Results from FOCUS4-D

The full report of the FOCUS4-D trial has been published.88 Parts of this chapter have been reproduced from Adams et al.88 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Scientific rationale for FOCUS4-D

Signalling through HER receptors (e.g. EGFR, HER2, HER3 and HER4) and their downstream pathways is a key mechanism that promotes proliferation and the malignant phenotype in cancer. Epidermal growth factor has been recognised as a key pathogenic surface receptor in CRC for many years. A greater understanding of CRC biology followed by use of EGFR-targeted treatments has led to the discovery of the importance of BRAF, PIK3CA, KRAS and NRAS mutations in predicting a lack of response to EGFR-targeted therapy.3 The monoclonal antibodies cetuximab and panitumumab were developed to target EGFR on the surface of cancer cells. After licensing, cetuximab and panitumumab were reported to be ineffective in patients whose cancer expressed a somatic mutation in the NRAS or KRAS genes.14 Simple EGFR inhibition, therefore, has limitations because both de novo and acquired resistance can arise and result in a lack of benefit for most of these biomarker-selected patients, thus driving interest in novel approaches to inhibit this pathway. However, EGFR remains a valid target for CRC within the population of patients who are wild type for KRAS, NRAS and BRAF, with proven clinical benefit.

The proteins HER2 and HER3 (also known as ERBB2 and ERBB3, respectively) heterodimerise with EGFR and are mechanisms of resistance to EGFR inhibition. Human epidermal growth factor-3 is a membrane-bound receptor protein that has extracellular heregulin- and neuregulin-binding domains but does not have an intracellular kinase domain, relying on heterodimerisation to other family members for downstream effects. Human epidermal growth factor-3 expression is associated with poor prognosis in CRC.89 The protein has a central role in driving oncogenic signals in tumours,90 and preclinical and clinical data have led to the hypothesis that HER3 is an escape pathway to EGFR blockade through a compensatory shift to HER3 signalling, predominantly through the PI3K/AKT pathway.91–93 Moreover, clinical data indicate that HER3 overexpression predicts the lack of efficacy of panitumumab.94 Human epidermal growth factor-2 is another member of the EGFR family of membrane-bound receptors. It is an orphan receptor, having no known associated ligands, but is activated through homodimerism and heterodimerisation with other EGFR family members, resulting in intracellular phosphorylation and cascaded downstream signalling. Growing in vitro and in vivo evidence suggests that HER2 might be overexpressed more frequently in patients with EGFR-dependent (all wild type) tumours.95 Furthermore, this protein might be upregulated in acquired EGFR inhibitor resistance, and concomitant blockade of HER2 could increase the efficacy and prolong activity in a synergistic fashion. Bertotti et al.96 developed a range of CRC xenograft models from genetically well-characterised, metastatic CRC samples. A cohort of these so-called xenopatients showed amplification of HER2, particularly in RAS wild-type tumours, suggesting enrichment in this cohort, which showed high and sustained sensitivity to combination EGFR and HER2 inhibition. Inhibition of EGFR, HER2 and HER3 signalling is postulated to reduce de novo resistance, thereby increasing the proportion of patients showing benefit when compared with inhibition of EGFR only. Furthermore, this inhibition might slow the development of acquired resistance in patients with initially EGFR-dependent tumours, thereby providing greater clinical benefit than EGFR inhibition alone.

FOCUS4-D was designed to assess the efficacy of AZD8931 in patients with CRC whose tumours were wild type for BRAF, PIK3CA, KRAS and NRAS mutations, and reached its first interim analysis trigger point in March 2016.

