Randomized trials comparing high-dose methotrexate (HDMTX) plus leucovorin (LV) with standard-dose methotrexate (SDMTX) have not detected a therapeutic advantage for the HDMTX arm despite compelling evidence from experimental systems. We hypothesized that these negative trials might reflect modulation of the antitumor effect of methotrexate (MTX) by LV. To test this we randomized 61 patients with locally advanced and recurrent squamous cell head and neck cancer to receive SDMTX (40 mg/m2 weekly for 8 weeks) and either LV or placebo "rescue" starting 24 hours later. Of the 61 randomized patients, there were protocol violations in two cases, leaving 59 patients evaluable for response using standard criteria, and for toxicity using the Eastern Cooperative Oncology Group (ECOG) scale. Of the 29 patients randomized to MTX plus LV, there were five responders (17.2%) compared with 11 of 30 (36.7%) patients randomized to MTX plus placebo (P = .047). Response was influenced independently by age, gender, and by previous treatment. Toxicity overall was more severe in patients randomized to MTX plus placebo (P = .016). This was accounted for primarily by differences in toxicities related to bone marrow function (neutrophil and platelet counts), stomatitis, and elevations of liver function tests. MTX therapy was more often interrupted for toxicity in the placebo group (P = .007) and discontinued for progressive disease in the LV group (P = .07). These results indicate that at the doses of MTX and LV used, LV modulates the antitumor effect as well as the toxicity of MTX in patients with head and neck cancer.