Critical heat flux (CHF) bundle data from the Heat Transfer Research Facility of Columbia University are used to check the validity of the CHF approaches used in thermal-hydraulic system analysis codes for light water reactors. The CHF approaches assessed include the Biasi et al. correlation of TRAC, the Groeneveld et al. CHF table lookup approach of RELAP5/MOD3, the CHF table lookup approach of CATHARE, and the CHF approach of RETRAN. Depending on system pressure, RETRAN uses the B&W2, Barnett, and modified Barnett correlations and a linear interpolation scheme to predict CHF. Results show that among these CHF approaches, the Groeneveld et al. approach has the best prediction accuracy and the smallest uncertainty in the estimation of the HTRF bundle data. On the average, the Groeneveld et al. approach overpredicts the uniform axial heat flux distribution by 3.6% and the nonuniform axial heat flux distribution by 0.9%. The performance of the RETRAN approach is comparable with that of the Groeneveld et al. approach for uniform axial heat flux. In general, the accuracy and the uncertainty of all the approaches, except that of CATHARE, are worse under a nonuniform axial heat distribution than under a uniform axial heat distribution. All the CHF approaches assessed have a tendency to overpredict the HTRF bundle data at low pressure, low measured CHF, and high CHF quality. The performance of the Groeneveld et al. approach is improved through a CHF table update and modification of the bundle correction factor using the HTRF bundle data.