Objective To explore the risk factors underlying congenital syphilis (CS) and to build a hazards model to assess the risk of CS in offspring born to mothers with syphilis treated in gestation. Methods This prospective study observed 554 pregnant women with syphilis and their offspring recruited from August 2002 to May 2007 in Shenzhen Centre for Chronic Disease Control and Prevention. After treatment, all the women were followed up until the diagnosis of CS in their offspring was confirmed or denied. Comparisons were made between women bearing infants with CS and women bearing infants without CS to reveal the risk factors for CS. ORs and their 95% CI were calculated for each risk factor by using logistical regression analysis. Results Twenty-nine (5.2%) infants were diagnosed with CS. Univariable analyses showed that the reciprocal logarithm of the titre of non-treponemal antibodies in mothers (log (1/T); OR=11.18, p<0.001), gestational week (GW) at treatment (OR=1.10, p<0.001) and the interaction between these two variates (OR=1.09, p<0.001) was associated with CS. Multivariable analysis showed that only the interaction was significantly associated with CS (OR=1.09, p=0.047). Conclusions The risk of CS could be predicted by the interaction between GWxlog (1/T). Early treatment given to women with syphilis during antenatal care may be the only effective method to decrease the risk of CS.