Study patient population and tumour rupture classification

Of the 400 patients randomised, we excluded three patients who were randomised without signing informed consent, 15 patients who did not have GIST at the central review of tumour histology, and 24 patients who had intra-abdominal metastases resected at surgery, which left 358 patients in the trial Efficacy Population, 181 in the 1-year arm and 177 in the 3-year arm (Fig. 1). The median duration of imatinib treatment in the 1-year and the 3-year groups was 12.0 months and 36.0 months, respectively. None of the patients received adjuvant imatinib longer than 37.2 months.

Fig. 1

CONSORT diagram of the study population.

Seventy-three (20.4%) of the 358 patients were reported to have GIST rupture to the trial database. In 12 (16.4%) of the 73 cases the medical records were not obtained for review. Forty-eight (78.7%) of the remaining 61 GISTs were classified as having undergone a major rupture, nine (14.8%) a minor rupture, and in four (6.6%) cases the tumour had ruptured into the bowel lumen without a spillage into the peritoneal cavity. These four GISTs were considered unruptured [8, 20] and were analysed together with 285 patients with non-ruptured GIST in the statistical analyses. Therefore, the final subgroup of patients with tumour rupture consisted of 69 patients, and the subgroup without a tumour rupture of 289 patients (Fig. 1). Most (n = 39, 56.5%) of the 69 ruptures occurred prior to surgery.

Patient and tumour characteristics

The median age of the 358 patients was 61 years (range, 22 to 84 years), and 184 (51.4%) were male. Ruptured GISTs were more frequently non-gastric compared with non-ruptured tumours, and they harboured more frequently KIT exon 9 mutations (Table 1). There was no statistical difference in the distributions of tumour size or mitotic counts between the rupture group and the non-rupture group.

Table 1 Characteristics of the patients with and without tumour rupture.

Thirty-one (44.9%) of the 69 patients with rupture were allocated to adjuvant imatinib for 1 year and 38 (55.1%) for 3 years. The characteristics of the 69 patients and their tumours by the random allocation group are provided in Supplementary Table S2. Eighteen (26.1%) of the 69 patients with rupture discontinued adjuvant imatinib before the scheduled duration was reached. Eight (44.4%) of these 18 patients stopped imatinib since GIST recurred while the patient was on imatinib (all in the 3-year group). One patient in the 1-year group stopped taking imatinib after 2.6 months, and in the 3-year group the median duration of imatinib administration was 21.0 months (range, 3.7–33.7 months) in the subset of nine patients who stopped imatinib for another reason than GIST progression.

Survival of patients with GIST rupture

Fifty-two (75.4%) RFS events occurred in the subset of 69 patients with rupture and 119 (41.2%) among the 289 patients with no rupture during a median follow-up time of 10.0 years, and 27 (39.1%) and 53 (18.3%) patients died, respectively. Patients with rupture had inferior RFS and OS compared with patients with non-ruptured GIST (HR 2.34, 95% CI, 1.69–3.26; p < 0.001; and HR 2.36, 95% CI, 1.48–3.76; p < 0.001, respectively; Fig. 2). Of the patients with rupture, 20.8% were alive without recurrence 10 years after the date of randomisation, and 58.8% were alive. The RFS and OS of the patients with minor rupture were also inferior compared to patients with no rupture (HR 2.24, 95% CI, 1.04–4.81; p = 0.039, and HR 2.82, 95% CI, 1.02–7.81; p = 0.037, respectively). The RFS and OS of the 12 patients whose rupture could not be classified resembled those of the patients with minor rupture (supplementary Fig. S1). In a multivariable analysis that contained tumour rupture (rupture vs. no rupture), tumour site (non-gastric vs. gastric), mitotic count (continuous covariable), size (continuous covariable) and the treatment group (1-year vs. 3-years) as covariables, presence of GIST rupture was independently associated with unfavourable RFS (HR 2.34, 95% CI 1.66–3.28; p < 0.001) together with non-gastric tumour site (HR 2.91, 95% CI 2.10–4.07; p < 0.001), large GIST size (HR 1.05, 95% CI 1.03–1.08; p < 0.001), a high mitotic count (HR 1.03, 95% CI 1.02–1.03; p < 0.001), and the 1-year treatment group (HR 1.61, 95% CI 1.17–2.20; p = 0.003). GIST rupture was independently associated also with unfavourable OS (HR 2.50, 95% CI 1.53–4.08; p < 0.001) as were the mitotic count (HR 1.02, 95% CI 1.01–1.03; p < 0.001), the 1-year treatment group (HR 1.98, 95% CI 1.22–3.20; p = 0.006), and a non-gastric tumour site (HR 1.68, 95% CI 1.05–2.70; p = 0.032), whereas tumour size was not (p = 0.301). When age at study entry was added as the sixth covariable to this analysis, GIST rupture was still independently associated with OS (Supplementary Table S3).

