FOXTROT Trial Lancet Onc2012
FOXTROT Trial Lancet Onc2012
FOXTROT Trial Lancet Onc2012
Summary
Lancet Oncol 2012; 13: 1152–60 Background Preoperative (neoadjuvant) chemotherapy and radiotherapy are more effective than similar postoperative
Published Online treatment for oesophageal, gastric, and rectal cancers, perhaps because of more effective micrometastasis eradication
September 25, 2012 and reduced risk of incomplete excision and tumour cell shedding during surgery. The FOxTROT trial aims to
http://dx.doi.org/10.1016/
investigate the feasibility, safety, and efficacy of preoperative chemotherapy for colon cancer.
S1470-2045(12)70348-0
See Comment page 1073
Methods In the pilot stage of this randomised controlled trial, 150 patients with radiologically staged locally advanced
*See Acknowledgments for full
list of FOxTROT collaborators
(T3 with ≥5 mm invasion beyond the muscularis propria or T4) tumours from 35 UK centres were randomly
assigned (2:1) to preoperative (three cycles of OxMdG [oxaliplatin 85 mg/m², l-folinic acid 175 mg, fluorouracil
Correspondence to:
The FOxTROT Trial, Birmingham 400 mg/m² bolus, then 2400 mg/m² by 46 h infusion] repeated at 2-weekly intervals followed by surgery and a
Clinical Trials Unit, Robert Aitken further nine cycles of OxMdG) or standard postoperative chemotherapy (12 cycles of OxMdG). Patients with KRAS
Institute, University of wild-type tumours were randomly assigned (1:1) to receive panitumumab (6 mg/kg; every 2 weeks with the first
Birmingham, Edgbaston,
Birmingham B15 2TT, UK
6 weeks of chemotherapy) or not. Treatment allocation was through a central randomisation service using a
FOxTROT-trial@contacts. minimised randomisation procedure including age, radiological T and N stage, site of tumour, and presence of
bham.ac.uk defunctioning colostomy as stratification variables. Primary outcome measures of the pilot phase were feasibility,
safety, and tolerance of preoperative therapy, and accuracy of radiological staging. Analysis was by intention to treat.
This trial is registered, number ISRCTN 87163246.
Findings 96% (95 of 99) of patients started and 89% (85 of 95) completed preoperative chemotherapy with grade 3–4
gastrointestinal toxicity in 7% (seven of 94) of patients. All 99 tumours in the preoperative group were resected, with
no significant differences in postoperative morbidity between the preoperative and control groups: 14% (14 of 99)
versus 12% (six of 51) had complications prolonging hospital stay (p=0·81). 98% (50 of 51) of postoperative
chemotherapy patients had T3 or more advanced tumours confirmed at post-resection pathology compared with 91%
(90 of 99) of patients following preoperative chemotherapy (p=0·10). Preoperative therapy resulted in significant
downstaging of TNM5 compared with the postoperative group (p=0·04), including two pathological complete
responses, apical node involvement (1% [one of 98] vs 20% [ten of 50], p<0·0001), resection margin involvement (4%
[ four of 99] vs 20% [ten of 50], p=0·002), and blinded centrally scored tumour regression grading: 31% (29 of 94) vs 2%
(one of 46) moderate or greater regression (p=0·0001).
Interpretation Preoperative chemotherapy for radiologically staged, locally advanced operable primary colon cancer is
feasible with acceptable toxicity and perioperative morbidity. Proceeding to the phase 3 trial, to establish whether the
encouraging pathological responses seen with preoperative therapy translates into improved long-term oncological
outcome, is appropriate.
N2), site of primary tumour, and defunctioning colos- patients whose tumours tested as KRAS wild-type were
tomy (yes, no). Treatment was open-label. entered in the panitumumab randomisation.
Soon after the start of FOxTROT, the poor effectiveness
of EGFR monoclonal antibodies in tumours harbouring Procedures
a KRAS mutation18–20 mandated introduction of pre- Chemotherapy was the standard UK modified de
operative KRAS testing of diagnostic biopsy samples, Gramont (OxMdG) regimen,32 consisting of cycles of
with only KRAS wild-type tumours eligible for the oxaliplatin at 85 mg/m² combined with l-folinic acid
panitumumab randomisation. A two-stage consent and 175 mg/m2 plus fluorouracil 400 mg/m² by intravenous
randomisation procedure was introduced. Patients were bolus, followed by a 46 h infusion of 2400 mg/m² through
first invited to consent to molecular testing, then to an indwelling line, repeated at 2-weekly intervals.
