From 1993 to 2006, RFLP typing was considered the gold standard in typing of M. tuberculosis isolates, especially for strains harboring multiple IS6110 copies, like the ones of the Beijing genotype family. However, this typing method is technically demanding and time consuming . Furthermore, the discriminatory power of RFLP typing among strains with a low number of IS6110 copies (<= 5 copies) is very poor . Therefore, in recent years VNTR typing has increasingly been explored in the molecular epidemiology worldwide and with the proposal on international standardization of this technique in 2006 it has in fact become the new gold standard . VNTR typing introduced major advantages in typing in comparison to RFLP typing, such as its ease in use, its suitability for standardization, and that the results that are displayed in numbers can be analyzed easily and exchanged efficiently between laboratories. Moreover, the turnaround time of VNTR typing is much shorter than that of RFLP typing, because it is PCR-based and only a little amount of mycobacterial DNA is required.
Many researchers have carried out comparisons between RFLP and 12 or 15 loci VNTR typing methods for discriminating M. tuberculosis isolates [3, 4, 12]. Their findings showed that the discriminative power of 12 loci VNTR was lower than that of 15 loci VNTR (with HGDI of 0.978-0.995)  and 15 loci VNTR has high level of discrimination with HGDI 0.990-0.995 [4, 12], but this was still lower than that of RFLP typing (0.998) . However, Supply et al.  proposed to apply 24 instead of 15 loci in VNTR typing, and this improved the level of discrimination significantly.
In our study, we compared the performance of 15 and 24 loci VNTR typing and RFLP typing using 95 Beijing strains and we found that the discrimination index (HGDI) of 15 loci VNTR was the lowest (0.992), followed by both RFLP typing and 24 loci VNTR typing (0.994). However, the differences observed were small. The HGDI of some loci (VNTR 154, VNTR 2461,VNTR 3171) were low in our study, which means that these loci are less useful in discriminating Beijing strains in the South of Vietnam and presumably elsewhere. The HGDI of VNTR 2461, VNTR 577, VNTR 2163b, VNTR 580, VNTR 802, VNTR 960, VNTR 1644 and VNTR 4348, were similar to that observed in a previous study in Hong Kong  (Table 1), whereas the HGDI of VNTR 2996, VNTR 4052, and VNTR 2165 were significantly higher than the ones in that study  (Table 1), for unknown reasons. It may be that because BCG vaccination has been introduced much earlier in Hong Kong than in Vietnam, the ongoing selection of particular strains of the Beijing lineage  may be more advanced in the former than in the latter area and the mentioned loci may have a different level of discriminative power among the circulating strains in both areas.
Our study further found the HGDI of VNTR 1955, 2163b and 2165 to be very high (>0.50) and the best differentiation, similar to two previous studies [4, 12], was obtained with VNTR 2163b (Table 1 and Figure 2).
Some of the HGDI of individual loci in our study were significantly different to the ones found in the study of Alonso et al. (Table 1), because we performed VNTR typing of exclusively Beijing strains, whereas Alonso et al. carried out VNTR typing on a strain collection consisting of 32% LAM, 28% Haarlem and only 2% Beijing strains.
A disadvantage of VNTR typing encountered in this study was that six strains revealed double alleles in a single locus, and two strains even in two and more than two loci. It is not clear whether the latter observation was associated with a mixed infection . However, the revealed genomic instability in particular loci decreases the utility of VNTR typing significantly, as this hampers a reliable interpretation. Also in RFLP typing transposition of IS6110 sometimes interfered with a reliable interpretation, but such a genetic turn-over was observed less frequently . However, we cannot exclude the possibility that these multiple alleles may reflect important phenomena in the epidemiology of TB currently unknown, and these observations, although technically demanding, may be associated with the ongoing adaptation of M. tuberculosis to the current TB control measures.
A major limitation of this study was that we did not have epidemiological information available to verify the transmission links indicated by both typing methods. It was therefore, not possible to ascertain the validity of epidemiological links indicated.