AZD8931 is an orally active, equipotent tyrosine kinase inhibitor of EGFR, HER2 and HER3 signalling. In vitro analysis of AZD8931 in ligand-driven cell assays shows potency exceeding that of gefitinib (Iressa, AstraZeneca) and lapatinib (Tykerb, Novartis) in this system for EGFR, HER2 and HER3 inhibition.97 In pharmacokinetic and dynamic analyses, a direct relation was recorded between the total concentration of AZD8931 in plasma and the inhibition of EGFR phosphorylation.97 Patients with PTEN loss or somatic BRAF, PIK3CA, KRAS and NRAS mutations were excluded from FOCUS4-D. The rationale for the exclusion of patients with BRAF mutations relates to two specific sets of data. First, BRAF-mutant CRC has significantly worse prognosis than disease with wild-type KRAS and NRAS.98 Second, despite some conflicting evidence, BRAF-mutant tumours gain no additional benefit from targeted EGFR inhibition.99 Patients harbouring tumours with a PIK3CA mutation or PTEN loss were excluded specifically because data suggest that the PI3K/AKT pathway is a resistance mechanism to EGFR inhibition.100

In the Phase II MRC COIN-B trial,9 patients who had metastatic KRAS wild-type tumours were treated with an intermittent strategy of oxaliplatin, intravenous fluorouracil and folinic acid and were randomised to either intermittent cetuximab (with intermittent chemotherapy) or continuous cetuximab (including single-agent maintenance through the interval between chemotherapy). These trial data showed that the use of maintenance cetuximab in the interval between chemotherapy was associated with an improvement in the duration of the progression-free interval from 3 months to 6 months.9 FOCUS4-D builds on these data to test if, in a more specific molecular cohort, AZD8931 could improve PFS in the interval off chemotherapy.

Study methods

Patients’ registration and biomarker assessment

The FOCUS4-D trial was undertaken at 18 hospital sites in the UK. Patients aged ≥ 18 years with newly diagnosed locally advanced or metastatic CRC were eligible for registration in the FOCUS4 trial programme, either at the start of a 16-week regimen of first-line chemotherapy or up to 12 weeks into the regimen. Patients could not be registered into FOCUS4 unless a CT scan had been carried out within 4 weeks of starting chemotherapy (pre-registration CT scan). CT scans were obtained 8 weeks and 16 weeks after chemotherapy was started. A CT scan performed at the end of the 16-week regimen was compared with the preregistration CT scan by local radiologists, using RECIST, version 1.1. Patients whose tumours had progressed were ineligible for randomisation and any further involvement with FOCUS4 ended at this time.

The 16-week duration of first-line chemotherapy was determined by reviewing data from three key trials: COIN,16 CAIRO3101 and AIO0207.102 This period accounted for factors including maximum tumour response and tolerability by patients.

During the 16-week period of first-line chemotherapy, a sample of the patient’s tumour was sent to one of two dedicated FOCUS4 biomarker laboratories in Leeds and Cardiff (UK), respectively, for assessment and stratification of the patient into one of four molecular groups. The molecular stratification hierarchy for the FOCUS4 trial programme is presented in Chapter 2 (see Figure 1). The technical components of the biomarkers and interlaboratory quality assurance are also described in Chapter 2.17

Patients whose tumours had remained stable or responded to treatment, as shown by the CT scan, were assessed for eligibility for FOCUS4-D. The biomarker assessment had to show wild-type status for BRAF, PIK3CA, KRAS and NRAS, as well as PTEN expression on immunohistochemistry. The FOCUS4-D trial design with eligibility criteria is presented in Figure 10. Full inclusion and exclusion criteria are available from the trial website (www.focus4trial.org; accessed 13 May 2022).

FIGURE 10 FOCUS4-D trial schema. a, Criteria were aged older than 17 years, no brain metastases, adequate organ function, WHO performance status of 0–2, not pregnant and CT scan within 4 weeks before randomisation. PTEN, phosphatase and tensin homologue; RECIST, Response Evaluation Criteria in Solid Tumours. (Figure not reproduced here; see the original article.)

