About 50% of patients infected with human immunodeficiency virus (HIV) are women, and among them, 80% to 90% are of reproductive age.1,2 Improved prognosis and high accuracy of preventive measures enabling a low rate of mother-to-child virus transmission, allow infected women, in the era of highly active anti-retroviral therapy, to more optimistically consider pregnancy.3,4 Thus, pregnancy in HIV-infected women is becoming more common. In addition to monitoring infection and anti-retroviral therapy, physicians must deal with the specific care related to the pregnancy and antenatal screening.
In developed countries, screening for Down syndrome is currently the standard of care and is offered to each singleton pregnant woman. Multiple-marker testing identifies pregnancies at risk for Down syndrome. Up to four biochemical markers–alpha-fetoprotein (AFP), human chorionic gonadotrophin (hCG), unconjugated estriol, and inhibin A–can be measured in maternal serum between 14 and 18 weeks of gestation and a risk based on maternal age and biochemical marker results calculated.5 To increase sensitivity without increasing the false-positive rate, this biochemically estimated risk can then be combined with risk based on nuchal translucency measurement. When the combined risk for Down syndrome is greater than that for the 95th percentile of the population, recommendations for undergoing amniocentesis are proposed to assess fetal karyotype.6 Amniocentesis is associated with several complications, especially fetal loss, the rate of which ranges from 0.5% to 1%.7 In HIV-infected women, even if few data are available, amniocentesis may also be associated with vertical virus transmission.8 Therefore, the false-positive rate with screening for Down syndrome is, especially in this population, of great concern so as to not add the risk of virus transmission to that of fetal loss with an unnecessary procedure.
A few recent studies9–11 have suggested an influence of HIV infection and anti-retroviral therapy on maternal levels of AFP and hCG during pregnancy. Human chorionic gonadotrophin levels increased and AFP levels decreased with infection. Therefore an increased false-positive rate with marker screening is suggested in this population. We aimed to estimate the influence of HIV infection and anti-retroviral therapy on maternal levels of AFP and hCG and the false-positive rate of maternal biochemical screening for Down syndrome to evaluate the relevance of this screening for HIV-infected pregnant women.
MATERIALS AND METHODS
Approval for this study was obtained from the institutional review board Comité de Protection des Personnes se prêtant à la Recherche Biomédicale Ile de France 3. This case–control study involved all known HIV-infected pregnant women with a single pregnancy who had a serum screening between 14 and 18 weeks of gestational age from January 1, 2001, through December 31, 2005, and delivered in our center. For both case and control groups, all late miscarriages, stillbirths, and malformation were included for representation of the screened population. We matched each subject with a control subject who had delivered during the same period, had a singleton pregnancy, and had undergone serum screening for Down syndrome in the same laboratory. We adjusted for all known confounding factors. Nevertheless, marker levels were stated in terms of multiple of the median (MoM) and thus were standardized on the most important confounding factors: gestational age, tobacco use, weight, and diabetes mellitus.5,12–14 Therefore, to adjust for confounding factors not taken into account in the calculation of this standardized rate, we matched cases and controls on geographic origin and fetal sex.
We adopted a second- trimester double test based on serum AFP and total hCG levels. Women underwent maternal serum sampling after giving signed, informed consent to undergo serum screening for Down syndrome. Serum AFP and total hCG concentrations were measured by the Immulite 2000 immunoassay analyser (DPC-Siemens Diagnostic, Tarrytown, NY). Risks were calculated using the software Prisca 4.0 (Typolog, Tornesch, Germany) with corrective factors for weight, tobacco use, and diabetes mellitus. Given results are combined risks, integrating maternal age, and biochemical markers.
Collected data were demographic data, geographical origin, fetal sex, tobacco use, diabetes mellitus, presence or absence of anti-retroviral therapy at the time of serum sample, details of anti-retroviral therapy use, viral load, CD4+ cell count, obstetric and neonatal outcomes, aneuploidy, MoM total hCG and AFP levels, combined risk (risk based on maternal age and biochemical risk), and false-positive rate of the combined risk. Because maternal age is a determinant in calculating the combined risk, we analyzed a risk based only on biochemical markers, the biochemical likelihood ratio. Likelihood ratios were obtained from age risks and combined risks by dividing the combined odds by the age odds.15 For combined screening (biochemical plus age risk), the threshold of risk commonly used is 1:250, corresponding to the 95th percentile of the general population. Thus, as Down syndrome is a rare event, the 95th percentile is considered the threshold for a false-positive rate of 5%. In the same way, we defined the threshold of biochemical likelihood ratio at the 95th percentile of the population. We therefore calculated the biochemical likelihood ratio for all women who received serum screening in our center’s laboratory during the study period. From 11,706 maternal serum screenings, the biochemical likelihood ratio threshold was 2.424. Beyond this threshold, the biochemical test result was considered positive.
