Abstract Background Studies that explore the controversial association between parity and anaemia-in-pregnancy (AIP) were often hampered by not distinguishing incident cases caused by pregnancy from prevalent cases complicated by pregnancy. The authors' aim in conducting this study was to overcome this methodological concern. Methods A retrospective cohort study was conducted in Oman on 1939 pregnancies among 479 parous female participants with available pregnancy records in a community trial. We collected information from participants, the community trial, and health records of each pregnancy. Throughout the follow-up period, we enumerated 684 AIP cases of which 289 (42.2%) were incident cases. High parity (HP, ≥ 5 pregnancies) accounted for 48.7% of total pregnancies. Two sets of regression analyses were conducted: the first restricted to incident cases only, and the second inclusive of all cases. The relation with parity as a dichotomy and as multiple categories was examined for each set; multi-level logistic regression (MLLR) was employed to produce adjusted models. Results In the fully adjusted MLLR models that were restricted to incident cases, women with HP pregnancies had a higher risk of AIP compared to those who had had fewer pregnancies (Risk Ratio, RR = 2.92; 95% CI 2.02, 4.59); the AIP risk increased in a dose-response fashion over multiple categories of parity. In the fully adjusted MLLR models that included all cases, the association disappeared (RR = 1.11; 95% CI 0.91, 1.18) and the dose-response pattern flattened. Conclusions This study shows the importance of specifying which cases of AIP are incident and provides supportive evidence for a causal relation between parity and occurrence of incidental AIP.