Pemetrexed is an important component of first line treatment in patients with non-squamous non-small cell lung cancer. However, a limitation is the contraindication in patients with renal impairment due to hematological toxicity. Currently, it is unknown how to safely dose pemetrexed in these patients. The aim of our study was to elucidate the relationship between pemetrexed exposure and toxicity to support the development of a safe dosing regimen in patients with renal impairment. A population pharmacokinetic/pharmacodynamic analysis was performed based on phase II study results in three patients with renal dysfunction, supplemented with data from 106 patients in early clinical studies. Findings were externally validated with data of different pemetrexed dosing regimens. Alternative dosing regimens were evaluated using the developed model. We found that pemetrexed toxicity was driven by the time above a toxicity threshold concentration. The threshold for vitamin-supplemented patients was 0.110 mg/mL (95% CI: 0.092-0.146 mg/mL). It was observed that in patients with renal impairment (estimated glomerular filtration rate [eGFR]: <45 mL/min) the approved dose of 500 mg/m2 would yield a high probability of severe neutropenia in the range of 51.0% to 92.6%. A pemetrexed dose of 20 mg for patients (eGFR: 20 mL/min) is shown to be neutropenic-equivalent to the approved dose in patients with adequate renal function (eGFR: 90 mL/min), but would result in an approximately 13-fold lower area under the concentration-time curve. The pemetrexed exposure-toxicity relationship is explained by a toxicity threshold and substantially different from previously thought. Without prophylaxis for toxicity, it is unlikely that a therapeutic dose can be safely administered to patients with renal impairment.