icon-    folder.gif   Conference Reports for NATAP  
  Conference on Retroviruses
and Opportunistic Infections
Seattle, Washington
March 4-7, 2019
Back grey_arrow_rt.gif
Integrase Inhibitor Exposure and Central Nervous System & Neural Tube Defects – Data from the Antiretroviral Pregnancy Registry
  Jessica Albano, PhD, MPH
9th International Workshop on HIV & Women Seattle, Washington 02 March 2019
Reported by Jules Levin







The figure [see this figure below following text] shows data from individual drugs that the APR has data on 200 or more first trimester exposures with birth outccomes. The name of the drug is supposed to be along the y axis (I think in the slide set it was missing) - the next column shows # defects/# live births for that drug, next column the % from that which is prevalence, the next 2 columns the lower and upper 95% confidence intervals, which are then shown graphically. There are two population based compaisons - one is the Metropolitan Atlanta Congenital Defects Program, in which the birth defect rate is 2.7% and the other is the Texas Birth Defects Registry which has a birth defect prevalence of 4.2% - these data reflect the boundaries of what would find for birth defects in the general population without HIV. Thus, we compare the prevalence for each drug (along with 95% CI) to these two rates and see if the rate for the drug is significantly different that them - as you see, all the 95% CI for all the ARV drugs overlap the normal range. Overall, the prevalence for any ARV with first trimester exposure is 2.7%, same as general population which is reassuring.
With 200 1st trimester exposures we can rule out a 2-fold increase in overall birth defects, which have a prevalence of about 2%.
However, once you want to look at specific organ systems, you need far more exposures to rule out an increased risk, depending on the prevalence of the birth defect. For example, in order to rule out a 3-fold increase in a birth defect with a prevalence of 0.1% - like neural tube defects - you need at least 2000 1st trimester exposures. That is what the below figure shows. The relative risk for a detectable defect is the y axis, the number of exposed births is the x axis. In blue you see with 200 exposures and an overall defect prevalence of 3%, you can rule out a 2-fold increase in risk for that defect with a specific drug. In orange you see for a defect like NTD with a prevalence of 0.1%, to rule out a 3-fold increase in risk you need at least 2000 exposures. Thus the ability to detect an increase in risk of a defect is related to the prevalence of the defect and the number of exposures you have.


The APR methodology is that there are several different types of reports. The PROSPECTIVE data is from women on ARV that the provider reports to the registry while she is still pregnant and then the registry follows up for outcome. This is unbiased as the outcome is unknown at the time of report. This is the primary analysis as it is unbiased and we have both a denominator (number of exposures) and a numerator (number with defects). The registry also gets RETROSPECTIVE reports - that is reports that come into the registry AFTER the infant is born and the outcome is known - these cases are biased because they are more likely to be births with a defect and we don't have a denominator - number of exposures - only the report. Thus, the retrospective reports are only useful to look at for trends and signals and not as the primary analysis. Finally, we get reports from clinical trials or cohorts - these go into a separate grouping - this would be like a perinatal trial that is reporting all outcomes after it is over, or a national registry like the UK/Ireland NSHPC registry. These are again not used in primary analysis, are used for trends. The below figure shows this.


Thus - in the propsective analysis, there are 174 preconception DTG exposures, with one CNS defect but it is not a neural tube defect. In the prospective data, there are no neural tube defects but only 174 relevant exposures for DTG, 186 for EVG, and 244 for RAL. From the above discussion you can see this is far too few exposures to be able draw conclusions about a defect with a low prevalence - in the US the prevalence of neural tube defects is only 0.06% because we have food folate fortification. Even if the prevalence was increased 10-fold by a drug - from 0.06% to 0.6%, you would not expect to see even 1 neural tube defect with this small number of exposures - with 200 exposures, you could easily see no defects even if the prevalence was 0.6%.
The RETROSPECTIVE data were also shown, and here we have several neural tube defect cases but we have no denominator for number of exposures. So these are not informative in telling us if there is an increased risk or not. You also can't say what a prevalence is from these data as you only have cases and not number exposed. One would expect if you had say 1000 exposures to a drug that did NOT cause neural tube defect that you might see 1 neural tube defect (with a prevalence of 0.1%). This is because neural tube defects occur even without drug exposure and you would expect to see SOME defects. So while we show the retrospective data, you can only say that one might expect to see some NTD and you can't use the data to say if a drug is or is not associated with an increase in NTD, only the prospective data can do that. Included in these retrospective reports are the data from the Botswana study which were reported to the Registry recently and a few other cases, as shown in the table.
Thus - only the propsective data are informative in terms of looking at risk, and the number of exposures remain too small for us to definitively say whether there is or is not an increased risk, as the prevalence in the general population is <0.1% which means that you need 2000 or more exposures to be able to say there is no increase in risk. So if you saw say 1 NTD defect with DTG with 1000 periconception exposures, that would be a prevalence consistent with the general population, prevalence of 0.1%. If we saw 5 defects with 1000 periconception exposures, that would be a 5-fold higher rate that the general population (0.5%) and would be concerning that there might be an increase.