Caris Molecular Intelligence for guiding cancer treatment

Dean et al. (2016)

Study size, design and location

54 patients with heavily pre-treated advanced cancer or with rare tumours in a prospective, single-centre study. Australia.

Intervention and comparator(s)

CMI‑guided therapy compared with clinicians' choice alone.

Key outcomes

CMI‑guided therapy recommendations differed from clinician recommendations in 89% of the study population. CMI‑guided therapy showed clinical benefit (defined as improved quality of life or performance status, symptoms, body weight or response rate based on imaging using RECIST) in 61% of heavily pre-treated tumours, 69% of rare tumours, and 43% patients who had previous standard first-line therapy. 60% of evaluable patients had a PFS ratio of 1.3 or higher.

Strengths and limitations

The study did not use randomisation and each person acted as their own control, so the results could not be compared with the clinician's initial choice of next best therapy. The study included a broad range of tumour types, which limits conclusions for different cancers.

Herzog et al. (2016)

Study size, design and location

224 patients with advanced stage recurrent epithelial ovarian cancer in a retrospective, registry-based observational multicentre study. Location unclear.

Intervention and comparator(s)

CMI; 2 cohorts were compared based on the matching of treatment to CMI recommendations (n=121, matched cohort) or patients who had at least 1 treatment associated with potential lack of benefit based on CMI (n=103, unmatched cohort).

Key outcomes

The matched cohort had a significantly greater improvement in OS from the time of molecular profiling (median 36 months) compared with the unmatched cohort (median 27 months). Patients who had more than 1 drug in the lack-of-benefit category trended towards worse OS than those who had only 1 drug in this category.

Strengths and limitations

The study did not use randomisation, but it did report that bias was accounted for (statistical analysis not specified) in age, race, stage, histology, grade and site of biopsy. Potential for bias remains in selection bias because of clinicians choosing to profile some patients over others, and choosing whether or not to follow biomarker recommendations based on unrecorded patient characteristics.

Jameson et al. (2014)

Study size, design and location

28 patients with previously treated metastatic breast cancer in a prospective, multicentre pilot study. US.

Intervention and comparator(s)

CMI‑guided therapy compared with clinicians' choice alone.

Key outcomes

Of the 28 patients enrolled, 25 were evaluable. MMP was done on fresh core biopsies and the results were sent to a committee which selected CMI‑guided treatment for all. None of the CMI‑guided therapies were the same as those the clinician would have chosen.

11 patients (44%) had a PFS ratio of 1.3 or higher after treatment and median survival was 10 months. Partial responses were noted in 5 patients, stable disease in 8 and no progression at 4 months in 9.

12 patients had a PFS ratio of less than 1.3 and median survival was 4 months; 2 patients had a scan outside of the required time frame and therefore were not considered responders.

Strengths and limitations

The sample size was powered to identify PFS ratio using a type I error rate of 5% and a power of 90%. No investigational agents were included as therapy in this study.

Spetzler et al. (2015)

Study size, design and location

1,180 patients with a variety of solid tumours in a prospective study. US.

Intervention and comparator(s)

CMI‑guided therapy compared with clinicians' choice alone. Cohort 1 (n=510) consisted of patients who had 1 or more drugs predicted to be of benefit and no drugs predicted to lack benefit. Those that had at least 1 drug predicted to lack benefit were placed in cohort 2 (n=500).

Key outcomes

Survival analysis of cohort 1 versus cohort 2 showed a median increase in OS of 274 days (978 versus 704 days). Clinicians indicated (by a self-report questionnaire) that CMI influenced their decision in 58% of cases. Of these, 97% had a drug from the benefit category and 46% did not have any lack-of-benefit category drugs.

Strengths and limitations

This was a published abstract, so detailed methodology is limited. It is unclear how the study determined whether CMI influenced clinician decisions.

Von Hoff et al. (2010)

Study size, design and location

86 patients with refractory metastatic cancer (mainly breast, colorectal and ovarian; 32 had other, rarer cancers) in a prospective, pilot cohort, multicentre study. US.

Intervention and comparator(s)

CMI‑guided therapy compared with clinicians' choice alone.

Key outcomes

In 84 patients with a molecular target detected, 66 (78%) were treated according to CMI results. Of these, 18 (27%) had a PFS ratio of 1.3 or higher and a median survival of approximately 10 months.

Strengths and limitations

The study did not use randomisation and each person acted as their own control. Ascertainment bias may exist with the PFS ratio influenced by frequency of tumour evaluation in the control and evaluation periods. The study was limited by attrition (38 of the original 106 patients who consented were lost to follow‑up; 20 withdrew before CMI, mainly because of not wanting additional therapy or worsening condition, and 18 withdrew after having CMI, mainly because of worsening condition and withdrawing consent).

Abbreviations used: CMI, Caris Molecular Intelligence; MMP, multi-omic molecular profiling; OS, overall survival; PFS, progression-free survival; RECIST, Response Evaluation Criteria in Solid Tumors.