In addition to the eligibility criteria, patients had to have had a minimum of a 3-week gap after the last dose of chemotherapy or biological therapy and before the first dose of trial drug; a WHO performance status of 0–2; and adequate organ function, which was ascertained by an estimated creatinine clearance > 50 ml/minute (according to local estimation method), serum bilirubin less than 1.5 times the upper limit of normal (ULN), alanine aminotransferase, aspartate aminotransferase and alkaline phosphatase less than 2.5 × ULN in the absence of liver metastases and less than 3.0 × ULN in presence of liver metastases, and a left ventricular ejection fraction greater than 50% by multigated acquisition scan or echocardiography. Ethics approval was granted by the National Research Ethics Service South Central Oxford—Panel C ethics committee. Regulatory approval was granted by the UK MHRA. All trial procedures and processes complied with the International Conference On Harmonisation’s Good Clinical Practice guidelines.103 Patients were asked to sign a consent form for registration and for analysis of the biomarker panel; additional written informed consent was obtained before randomisation into FOCUS4-D.

Randomisation and masking

Randomisation was undertaken by telephone to a centrally managed service at the MRC CTU at UCL. Patients were allocated to either AZD8931 or placebo by minimisation with a random element of 20% (both patients and clinicians were masked to treatment allocation). Minimisation factors were treating hospital site, site of primary tumour (right colon, left colon or rectum), WHO performance status (0–2), 16-week CT scan result (stable disease, partial response or complete response), number of metastatic sites (none, one, or two or more) and first-line chemotherapy regimen (fluorouracil, capecitabine or neither; both oxaliplatin and irinotecan, oxaliplatin only, irinotecan only, or neither; and cetuximab or panitumumab, bevacizumab, or no monoclonal antibody). AZD8931 and placebo were identical in appearance and were supplied by AstraZeneca Ltd. The packaging, labelling and distribution of AZD8931 and placebo were all undertaken by Fisher Services. Treatment was administered double-blind using an interactive web-based drug delivery system provided by Cenduit (Nottingham, UK).

Procedures

Patients were asked to take oral AZD8931 (40 mg twice daily) or placebo until disease progression, death or toxic effects. Patients were also requested to complete a diary card on which the number of pills taken every day was recorded. Patients were assessed every 8 weeks by CT scan, which was reviewed at the treating hospital site according to RECIST, version 1.1. Measurements were collated centrally to ensure appropriate delineation of response and progression. Toxic effects and symptoms were assessed locally every 4 weeks from the start of trial treatment, using the NCI Common Terminology Criteria for Adverse Events (version 3.0).104 Patients were followed up until progressive disease was identified on a CT scan, at which point the patient was recommended to restart first-line chemotherapy. Treatment was stopped in the event of grade 3 or worse toxic effects or persistent toxicities judged medically significant or not tolerated by the patient, until the toxicity resolved to grade 1 or better. After stopping treatment (for a maximum of 28 days), trial therapy could be re-initiated at a reduced dose. For the first event, treatment was restarted at 40 mg in the morning and 20 mg at night; for the second event, treatment was restarted at 20 mg twice per day. A third event resulted in treatment discontinuation. Any stoppage for ≥ 28 days was not permitted and the trial therapy was discontinued from.

Outcomes

The primary outcome of FOCUS4-D was PFS, defined as the time from randomisation to either disease progression (according to RECIST criteria) or death from any cause. The trial could be extended to include OS as a secondary outcome if adequate drug activity was seen for PFS in the early interim analyses. Additional secondary end points included safety and toxicity.

