Objective <p>We aimed to investigate the relationship between the prognostic nutritional index (PNI) during the third trimester and the risk of preeclampsia (PE) in twin pregnancies.</p> Method <p>A total of 2998 twin pregnancies were enrolled in Fujian Maternal and Child Health Hospital from January 2015 to December 2021, including preeclampsia group (<i>n</i> = 421) and control group (<i>n</i> = 2577). The significance of the characteristic variables in predicting PE in twin pregnancies were calculated using the random forest algorithm (Boruta package) and the correlation between PNI and PE in twin pregnancies was examined in three distinct models using multivariable logistic regression corrected for confounders. Receiver operating characteristics (ROC) curves were used to evaluate the ability for PNI to predict PE in twin pregnancies.</p> Results <p>PNI (37.92 ± 3.86 vs. 40.57 ± 3.63, <i>P</i> &lt; 0.001) was significantly lower in the PE group than in the control group. After adjusting for all covariates, the PNI was negatively associated with PE in twin pregnancies (OR = 0.780; 95% CI: 0.753, 0.808). Meanwhile, the higher PNI remained an independent protective factor for PE in twin pregnancies compared to lower PNI (OR, 95% CI: 0.410, 0.438–0.530; 0.144, 0.103–0.201) in sensitivity analysis. ROC curve analysis revealed an area under curve (AUC) of 0.691 for PNI and the cut-off value of PNI was 40.162.</p> Conclusion <p>PNI was negatively correlated with the risk of PE in twin pregnancies, which may help in risk assessment for twin pregnancies.</p> Clinical trial number <p>Not applicable.</p>