Background Formulae of estimated glomerular filtration rate (eGFR) based on serum creatinine (Scr) are routinely used in oncology patients, however, they are inaccurate in some populations. Our aim was to assess the agreement of eGFR formulae and thereby build a nomogram to predict the reliability of estimates. Methods Measured GFR (mGFR) using isotope from 445 oncology patients were compared with eGFR from six formulae (Cockcroft-Gault, Modification of Diet in Renal Disease (MDRD), modified MDRD formulae for Chinese (C-MDRD), Chronic Kidney Disease Epidemiology (CKD-EPI) Collaboration, Wright and full age spectrum (FAS)). Bias, precision and accuracy of eGFR formulae were examined. We also evaluated statistics of agreement: the total deviation index (TDI), the concordance correlation coefficient (CCC) and the coverage probability (CP). Multivariate logistic regression was applied to identify characteristics associated with inaccurate eGFR and construct a predictive nomogram. Results All eGFR formulae tended to overestimate the eGFR. The percentage of patients with eGFR within 30% the mGFR ranged from 38.0 to 62.8%. Cockcroft-Gault and MDRD showed low bias and high precision. The MDRD formula exhibited lowest TDI, meaning that 90% of estimations ranged from - 36 to 36% of mGFR. Multivariate logistic regression showed that inaccuracy of MDRD was found in elderly patients or in patients with eGFR greater than 120 ml/min. A nomogram was constructed to help oncologists to predict the risk of inaccuracy of eGFR. The calibration curve showed good agreement. Conclusions Our results suggest that the error of eGFR by any formulae was common and wide in Chinese oncology patients. Our nomogram may assist oncologists in decision-making when mGFR is needed.