Statistical methods

FOCUS4 used a MAMS design, which allows pre-planned analyses to be carried out to inform a decision on whether or not to continue the trial.105 The nstage MAMS function in Stata, version 14.1, was used to calculate the operating characteristics for each molecular trial.106 For FOCUS4-D, a recruitment rate of nine patients per month was anticipated if 100 sites were open. For the placebo group, a median PFS of 4.6 months was assumed, based on a similar group of patients being followed up in the COIN trial.16 A target HR of 0.5 was sought for PFS, with a randomisation ratio of 1 : 1. A summary of the operating characteristics for FOCUS4-D, with predicted timelines for the staged interim analyses, is presented in Table 9. Power was maintained at 85% after correction for two interim analyses and one final analysis. The target recruitment was 174 patients, with the first interim analysis timed to occur when nine PFS events had occurred in the control group. At every interim analysis, the observed HR was compared with the critical HR for that interim analysis (see Table 9). If the HR was above the critical value, trial closure could be considered on the grounds of an absence of sufficient drug activity. When an interim analysis was triggered by the prespecified number of PFS events in the control group, data were analysed and the results presented at a closed and confidential meeting of the IDMC, who were not involved with the conduct of the study. Recommendations to continue or close the study were made to the TSC, who would recommend their decision. If a decision was made to close the study, the TSC would allow the TMG to see the data to ensure that there were no objections to study closure. When all committees agreed to trial closure, study sites, the pharmaceutical company (AstraZeneca Ltd) and the trial funders were to be informed, and no further patients were to be permitted entry into the study.

All analyses were undertaken according to a predefined SAP, which was agreed before any data inspection. The primary outcome was prespecified to be analysed by ITT (final analysis). An interim analysis was undertaken in the per-protocol sample, which was defined as patients who completed at least one cycle of trial treatment (≥ 28 days). Per-protocol analyses were carried out for sensitivity. Data were analysed with Stata, version 14.1.

Patients were censored according to the following criteria. For survival status, patients were censored on the date that they were last known to be alive, which was recorded as the latest of either the date of collection of a prescription from their hospital pharmacy or the date of a follow-up visit or CT scan. For patients who died before any follow-up visit or CT scan, the date of death was used as the date of the event and assumed death without progression. For PFS, patients were censored without progression on the date of the last CT scan showing no progression.

Kaplan–Meier curves were used to present survival data and Cox regression modelling to estimate HRs, which were adjusted for the stratification factors that were used to minimise patients into allocated groups. Because the numbers of patients and events were small, adjustments were undertaken using the method of inverse probability weighting107 and using the bootstrap method to estimate CIs. The proportional hazards assumption was tested by regressing scaled Schoenfeld residuals against the log of time.108 If evidence showed significant violation, a sensitivity analysis using restricted mean survival analysis was performed.105

Study results

Between 7 July 2014 and 7 March 2016, 32 patients were randomised (16 into each treatment group) across 18 hospitals in the UK (Figure 11). The groups were well balanced in terms of baseline characteristics (Table 10). The first interim analysis was triggered in March 2016, when nine PFS events had occurred in the placebo group. The IDMC reviewed the data as part of a closed confidential meeting and made a recommendation to close the trial based on insufficient drug activity; the observed HR did not fall below the critical HR threshold of 1.0, which was predefined as part of the multistage sample size calculations (see Table 9). This decision was subsequently endorsed by both the TSC and the TMG. Trial recruitment was closed, and during subsequent follow-up of patients already entered into the study further PFS events occurred, such that 31 out of 32 patients had a PFS event by the time of the final analysis on 1 August 2016. No patients were lost to follow-up. The median PFS was 3.48 months (95% CI 1.51 to 5.09 months) with placebo and 2.96 months (95% CI 1.94 to 5.62 months) with AZD8931 (adjusted HR 1.10, 95% CI 0.47 to 3.57; p = 0.95) (Figure 12).

FIGURE 11 FOCUS4-D trial profile. a, Reasons for non-randomisation not recorded. PTEN, phosphatase and tensin homologue. CR, complete response; PR, partial response. (Figure not reproduced here; see the original article.)

FIGURE 12 Progression-free survival (ITT analysis) FOCUS4-D. (Figure not reproduced here; see the original article.)

