Patients with gestational trophoblastic neoplasia who have an International Federation of Gynaecology and Obstetrics (FIGO) risk score of 5 or 6 usually receive non-toxic single-agent chemotherapy as a first-line treatment. Previous studies suggest that only a third of patients have complete remission, with the remaining patients requiring toxic multiagent chemotherapy to attain remission. As stratification factors are unknown, some centres offer multiagent therapy upfront, resulting in overtreatment of many patients. We aimed to identify predictive factors for resistance to single-agent therapy to inform clinicians on which patients presenting with a FIGO score of 5 or 6 are likely to benefit from upfront multiagent chemotherapy. We did a multicentre, retrospective, cohort study of patients with gestational trophoblastic neoplasia presenting with a FIGO score of 5 or 6, who received treatment at three gestational trophoblastic neoplasia reference centres in the UK, Brazil, and the USA between Jan 1, 1964, and Dec 31, 2018. All patients who had been followed up for at least 12 months after remission were included. Patients were excluded if they had received a non-standard single-agent treatment (eg, etoposide); had been given a previously established first-line multiagent chemotherapy regimen; or had incomplete data for our analyses. Patient data were retrieved from medical records. The primary outcome was the incidence of chemoresistance after first-line or second-line single-agent chemotherapy. Variables associated with chemoresistance to single-agent therapies were identified by logistic regression analysis. In patient subgroups defined by choriocarcinoma histology and metastatic disease status, we did bootstrap modelling to define thresholds of pretreatment human chorionic gonadotropin concentrations and identify groups of patients with a greater than 80% risk (ie, a positive predictive value [PPV] of 0·8) of resistance to single-agent chemotherapy. Of 5025 patients with low-risk gestational trophoblastic neoplasia, we identified 431 patients with gestational trophoblastic neoplasia presenting with a FIGO risk score of 5 or 6. All patients were followed up for a minimum of 2 years. 141 (40%) of 351 patients developed resistance to single-agent treatments and required multiagent chemotherapy to achieve remission. Univariable and multivariable logistic regression revealed metastatic disease status (multivariable logistic regression analysis, odds ratio [OR] 1·9 [95% CI 1·1-3·2], p=0·018), choriocarcinoma histology (3·7 [1·9-7·4], p=0·0002), and pretreatment human chorionic gonadotropin concentration (2·8 [1·9-4·1], p<0·0001) as significant predictors of resistance to single-agent therapies. In patients with no metastatic disease and without choriocarcinoma, a pretreatment human chorionic gonadotropin concentration of 411 000 IU/L or higher yielded a PPV of 0·8, whereas in patients with either metastases or choriocarcinoma, a pretreatment human chorionic gonadotropin concentration of 149 000 IU/L or higher yielded the same PPV for resistance to single-agent therapy. Approximately 60% of women with gestational trophoblastic neoplasia presenting with a FIGO risk score of 5 or 6 achieve remission with single-agent therapy; almost all remaining patients have complete remission with subsequent multiagent chemotherapy. Primary multiagent chemotherapy should only be given to patients with metastatic disease and choriocarcinoma, regardless of pretreatment human chorionic gonadotropin concentration, or to those defined by our new predictors. None. For the Portuguese translation of the abstract see Supplementary Materials section.