To analyze the false-positive rate of the biochemical screening, we first compared cases, with no consideration of anti-retroviral therapy, with controls. Then, to study the possible effect of anti-retroviral therapy, we compared treated (anti-retroviral therapy) cases with their controls, untreated cases with their controls, and treated with untreated cases.
We opted for a matching strategy rather than a multivariate analysis because of the limitations due to our sample size and the low incidence of the outcome under study (false positive). Indeed, the number of predictors used in a logistic regression model is usually limited to 10% of the number of the outcome events we are trying to predict.16,17 So we were limited to two predictors at most.
Sample size was estimated so we could test an expected marker level of 0.8 MoM in cases compared with an expected level of 1 MoM in controls (alpha=0.05, power=0.8). Therefore, 43 patients were required for each group. As we planned to perform a sub-group analysis according to treatment, at least 86 HIV-infected women were required for an estimated treatment rate of about 50%. As data of more HIV-infected women were available, we decided to include all of them.
For matched samples, paired sample t test was used to compare means, Wilcoxon signed rank test for comparison of medians, and McNemar or Sign Test if required for categorical variables. For comparison of independent samples, t test, Mann–Whitney test, and χ2 test, or Fisher exact test was used.
Statistical analysis involved use of Stata software 9 (Stata Corp, College Station, TX). There were no missing data in variables used for matching, HIV status, anti-retroviral therapy, serum markers levels, risks, and pregnancy outcomes. For all others factors, there was less than 2% missing data. So we used no specific procedure for them.
Between January 1, 2001, and December 31, 2005, 339 HIV-infected women delivered singleton newborns in our institution. Among them, 132 (38.9%) were screened for Down syndrome by analyzing level of second-trimester serum markers (double test). Characteristics of patients with HIV (case women) and control women are summarized in Table 1. More than half of case women and matched control women (53%) originated from sub-Saharan Africa. Case women were slightly older than control women (32.9 years compared with 30.4 years, P<.01), and tobacco use was more common among case women (20.5% compared with 4.5%, P<.01). Less than half (47.7%) of patients were receiving anti-retroviral therapy at the time of screening. All treated patients received at least one nucleosidic reverse-transcriptase inhibitor, and 90% received highly active anti-retroviral therapy. CD4+ cell counts and viral loads are those obtained closest in time to the screening test.
Case women and control women did not differ in screening results. Nevertheless, HIV-infected women showed a lower but not significantly different median AFP MoM than control women (1.02 versus 1.11 MoM) (Table 2). Median total hCG MoM did not differ (1.01 versus 0.98 MoM). Qualitative analysis showed a higher but nonsignificant proportion of case women with serum AFP level lower than 0.5 MoM (4.5% versus 0.8%). Because case women were significantly older, the median age risk was higher for HIV-positive women than for control women (1:407 versus 1:729, P<.01), and the combined risk was significantly higher for case women (1:1339 versus 1:3269, P=.03) but not the biochemical likelihood ratio (0.29 versus 0.20). Case women and control women did not differ in false-positive rate for combined screening (combined risk greater than 1:250) (15.9% versus 12.9%) and biochemical screening (biochemical likelihood ratio greater than 2.424) (7.6% versus 6.8%).
The results of subgroup analysis are in Table 3. Untreated women (n=69) were similar to the respective control women (n=69), except for age and tobacco use. The median AFP MoM was significantly lower for untreated HIV-positive women than for the control women (0.91 versus 1.03 MoM, P<.01), with no difference in total hCG (0.96 versus 0.98 MoM). Likewise, as compared with their controls, the proportion of untreated HIV-positive women with serum AFP levels below 0.5 MoM was higher and the proportion with AFP levels above 2 MoM was lower despite nonsignificance. As they were older than control women, case women showed a higher mean age risk (1:504 versus 1:749, P<.01). The combined risk was also significantly higher for case women (1:1289 compared with 1:4178, P<.05) but not the biochemical likelihood ratio (0.3 compared with 0.18). The false-positive rates of combined screening (14.5% compared with 11.6%) and of biochemical screening (10.1% compared with 8.7%) did not significantly differ.
Treated HIV-infected women (n=63) were similar in demographics to their matched control women (n=63). The median AFP MoM did not differ (1.18 compared with 1.17 MoM, P=.85), nor did median total hCG MoM (1.03 compared with 0.98 MoM, P=.97). The mean age risk was higher for cases (1:346 compared with 1:707, P<.01), but combined risk (1:1536 compared with 1:2620) and biochemical likelihood ratio (0.23 compared with 0.23) did not differ. The false-positive rates of combined screening (17.5% compared with 14.3%) and biochemical screening (4.8% compared with 4.8%) did not differ.