randomisation for either preoperative plus postoperative Capecitabine could not be substituted for fluorouracil
chemotherapy or postoperative chemotherapy only. Only and folinic acid in this pilot study because of higher
toxicity when combined with panitumumab.17 Dose
reductions and delays of up to 4 weeks were allowed for
Preoperative plus Postoperative reversible toxicity. Preoperative chemotherapy duration
postoperative chemotherapy
chemotherapy (n=99) only (n=51)
was only 6 weeks (three cycles of OxMdG) to minimise
the risk of progression of chemoresistant tumours
Age (years)
(15–20% of advanced metastatic colon cancers progress
Median (IQR) 64 (59–68) 65 (56–69)
during 12 weeks of similar combination chemo-
Range 31–82 38–78
therapy).16,17,24,26 Surgery with curative intent was under-
Age group (years) taken at least 3 weeks after completing preoperative
<50 9 (9%) 4 (8%) therapy, to reduce perioperative morbidity,2 followed by a
50–59 16 (16%) 10 (20%) further 18 weeks (nine cycles) of OxMdG. CT scans were
60–69 50 (51%) 26 (51%) repeated before surgery in the preoperative group. For
≥70 24 (24%) 11 (22%) the patients who were not assigned to receive preoperative
Sex chemotherapy, postoperative chemotherapy duration was
Male 65 (66%) 32 (63%) 24 weeks (12 cycles) of OxMdG. If allocated, panitumumab
Female 34 (34%) 19 (37%) (6 mg/kg) was given by intravenous infusion at 2-weekly
Colonic obstruction 3/99 (3%) 1/51 (2%) intervals during the first 6 weeks of chemotherapy
WHO performance status (preoperative or postoperative).
0 67 (68%) 34 (67%) The primary outcomes of the pilot phase were feasibility,
1 30 (30%) 17 (33%) safety, tolerance of preoperative therapy, and the accuracy
2 2 (2%) 0 of radiological staging. Other key outcomes were
Location completion of planned surgery, perioperative morbidity,
Caecum 23 (23%) 11 (22%) timely completion of preoperative KRAS testing, and
Ascending colon 18 (18%) 11 (22%) downstaging of the resected tumour as measured by
Hepatic flexure 5 (5%) 3 (6%) histopathological tumour diameter and stage. Tumour
Transverse colon 7 (7%) 3 (6%) staging was standardised across centres, with training
Splenic flexure 3 (3%) 1 (2%) provided for participating specialist gastrointestinal
Descending colon 3 (3%) 3 (6%) radiologists and histopathologists. Pathological reports
Sigmoid 32 (32%) 15 (29%) and tumour blocks were collected centrally with tumour
Rectosigmoid 8 (8%) 4 (8%) regression grades scored masked to treatment allocation.
Radiological T-stage
T3 69 (70%) 35 (69%) Statistical analysis
T4 30 (30%) 16 (31%) The FOxTROT study aims to randomly assign at least
Radiological N-stage
1050 patients to detect a 25% proportional reduction
Nx 3 (3%) 1 (2%)
(roughly 8% absolute difference) in recurrence at 2 years
N0 23 (23%) 12 (24%)
(eg, 32% reduced to 24%) with 80% power at p<0·05.
N1 44 (45%) 22 (43%)
The prespecified sample size of 150 for the pilot phase
was chosen pragmatically as a sufficient number to
N2 29 (29%) 16 (31%)
assess the potential rate of recruitment and any large
Extramural vascular invasion 57/98 (58%)* 31/51 (61%)
differences in other primary outcomes. An independent
Mean (SD) 12·9 (8·8) 15·5 (9·9)
steering committee advised whether to continue to the
Range 1–50 5–50
full study. Comparisons of preoperative versus post-
One radiology form had missing extramural vascular invasion data. operative chemotherapy were by intention to treat
including all patients randomly assigned to treatment
Table 1: Patient characteristics at baseline radiology
groups, ignoring panitumumab allocation, and using
Results
8 had no postoperative
Between May 15, 2008, and Sept 21, 2010, 150 patients chemotherapy
from 35 UK centres were randomly assigned to receive
either preoperative plus postoperative chemotherapy 99 underwent surgery 40 started postoperative chemotherapy
(n=99) or standard postoperative chemotherapy alone 38 completed first three cycles (1 died, 1 stopped
after chest pain)
(n=51). Radiologically classified tumour characteristics
were similarly distributed across treatment groups
(table 1). Patient screening logs suggested that the main 15 had no postoperative
chemotherapy
reason for clinicians not entering radiologically eligible 2 had off-protocol treatment
patients was unsuitability for combination chemotherapy
because of age or frailty. Patients were also excluded
82 started postoperative chemotherapy 38 started second three cycles of postoperative
because of uncertainty about eligibility on CT staging, 79 completed first three cycles (3 stopped due chemotherapy
principally on whether depth of invasion beyond the to toxicity) 36 completed (2 stopped due to toxicity)
muscularis propria was 5 mm or more. Premature
scheduling of the operation date was also a common 1 stopped due to toxicity 1 possible liver metastasis
exclusion reason before the multidisciplinary teams and 1 missing data 1 missing data
individual clinicians adapted to the new care pathway.
