Cisplatin is an effective antineoplastic agent, but can cause renal tubular damage leading to urinary magnesium wasting and hypomagnesaemia. Cisplatin and 5-fluorouracil, when used in combination, have synergistic antitumour activity in upper gastrointestinal malignancies, but it is unclear whether they have additive effects on renal magnesium loss. To determine the optimal regimen for magnesium supplementation in these patients, we have conducted a randomised trial of routine intravenous magnesium supplements compared with magnesium given on an 'as required' basis. 32 patients were randomised to receive magnesium intravenously in prehydration and posthydration fluids with cisplatin chemotherapy, or to receive magnesium only when the serum level was low. 5-fluorouracil was given as a continuous infusion. Serum magnesium was measured on admission for each cycle of chemotherapy and an interim measurement performed between each cycle. 28 patients were evaluable. All patients randomised to receive magnesium on an 'as required' basis had at least one episode of hypomagnesaemia. On subsequent admissions for chemotherapy (cycles 2 and 3), the mean serum magnesium level was significantly lower in these patients compared with patients who received magnesium routinely (P < 0.05). After omission of magnesium from the first cycle of cisplatin, magnesium supplements were necessary in 50% of subsequent cycles, usually by the second or third cycle. Moreover, there were several instances of symptomatic hypomagnesaemia requiring further intravenous supplements in mid-cycle. Patients treated with a combination of cisplatin and 5-fluorouracil should be given intravenous magnesium supplements with each cycle of cisplatin chemotherapy. Nevertheless, episodes of hypomagnesaemia still occur, and additional intravenous supplements may be required, highlighting the importance of measuring this electrolyte.