A Randomised, Double Blind, Placebo-Controlled Pilot Study of Oral Artesunate Therapy for Colorectal Cancer

Abstract

Background: Artesunate is an antimalarial agent with broad anti-cancer activity in in vitro and animal experiments and case reports. Artesunate has not been studied in rigorous clinical trials for anticancer effects.

Aim: To determine the anticancer effect and tolerability of oral artesunate in colorectal cancer (CRC).

Methods: This was a single centre, randomised, double-blind, placebo-controlled trial. Patients planned for curative resection of biopsy confirmed single primary site CRC were randomised (n = 23) by computer-generated code supplied in opaque envelopes to receive preoperatively either 14 daily doses of oral artesunate (200 mg; n = 12) or placebo (n = 11). The primary outcome measure was the proportion of tumour cells undergoing apoptosis (significant if > 7% showed Tunel staining). Secondary immunohistochemical outcomes assessed these tumour markers: VEGF, EGFR, c-MYC, CD31, Ki67 and p53, and clinical responses.

Findings: 20 patients (artesunate = 9, placebo = 11) completed the trial per protocol. Randomization groups were comparable clinically and for tumour characteristics. Apoptosis in > 7% of cells was seen in 67% and 55% of patients in artesunate and placebo groups, respectively. Using Bayesian analysis, the probabilities of an artesunate treatment effect reducing Ki67 and increasing CD31 expression were 0.89 and 0.79, respectively. During a median follow up of 42 months 1 patient in the artesunate and 6 patients in the placebo group developed recurrent CRC.

Interpretation: Artesunate has anti-proliferative properties in CRC and is generally well tolerated.

Keywords: Artesunate; Colorectal cancer; Dihydroartemisinin; Ki67; Neutropaenia.

Figures

Fig. 1

Patient flow diagram. Assessment for eligibility was recorded after 6 patients had been randomised, so that these added to the 17 patients randomised after screening give the total number of randomised patients (n = 23).

Fig. 2

2a Ki67 staining in treatment groups. Individual results for Ki67 epithelial cell staining (% positive), the grand means in each treatment group as predicted by the random effects model and individual predications are displayed. Interquartile intervals are presented for the raw measurements and the 95% credible intervals correspond to the individual and group predicted means. This analysis has been carried out under missing at random assumption (MAR) with large uncertainty around the individual means of missing individuals, as expected and shown for 3 individuals randomised to receive artesunate but not able to be analysed for reasons given in results. These results correspond to non-informative prior assumption with regards to the difference between the two groups. 2b Sensitivity analysis to various prior information on the difference between the two groups with respect to Ki67 — the analysis has been justified by published experimental results. The probability that the difference between artesunate and placebo is negative remains high even under a skeptical prior of no effect (0.77).

Fig. 3

Immunohistochemical staining of biomarkers in colorectal cancer. (a) Detection of apoptotic cells by the Tunel assay, (b) Ki67, (c) p53, (d) EGFR, (e) c-MYC, (f) CD31, (g) VEGF, (h) negative control (without primary antibody). Magnification: × 250. For determination of protein expression the UltraVision polymer detection method (kit from Thermo Fisher Scientific GmbH, Dreieich, Germany) was used as detailed in Supplementary methods. The immunostained slides were scanned by Panoramic Desk (3D Histotech Pannoramic digital slide scanner, Budapest, Hungary) and interpreted (Quantification of immunostained slides) by panoramic viewer software (NuclearQuant and membraneQuant, 3DHISTECH) in which positive stained nucleus or membrane were counted in each defined annotated area. Evaluation parameters included number of overall detected objects (nucleus or membrane) in each annotated area, average of positivity and intensity. Nuclear stainings (Ki67, p53, c-MYC, TUNEL) were quantified using the Nuclear Qant software and Membrane-bound and cytosolic stainings were quantified by the MembraneQuant software (3D histoQuant). Results are in Table 2.

Fig. 4

Survival recurrence curves predicted by Cox proportional hazards model. Patient CRC21 was assumed to be missing completely at random (please see Supplementary Table 1 for a full sensitivity analysis). In the placebo group 2 patients died within a year (108, 170) days leaving 10 (83%) in the study, another 2 within the next year (383, 663 days) leaving 8 (66%) in the study and the other two died within the third year of the follow up (749 and 990, respectively) leaving 50% patients beyond the third year. The only death in the artesunate group happened after 552 days leaving 9 patients (90%) surviving beyond the third year. These crude estimates support the estimates from the data above.

Fig. 5

Adverse events. a. Patient CRC 04's haemoglobin (filled red circles, g/dl), total white cell count (filled blue squares, × 10− 9/L), neutrophil count (green triangles, × 10− 9/L) and platelet count (grey diamonds, x) are shown from the start of the study (Day 0). b. CRC 07's serum carcinoembryonic antigen levels are shown from the start of the study (Day 0). c. CRC 07's haematological results are shown from the start of the study (Day 0) with symbols as in a. Tx is transfusion of red cells.