Two patients did not receive any treatment (one patient in each group) and one further patient in the placebo group did not complete their first cycle of treatment. Therefore, 29 patients were included in the per-protocol sample (14 in the placebo group and 15 in the AZD8931 group). The Cox regression analysis within the per-protocol sample (for sensitivity) generated an adjusted HR of 1.15 (95% CI 0.49 to 4.75; p = 0.92) in favour of placebo. No statistical evidence was found that the proportional hazards assumption had been violated (Grambsch–Therneau test on log-time scale: p = 0.25 for the ITT analysis, and p = 0.35 for the per-protocol analysis); however, as a result of the small sample size, a sensitivity analysis using restricted mean survival time analysis at 4 months was carried out. However, this test did not alter the results for either the ITT analysis [adjusted restricted mean 2.90 months (95% CI 2.34 to 3.46 months) in the placebo group vs. 2.95 months (95% CI 2.37 to 3.53 months) in the AZD8931 group; p = 0.90 for difference] or the per-protocol analysis [adjusted restricted mean 3.08 months (95% CI 2.61 to 3.56 months) vs. 3.05 months (95% CI 2.48 to 3.63 months); p = 0.94 for difference].

Of the 32 randomised patients, 11 (34%) received first-line irinotecan and fluorouracil. After disease progression, only four (36%) of these patients returned to their first-line therapy. Reasons for not returning to their first-line therapy were the clinician’s decision because of toxicity or disease progression. A further eight (25%) patients received first-line oxaliplatin plus capecitabine, and four (50%) of these patients returned to this regimen after disease progression. Reasons for not returning to first-line therapy were the clinician’s decision because of toxicity (n = 1), disease progression (n = 1) and no symptoms being reported (n = 1) and the patient entering another trial (n = 1). A further seven (22%) patients received irinotecan and fluorouracil plus cetuximab or panitumumab as first-line therapy, and only one (14%) patient returned to their first-line regimen after disease progression. The reasons for not returning to first-line therapy were patient’s choice (n = 4) and the patient was considered for resection (n = 1), and there was also missing data on the reasons provided (n = 1). Other first-line therapies included oxaliplatin and capecitabine plus bevacizumab (n = 2), oxaliplatin and fluorouracil (n = 2), capecitabine alone (n = 1) and fluorouracil alone (n = 1). In total, 11 out of 32 (34%) patients restarted their first-line therapy at the time of data collection.

No patterns were noted in the reasons given for not restarting therapy. As at November 2017, not all data for post-progression treatments were available. Five of the 32 randomised patients needed dose reductions: one in the placebo group and four in the AZD8931 group. The patient in the placebo group had their evening dose reduced from 40 mg to 20 mg on their second cycle because of diarrhoea (grade 3). The dose reductions in the AZD8931 group were all evening dose reductions from 40 mg to 20 mg on the second or third cycles and were because of skin rash (grade 2, n = 1; grade 3, n = 3).

One patient in the AZD8931 group subsequently had a morning dose reduction from 40 mg to 20 mg on their fourth cycle, which was also because of skin rash (grade 3), and another patient subsequently had a morning dose reduction from 40 mg to 20 mg on their sixth cycle, which was for dry eyes (grade 3). All dose reductions were maintained within the currently available data. Only one patient explicitly gave the reason for stopping trial treatment as toxicity. This patient was assigned AZD8931 and reported grade 3 dry eyes, grade 2 skin ulceration and grade 2 skin rash. No serious adverse events were reported for this patient. Overall, few toxic effects were reported. Skin rash was the most frequent grade 3 adverse event, recorded in 3 out of 15 patients (20%) in the AZD8931 group compared with no patients in the placebo group. Diarrhoea was the most frequent grade 3 adverse event in the placebo group, recorded in one patient (6%) compared with one patient (7%) in the AZD8931 group (Table 11). No grade 4 adverse events were reported. Five serious adverse events required admission to hospital: four in the AZD8931 group [epigastric pain (grade 3); back pain (grade 3), hyperbilirubinaemia (grade 4) and device-related infection (grade 2); duodenal ulcer (grade 3); and dehydration (grade 3, but subsequently fatal)] and one in the placebo group [chest infection (grade 3)]. No treatment-related deaths were reported. Patients’ compliance was 85% in the placebo group and 75% in the AZD8931 group.