138 of the 150 patients were eligible for KRAS testing to 77 started second three cycles of postoperative 34 started third three cycles of postoperative
establish eligibility for the panitumumab randomisation: chemotherapy chemotherapy
76 completed (1 stopped, reason unknown) 33 completed (1 stopped due to toxicity)
two were entered before KRAS testing was introduced,
six were entered while the panitumumab randomisation
was suspended, and four were from centres not taking 1 possible cerebrovascular accident 1 missing data
1 unknown
part in the panitumumab randomisation (appendix).
Biopsy samples were obtained for 98% (135 of 138) of
74 started final three cycles of postoperative 32 started final three cycles of postoperative
these patients and 96% (130 of 135) were successfully chemotherapy chemotherapy
tested. 72% (95 of 132 with known KRAS status, including 67 completed (4 stopped due to toxicity, 29 completed (2 stopped due to toxicity,
the two patients retrospectively tested) were KRAS wild- 1 venous access problems, 2 unknown) 1 venous access problems)
82% (67 of 82) completed all 18 weeks of postoperative 12% (six of 51) had pT3 tumours without additional risk
treatment. Toxicity was the most common reason for factors. Radiology was less accurate in discriminating
stopping treatment, with nine patients stopping at some between T3 and T4 stage, with 47% (eight of 17) of pT4
point during postoperative chemotherapy (four during or tumours also T4 on radiological assessment and 50%
after the first three cycles of postoperative chemotherapy, (8 of 16) of those T4 on radiology were pT3 on pathological
one after the second three cycles, and four during or after assessment. Radiological assessment was more sensitive
the final three cycles). in detecting nodal spread with 83% (20 of 24) of patho-
Of 51 patients allocated postoperative chemotherapy, logical node-positive patients also node-positive on radio-
94% (48 of 51) had resectional surgery with one dying logical assessment. However, specificity was low with
beforehand and two who were not resected because of radiological staging tending to overestimate tumour
inoperable peritoneal spread detected at surgery (figure 2). spread: 44% (16 of 36) of those node-positive on radiology
78% (40 of 51) started postoperative chemotherapy with were node-negative on pathological assessment. Simi-
11 not having postoperative chemotherapy because of low- larly, 52% (14 of 27) of those with extramural vascular
risk pathology (seven) or dying beforehand (four). 95% invasion on radiology were EMVI-negative on patho-
(38 of 40) of those starting completed the first 6 weeks and logical assessment.
72% (29 of 40) completed all 24 weeks of chemotherapy. One patient in each group needed acute surgery
Toxicity caused six patients to discontinue chemotherapy. because of incipient obstruction, with both proceeding to
Thus, a higher proportion of patients started pre- resectional surgery. No significant differences were seen
operative than postoperative chemotherapy (96% [95 of in complication rates (table 2) or median time to hospital
99] vs 78% [40 of 51]; p=0·001), and chemotherapy discharge (preoperative plus postoperative chemotherapy
completion rates were also higher in the preoperative group: 7 days, IQR 5–10; postoperative only: 6 days, 3–8;
therapy group with 68% (67 of 99) of those allocated pre p=0·18) although there was a greater proportion of wound
plus postoperative chemotherapy completing 24 weeks of infections in the preoperative group (table 2). The
treatment compared with 57% (29 of 51) of postoperative proportions of patients with complications that prolonged
chemotherapy patients (p=0·19). hospital stay, or had procedures that resulted in a stoma
Accuracy of radiological staging was assessed by or required further adominal surgery for anastomotic
pathological examination of resected tumours from the complications, were much the same for the two groups
postoperative chemotherapy only group. 86% (43 of 50 (table 2). The median delay after surgery to starting
[including two that were unresected]) of tumours had chemotherapy was 47 days (IQR 40–55) for patients
adverse features (inoperable tumour, positive lymph allocated preoperative chemotherapy and 53 days (44–57)
nodes, extramural vascular invasion [EMVI], or depth of for patients allocated postoperative chemotherapy.
