Restriction fragment length polymorphism (RFLP) analysis is one of the tools commonly used to study the molecular epidemiology of porcine reproductive and respiratory syndrome viruses (PRRSVs). As PRRSVs are genetically variable, the stability of the RFLP pattern of a PRRSV during in vivo replication was evaluated by carrying out 13 sequential pig-to-pig passages (P1 to P13) of PRRSV ATCC VR-2332 in three independent pig lines for a total of 727 days. During P1 the pigs were inoculated with a homogeneous inoculum (CC-01) prepared through a series of plaque purifications, and during P2 to P13 the pigs were inoculated with a tissue filtrate from the corresponding pig in the previous passage. Fifteen viral plaque clones were directly isolated from CC-01 and the day 7 serum of each pig of each passage, open reading frame 5 of the clones was sequenced, and the clones were compared to CC-01 to assess the mutation rates and RFLP patterns (obtained by digestion with MluI, HincII, and SacII) over time. Among the 495 viral clones recovered during the passages, 398 clones, including CC-01, had pattern 2-5-2 (MluI-HincII-SacII); however, the remaining 97 viral clones showed different patterns (2-6-2 [P2], 1-5-2 [P3], 2-5-4 [P7], and 2-1-2 [P10]). Importantly, the MluI site that was reported to be present in only one of the PRRS modified live virus vaccine strains (Ingelvac) and its parental strain (ATCC VR-2332) can disappear during in vivo replication. Furthermore, sequence homology between CC-01 and clones with pattern 2-5-2 or clones with other patterns differed by 0.05 to 1.58% and 0.5 to 1.45%, respectively, suggesting that RFLP analysis cannot accurately predict genetic relatedness between PRRSVs. Collectively, precaution should be taken when the molecular epidemiology of PRRSVs is evaluated by RFLP analysis.