Discussion and conclusions

In FOCUS4-D, AZD8931 failed to pass the first stage of assessment within the MAMS trial design, that is the predefined critical HR of 1.0 at stage 1 (lack of sufficient activity) was not reached. However, this trial has set a new paradigm in molecularly stratified trials in CRC.

Two Phase II, randomised, combination trials of AZD8931 have been carried out previously, both in patients with breast cancer. The MINT study109 was stopped because of a lack of efficacy (NCT01151215). In this trial, AZD8931 (40 mg or 20 mg twice daily) was compared with placebo in combination with anastrozole in postmenopausal women with hormone receptor-positive locally advanced or metastatic breast cancer who had never received endocrine treatment. The second randomised trial (THYME)110 did not meet its primary objective of prolonged PFS when AZD8931 was added to weekly paclitaxel in patients with advanced breast cancer and low expression of HER2 (NCT00900627). The primary hypothesis in the THYME study was that low HER2 expression was a driver for heterodimerisation, which would be inhibited by AZD8931.

In FOCUS4-D, it was not possible to investigate subgroups effectively because of the small sample size. Specific areas of interest include the effect of anti-EGFR agents in the chemotherapy induction phase and pre randomisation (six patients in total) and the role of PTL as a marker of response to treatment. The findings of these three trials indicate a need for greater understanding of EGFR family heterodimers and their dynamic interaction with targeted agents. To date, heterodimers have been poorly assessed in vitro and in vivo because of a paucity of effective methods, which might have hampered more complete scrutiny of therapeutic agents, such as FOCUS4-D, which was the first of the FOCUS4 molecular cohorts to reach its first MAMS-defined interim analysis. The observed HR at the interim analysis was above the critical threshold for continuation of the trial, and the decision to close the trial was clear cut and unanimous across all oversight committees. This outcome was considered to be a success of the trial design, in that a decision to continue might have wasted clinical trial resources on recruiting and following up a further 142 patients, and closing the trial prevented other patients from being given an ineffective but potentially harmful treatment. FOCUS4-D was designed to detect a target HR of 0.5. Therefore, the probability of a final analysis of 174 patients showing a HR of 0.5 when the noted HR at this interim analysis was 1.10 was exceptionally small.

A limitation of the FOCUS4-D study was that there was some deviation from the assumptions that we made in the sample size calculations. The final sample size was small, with recruitment of only 32 of the maximum expected target of 174 patients. The median PFS in the placebo group was 3.5 months, which is lower than that anticipated in the sample size (4.6 months), and the higher event rate brought forward the trigger point for analysis (nine events in the placebo group). Owing to delays in site opening of this complex trial, recruitment was also slower than anticipated; therefore, this trigger occurred after randomisation of fewer patients than had been expected (32 instead of 54 patients). However, since all but one patient had a PFS event by the time of the final analysis, it was reassuring that further follow-up would not alter the conclusions of the study, and there is little reason to suspect that a further 22 patients (which would take the total sample size to the expected 54 patients at this analysis) would have responded any differently to the 32 patients reported here. In addition, the smaller number of patients in the stage 1 analysis of lack of sufficient activity might increase any effect on outcome between the two treatment groups posed by trial imbalances in prognostic factors and choice of first-line treatment. For instance, the trial size makes it impossible to assess with any meaningful interpretation the effects of previous first-line treatment by EGFR inhibition with cetuximab or panitumumab (four patients in the placebo group and three patients in the AZD8931 group).

FOCUS4-D has shown no evidence of efficacy of single-agent EGFR, HER and HER3 inhibition with AZD8931 in patients with advanced CRC whose tumours are wild type for BRAF, PIK3CA, KRAS and NRAS after first-line induction therapy. Toxic effects were low, apart from skin rash in about 20% of patients. Early planned interim analyses with prespecified efficacy thresholds for lack of sufficient drug activity might be helpful in closing trials that are very unlikely to show benefit, which could facilitate more efficient use of clinical trial resources. New agents are currently being investigated for testing in this cohort of patients with all wild-type tumours.