invasion ≥5 mm) on pathological examination, indicative Of those receiving chemotherapy, toxicity was similar
of a greater than 50% recurrence risk at 3 years. Radiology between groups in the first 6 weeks and in subsequent
accurately identified invasion of the muscularis propria, treatment (appendix). 34% (32 of 94) of preoperative and
with only 2% (one of 51) of patients in the postoperative 31% (12 of 39) of postoperative chemotherapy patients had
group having a pathological T2 (pT2) tumour; a further grade 3 or worse toxicity in their first 6 weeks of
chemotherapy (p=0·72). 7% (seven of 94) of preoperative
and 10% (four of 39) of postoperative chemotherapy
Preoperative plus Postoperative p value patients had grade 3 or worse gastrointestinal toxicity; 11%
postoperative chemotherapy
(nine of 85) versus 21% (eight of 38) required dose
chemotherapy(n=99) only (n=51)
reductions (p=0·12) and 11% (ten of 95) versus 5% (two of
Anastomotic leak 5 (5%) 2 (4%) 0·77
40) did not complete their first 6 weeks of chemotherapy
Wound infection with or without intra-abdominal 13 (13%) 4 (8%) 0·34 (p=0·31). Only one patient had surgery delayed, for
abscess*
2 weeks, because of toxicity of preoperative chemo-
Bronchopneumonia 2 (2%) 0 0·31
therapy—grade 4 neutropenia. For all 24 weeks of
Deep vein thrombosis 2 (2%) 0 0·31
chemotherapy, 49% of patients (47 of 95) in the preoperative
Rash 3 (3%) 0 0·21
chemotherapy group and 51% of patients (20 of 39) in the
Neutropenia 1 (1%) 0 0·47
postoperative group had any grade 3 or higher adverse
Death 0 1 (2%) 0·16
events, with the most common being haematological (29%
Other 12 (12%) 6 (12%) 0·72
[28 of 95] and 28% [11 of 39]) and gastrointestinal (19%
Complication prolonging hospital stay 14 (14%) 6 (12%) 0·81
[18 of 95] and 21% [eight of 39]) adverse events (appendix).
Procedure resulting in a stoma 12 (12%) 5 (10%) 0·66
Significant differences favouring preoperative therapy
Further abdominal surgery needed 4 (4%) 2 (4%) 0·96 were seen in apical node involvement in the resected
*All three patients with intra-abdominal abscess also had wound infection recorded. specimens (one of 98 vs ten of 50; p<0·0001), TNM5
staging (p=0·04), resection margin involvement (four of
Table 2: Perioperative complications in the preoperative plus postoperative chemotherapy group
99 vs ten of 50; p=0·002), and retroperitoneal margin
compared with the postoperative chemotherapy group
involvement (five of 94 vs eight of 44; p=0·016; table 3).
Data are n (%), mean (SD), or median (IQR). *The two extra tumours are the two that were unresected. Macroscopic evaluation was used. †Either T4, T3 with extramural
vascular invasion, or T3 with ≥5 mm invasion of the muscularis propria. MH=Mantel-Haenszel test, TNM5=American Joint Committee on Cancer, fifth edn.
31% of (29 of 94) tumours in the preoperative group with depth of spread beyond the muscularis propria in the
blinded centrally scored regression grades showed preoperative group compared with controls. There were
moderate to complete regression as compared with 2% two complete pathological responses and seven T2
(one of 46) in the control series (p=0·0001). There were tumours in the preoperative group, with only one T2
reductions in tumour diameter, tumour thickness, and tumour in the control group. However, downstaging as
assessed by T stage was not significant (table 3). The of patients in the preoperative chemotherapy group had
proportion of tumours with involved lymph nodes was apparent progression between diagnostic radiology and
lower (40% [39 of 98] vs 52% [26 of 50], test for association post-resection pathology but, reassuringly, no patients
p=0·039) in the preoperative chemotherapy group, with developed incipient obstruction during the 6-week
the number of lymph nodes identified for histology preoperative chemotherapy. A central review of CT scans is
assessment being similar between groups. being undertaken to investigate whether the apparent
55% (11 of 20) of the tumours apparently downstaged upstaging was due to inaccurate initial radiological staging
between initial radiological assessment and pathological or true progression and will be reported separately. The
examination were also downstaged on radiology assess- absence of measurable adverse effects in terms of
ment after preoperative chemotherapy. None were prolonged hospital stay, stoma formation, reoperation
upstaged. Conversely, 35% (six of 17) of the tumours rates, and leak rates is, again, encouraging. A trend
apparently upstaged between initial radiological assess- towards increased wound and chest infection was expected
ment and pathological examination were also upstaged and seen, but was not significant. Compliance with
on postchemotherapy radiology. None were downstaged. postoperative chemotherapy, a potential surrogate marker
Reductions in the mean depth of spread beyond the for surgical morbidity, was not reduced in the preoperative
muscularis propria (12·8 [SD 8·4] to 9·0 [7·9] mm; group. Diarrhoea and skin rash were lower than previously
p=0·0022) and in the maximum tumour thickness reported with combination oxaliplatin, fluoropyrimidine,
(24·9 [12·2] to 19·0 [12·8] mm; p=0·0018) were also and EGFR antibody therapy,17 probably because only 31%
seen in the second radiological assessment compared received panitumumab—the panitumumab comparison
with baseline. is still masked—and also because of the short (6-week)
duration preoperative treatment.
