The interesting article by Lin et al.1 stimulated us to put forward a few comments/queries regarding their study design and its interpretation. Their study used population-based routinely coded information from the Taiwan National Health Insurance Research Data Set. In such data sets, the coding procedure, data entry and subsequent retrieval, is more likely to be inaccurate compared with individual case review with predetermined criteria.2,3 We conclude this from our own experience of conducting a study on antepartum haemorrhage in UK.4 The accuracy of such data can be improved by regular audit, and cross-checking with actual case records. Without such validation, the accuracy of any data set remains questionable. Even the high penalty for false insurance claims does not overcome basic deficiencies in classification and coding.
In Lin et al.’s1 comparative study, the control group was matched for age and year of delivery, but they ignored parity, an important variable affecting birthweight. In our earlier studies, we matched the controls for age, parity and timing of delivery,2,3 and found that the mean birthweight of infants in the group with active pulmonary tuberculosis was 215 g less than in the controls (2649 versus 2864 g; P < 0.001), primarily because of intrauterine growth restriction rather than preterm birth.2 In the current study, such matching would not have been difficult, and we consider that it should have been performed, particularly as the crude odds ratio for small for gestational age was borderline (1.22; 95% CI 1.00–1.49). Furthermore, the authors reported two continuous variables, gestational age and birthweight, as dichotomous variables,1 which is likely to have obscured the diagnosis of fetal growth restriction.2
Lin et al.’s study included women who had received anti-tuberculosis drugs during pregnancy, and analysed the pregnancy outcomes without addressing the site of tuberculosis. Tuberculosis at pulmonary and extrapulmonary sites could have different perinatal effects.2,3 As ICD-9-CM codes (011–018) were used, the authors could examine this aspect further.
It is also intriguing to note that Lin et al.1 selected a cohort with ‘a principal diagnosis’ of tuberculosis ‘within the 1 year preceding their index conceptions’. Many of these women would have completed a full course of anti-tuberculosis drugs (usually lasting between 6 and 9 months) before they became pregnant. As only 764 of the total 1820 women in this study received anti-tuberculosis drugs during pregnancy, we presume that the remaining 1056 women completed their course of medication before conception. Were pregnancy outcomes different between these two groups? Did the authors exclude all newly diagnosed cases of tuberculosis during pregnancy? If this was the case, then what was the mean time interval between diagnosis of tuberculosis and the index conceptions? It would also be helpful if the authors provided details of the anti-tuberculosis regimens used and the number of women who received anti-tuberculosis drugs during the different trimesters of pregnancy. Did the timing and duration of chemotherapy during the index pregnancies affect the perinatal outcomes? Although addressing some pregnancy outcomes, the study is also noticeably silent on the vital issue of maternal and perinatal mortality.
Surprisingly, Lin et al.1 claimed that ‘As far as we know, this is the first report of pregnancy outcomes among women with TB in Asia’. We have identified five previous studies from India alone, including two from our own group, referred to by the authors in their article.2,3 Although we appreciate the authors’ effort to address an important, but underestimated issue of perinatal effects of maternal tuberculosis, we think that additional information could be gleaned from their data by careful re-analysis.