Fig. 2: Survival outcomes of patients with and without tumour rupture.

Upper panel: recurrence-free survival; lower panel: overall survival. Five-year and 10-year survival rates are shown. Patients alive are indicated with a bar.

Influence of adjuvant imatinib on survival

There was no significant difference in RFS or OS in the subset of patients with ruptured GIST when the patients assigned to 3-year adjuvant imatinib were compared to those assigned to 1-year of imatinib (Fig. 3). The results remained essentially similar when these analyses were restricted to patients with major rupture (Supplementary Fig. S2).

Fig. 3: Influence of the duration of adjuvant imatinib on survival outcomes of patients with tumour rupture.

Upper panel: recurrence-free survival; lower panel: overall survival. Five-year and 10-year survival rates are shown. Patients alive are indicated with a bar.

KIT exon 11 deletion/indel mutations and survival

We next investigated the survival of patients with ruptured GIST in the largest mutational subgroup, patients with KIT exon 11 deletion/indel mutation. Thirty-one (46.3%) of the 67 patients with rupture and with mutation analysis results available had KIT exon 11 deletion/indel mutation. Patients allocated to 3 years of imatinib tended to have longer RFS compared to those allocated to 1 year of imatinib in this mutational subgroup (HR 0.45, 95% CI, 0.19–1.04; p = 0.056). A large drop in RFS occurred in the Kaplan-Meier plots once imatinib was stopped (Fig. 4). The patients allocated to 3-year adjuvant imatinib had longer OS than those allocated to 1 year of adjuvant imatinib when GIST harboured KIT exon 11 deletion/indel mutation (HR 0.09, 95% CI, 0.01–0.74; p = 0.016). Only 1 of the 17 patients assigned to the 3-year adjuvant imatinib group died during the follow-up despite a ruptured tumour, which resulted in a high 10-year OS rate of 94.1% (90.0% when the analysis was restricted to patients with a major rupture; supplementary Fig. S3).

Fig. 4: Influence of the duration of adjuvant imatinib on survival outcomes of patients with KIT exon 11 deletion/indel mutation and with tumour rupture.

Upper panel: recurrence-free survival; lower panel: overall survival. Five-year and 10-year survival rates are shown. Patients alive are indicated with a bar.

The numbers of patients with some other type of KIT or PDGFRA mutation or no mutation in these genes were too small for carrying out reliable survival analyses. All 13 ruptured GISTs with KIT exon 9 mutation recurred.

Treatment after GIST recurrence

The treatments for recurred GIST were administered after the trial primary endpoint had been met and outside of the trial protocol. Despite the treatments for recurred or overtly metastatic GIST were selected based on the institutional practice, we collected limited data about them when feasible, since they likely influence OS. Information about the first-line systemic treatment after GIST recurrence was available from 47 (68.1%) of the 69 patients with rupture. Thirty-four (72.3%) of the 47 patients received imatinib as the first-line treatment for advanced GIST, four (8.5%) sunitinib, three (6.4%) nilotinib, and six (12.8%) patients received no systemic treatment. Therefore, a total of 41 (87.2%) of the 47 patients with information available received tyrosine kinase inhibitor treatment after GIST recurrence as their first-line treatment for recurrent GIST.