Discussion Another key finding is that a high-risk cohort suitable
FOxTROT is, to our knowledge, the first randomised trial for combination chemotherapy can be identified by
to evaluate preoperative chemotherapy in primary colon radiological assessment across multiple sites. The risk of
cancer (panel). The pilot phase has provided clear evidence exposing patients with early stage (pT2) disease to
of downstaging with only 6 weeks of preoperative inappropriate chemotherapy was low, with only one
treatment. The significant reductions in apical node patient of 51 (2%) being inaccurately upstaged, justifying
involvement, incomplete resections, and two pathological the ongoing phase 3 trial. Chemotherapy for the 12% with
complete responses, are noteworthy since all have pT3 tumours without additional risk factors is not
previously been shown to correlate with protracted disease- unreasonable in view of the established efficacy of
free survival.11 Tumour regression grading after preoperative chemotherapy in stage 2 disease.34 In a parallel audit,35
chemotherapy also correlates with recurrence risk in other 93% of patients with radiologically staged T3 tumours
gastointestinal malignancies,33 and longer follow-up of the with less than 5 mm invasion of the muscularis propria
FOxTROT cohort should establish whether tumour were found to have coexisting high-risk pathological
regression grading is similarly prognostic in colon cancer. features that justify chemotherapy and the eligibility
We have also shown that preoperative therapy is criteria for the full FOxTROT study have been amended
practicable and safe with no increase in surgical morbidity accordingly. As radiological staging of primary colon
or mortality—a major focus of the feasibility study. With a cancer is underdeveloped, a series of workshops—
3-week delay after chemotherapy, surgery can be completed attended by 200 gastrointestinal radiologists—were held
within 10 weeks from diagnosis. A small proportion (17%) to standardise radiological selection criteria and similar
training would be advisable in other neoadjuvant trials.
Panel: Research in context The population studied in FOxTROT seemed typical of
those who might be considered for neoadjuvant therapy.
Systematic review The median age of 63 years is 10 years younger than the
We searched Medline and Embase for the MeSH and free terms: “colon” or “colonic“ near colon cancer population as a whole, but typical for trials
“cancer”, “carcinoma”, “tumor”, “tumour”, or “neoplasm” and drug therapy or chemotherapy of adjuvant chemotherapy. The distribution of tumours
and neoadjuvant or preoperative combined with appropriate study design filter as described in the study largely reflects that of colonic tumours in the
in section 6.4.11 of The Cochrane Handbook of Systematic Reviews of Interventions. No date or population, other than fewer tumours from the recto-
language restrictions were applied. No randomised trials comparing preoperative systemic sigmoid region, probably because of consideration for
chemotherapy with no chemotherapy in operable colon cancer were identified. radiotherapy, an exclusion criterion. The fairly high
Interpretation (20%) resection margin involvement in control patients
6 weeks of preoperative OxMdG chemotherapy for radiologically staged, locally advanced is probably explained by more rigorous histopathological
operable primary colon cancer is feasible with acceptable toxicity and perioperative assessment and by selection of high-risk patients.
morbidity and is associated with a significant pathological response. Further investigation Delivery of KRAS mutation analysis across 35 UK
of neoadjuvant chemotherapy of colon cancer is needed to evaluate long-term sites was another challenge, since delay in treatment
oncological outcome. would preclude clinical use. Our 9-day median time
from consent to KRAS testing to randomisation for