diagnosed advanced ovarian can- Surg Oncol 2006;​13:​1702-10. ment (ICON8): primary progression free
cer. N Engl J Med 2018;​379:​2495-505. 14. du Bois A, Reuss A, Pujade-Lauraine survival analysis results from a GCIG
5. González-Martín A, Pothuri B, Vergote E, Harter P, Ray-Coquard I, Pfisterer J. phase 3 randomised controlled trial. Lan-
I, et al. Niraparib in patients with newly Role of surgical outcome as prognostic cet 2019;​394:​2084-95.
diagnosed advanced ovarian cancer. N Engl factor in advanced epithelial ovarian can- 23. Pignata S, Scambia G, Katsaros D,
J Med 2019;​381:​2391-402. cer: a combined exploratory analysis of et al. Carboplatin plus paclitaxel once a
6. Coleman RL, Fleming GF, Brady MF, 3 prospectively randomized phase 3 multi- week versus every 3 weeks in patients
et al. Veliparib with first-line chemother- center trials: by the Arbeitsgemeinschaft with advanced ovarian cancer (MITO-7):
apy and as maintenance therapy in ovarian Gynaekologische Onkologie Studiengruppe a randomised, multicentre, open-label,
cancer. N Engl J Med 2019;​381:​2403-15. Ovarialkarzinom (AGO-OVAR) and the phase 3 trial. Lancet Oncol 2014;​15:​396-
7. Ray-Coquard I, Pautier P, Pignata S, Groupe d’Investigateurs Nationaux Pour les 405.
et al. Olaparib plus bevacizumab as first- Etudes des Cancers de l’Ovaire (GINECO). 24. Coleman RL, Spirtos NM, Enserro D,
line maintenance in ovarian cancer. N Engl Cancer 2009;​115:​1234-44. et al. Secondary surgical cytoreduction
J Med 2019;​381:​2416-28. 15. Harter P, Sehouli J, Reuss A, et al. Pro- for recurrent ovarian cancer. N Engl J Med
8. Parmar MKB, Ledermann JA, Colombo spective validation study of a predictive 2019;​381:​1929-39.
N, et al. Paclitaxel plus platinum-based score for operability of recurrent ovarian 25. Shi T, Zhu J, Feng Y, et al. Secondary
chemotherapy versus conventional plati- cancer: the Multicenter Intergroup Study cytoreduction followed by chemotherapy
num-based chemotherapy in women DESKTOP II: a project of the AGO Kommis- versus chemotherapy alone in platinum-
with relapsed ovarian cancer: the ICON4/ sion OVAR, AGO Study Group, NOGGO, sensitive relapsed ovarian cancer (SOC-1):
AGO-OVAR-2.2 trial. Lancet 2003;​ 361:​ AGO-Austria, and MITO. Int J Gynecol a multicentre, open-label, randomised,
2099-106. Cancer 2011;​21:​289-95. phase 3 trial. Lancet Oncol 2021;​22:​439-49.
9. Pfisterer J, Shannon CM, Baumann K, 16. Rosenberger WF, Lachin JM. Random- 26. Hall M, Savvatis K, Nixon K, et al.
et al. Bevacizumab and platinum-based ization in clinical trials: theory and prac- Maximal-effort cytoreductive surgery for
combinations for recurrent ovarian can- tice. New York:​John Wiley, 2002. ovarian cancer patients with a high tumor
cer: a randomised, open-label, phase 3 17. du Bois A, Reuss A, Pujade-Lauraine burden: variations in practice and impact
trial. Lancet Oncol 2020;​21:​699-709. E, et al. European Network of Gynaeco- on outcome. Ann Surg Oncol 2019;​26:​
10. Coleman RL, Brady MF, Herzog TJ, et al. logical Oncological Trial Groups’ require- 2943-51.
Bevacizumab and paclitaxel-carboplatin ments for trials between academic groups 27. Klein DA, Mann AK, Freeman AH,
chemotherapy and secondary cytoreduc- and industry partners — first update 2015. Liao C-I, Kapp DS, Chan JK. Chemothera-
tion in recurrent, platinum-sensitive ovar- Int J Gynecol Cancer 2015;​25:​1328-30. py alone for patients 75 years and older
ian cancer (NRG Oncology/Gynecologic 18. Grambsch P, Therneau T. Proportional with epithelial ovarian cancer — is inter-
Oncology Group study GOG-0213): a mul- hazards tests and diagnostics based on val cytoreductive surgery still needed?
ticentre, open-label, randomised, phase 3 weighted residuals. Biometrika 1994;​81:​ Am J Obstet Gynecol 2020;​222(2):​170.e1-
trial. Lancet Oncol 2017;​18:​779-91. 515-26. DOI:​10.2307/2337123. 170.e11.
11. Poveda A, Floquet A, Ledermann JA, 19. O’Brien PC, Fleming TR. A multiple Copyright © 2021 Massachusetts Medical Society.

n engl j med 385;23  nejm.org  December 2, 2021 2131

The New England Journal of Medicine
Downloaded from nejm.org on March 12, 2023. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.