BackgroundPregnancies following assisted reproductive technology (ART) may have elevated potential risk of pregnancy loss (PL) when compared to natural conception. However, rare studies comprehensively analyzed the IVF/ICSI cycle-dependent factors for loss of clinical pregnancy. Therefore, we aimed to determine the ART subgroup-specific risks of PL throughout pregnancy and explore different risk factors for early miscarriage and late miscarriage among pregnancies conceived through ART.MethodsA retrospective cohort study was launched in two infertility treatment centers in Nanjing and Changzhou including 5485 IVF/ICSI embryo transfer cycles with known outcomes after clinical pregnancy by the end of 2015. Cox proportional hazards regression analysis was performed to estimate the hazard ratios and their 95% confidence intervals. The associations between survival time during pregnancy and demographics and clinical characteristics of clinical pregnancies were estimated using the Kaplan-Meier method and the Log-rank test.ResultsThe overall PL rate in current ART population was 12.5%. Among the 685 pregnancy loss cycles, a total of 460 ended as early miscarriage, 191 as late miscarriage. We found couples in ART pregnancies demonstrated a significantly increased risk of PL as maternal age (HR = 1.31, Ptrend < 0.001) grows. Pregnancies received controlled ovarian hyperstimulation (COH) protocol like GnRH antagonist protocol (HR = 3.49, P < 0.001) and minimal stimulation protocol (HR = 1.83, P < 0.001) had higher risk of PL than GnRH-a long protocol. Notably, in contrast to fresh cycle, women who received frozen cycle embryo had a significant increased risk of early miscarriage (P < 0.001), while frozen cycle was linked with lower risk of late miscarriage (P = 0.045). In addition, four factors (maternal age, COH protocol, cycle type and serum hCG level 14 days after transfer) had independent impact on miscarriage mainly before 12 weeks of gestational age.ConclusionsWith these findings in this study, clinicians may make it better to evaluate a patient’s risk of PL based on the maternal age at the time of treatment, COH protocol, cycle type and serum hCG level 14 days after transfer and the gestational week of the fetus, and we hope that it contributes to future study on its etiology and guide the clinical prevention and treatment.Electronic supplementary materialThe online version of this article (10.1186/s12958-018-0390-6) contains supplementary material, which is available to authorized users.