Trend estimation in international comparative large‐scale assessments relies on measurement invariance between countries. However, cross‐national differential item functioning (DIF) has been repeatedly documented. We ran a simulation study using national item parameters, which required trends to be computed separately for each country, to compare trend estimation performances to two linking methods employing international item parameters across several conditions. The trend estimates based on the national item parameters were more accurate than the trend estimates based on the international item parameters when cross‐national DIF was present. Moreover, the use of fixed common item parameter calibrations led to biased trend estimates. The detection and elimination of DIF can reduce this bias but is also likely to increase the total error.