Treated (n=63) and untreated (n=69) women did not differ in mean age, ethnic origin, or smoking habits, but fetal sex ratio significantly differed (28 males:35 females for treated compared with 43:26 for untreated women, P=.04). Viral load was lower for treated patients, but CD4+ cell counts and percentages did not differ between the groups. Median AFP MoM was significantly higher for treated than untreated women (1.18 compared with 0.91 MoM, P<.01) but not median total hCG MoM (1.03 compared with 0.96 MoM). As compared with treated HIV-infected women, the proportion of untreated HIV-infected women with AFP levels below 0.5 MoM was higher but nonsignificant. Nonetheless, significantly fewer untreated HIV-positive women had AFP levels above 2 MoM (0% compared with 11.1%). Age risk, combined risk, and biochemical likelihood ratio did not differ, nor did false-positive rates of combined screening (17.5% compared with 14.5%) or biochemical screening (4.8% compared with 8.7%). The rate of amniocentesis performed was not different between the groups (4.8% compared with 7.2%).
Figure 1 summarizes the results of screening outcomes. Median AFP MoM was nonsignificantly lower in HIV-infected women than in control women but was significantly different between untreated HIV-positive women and control women. Treated women showed the same median AFP MoM as control women. Whatever the comparison, median total hCG MoM, biochemical likelihood ratio, and false-positive rates for combined screening and biochemical screening did not differ.
In the three groups (treated and untreated case women and control women), Down syndrome was not diagnosed before or after birth.
Among HIV-positive women, only one case occurred of aneuploidy, a trisomy 13, diagnosed by amniocentesis after ultrasound suspicion; serum screening results were normal. Four isolated malformations occurred among cases: two nonlethal cardiac malformations and two nonlethal malformations of the urinary tract. One case of each occurred in treated and untreated groups.
Other pregnancy outcomes, rates of fetal loss (one case of late miscarriage in the untreated HIV-positive group) and preeclampsia, gestational age at delivery, birth weight, and Apgar score at 5 minutes were similar among treated, untreated patients and control women. Prematurity rate (gestational age less than 37 weeks) showed a trend, although not significant, for a twofold higher frequency in treated than in untreated women (22.6% compared with 11.9%). Finally, no women showed occurrences of mother-to-child transmission of infection.
In this case–control study, the false-positive rate with double-markers second-trimester Down syndrome serum screening, commonly used in European countries, seems not to be increased by HIV infection or therapy with anti-retroviral agents.
However, untreated HIV-infected women had a lower maternal median AFP MoM than treated women with HIV infection and control women, which suggests that HIV infection could be responsible for decreased concentration of AFP, which could be restored by anti-retroviral therapy. Alpha fetoprotein, at this stage of pregnancy, is exclusively synthesized by the fetal liver,18 and its release across the placenta is due to an active transportation process,19 so the lower concentration of AFP in untreated HIV-infected women may be that HIV is responsible for modified fetal-to-maternal transportation of this protein across the placenta.
Despite the decrease in AFP serum level, we found no effect of AFP level on biochemical likelihood ratio or anti-retroviral therapy–positive rate of biochemical screening for Down syndrome and no effect of HIV infection or anti-retroviral therapy on median total hCG MoM. The lack of effect of decreased level of AFP may have two explanations. First, AFP is the least sensitive marker for Down syndrome screening, but because it has relevance for neural tube-defect screening, it remains commonly used in multiple-marker screening.20 Its weight in calculation of Down syndrome risk is far below that of total hCG. Therefore, the observed difference in AFP levels could have only minor consequences on the calculation of Down syndrome risk, biochemical likelihood ratio, and false-positive rates of biochemical screening. Second, since our sample was small, we must acknowledge that statistical power might have been low. Indeed, to show an increase in false-positive rate from 5% to 10% (expected value), the sample size should have been 474 per group (alpha=0.05, power= 0.8). However, even if a difference in false-positive rate exists, it would be mild, with limited consequences, even with large-scale screening.
The proportion of untreated HIV-positive women with AFP levels above 2 MoM is lower in comparison with treated HIV-positive women and HIV-negative women. That suggests that threshold commonly used to screen neural tube defect is inadequate for untreated HIV-positive women. Due to absence of neural tube defect and limited sample size in our survey, we were unable to suggest a new threshold dedicated to this population.
To date, few studies have investigated maternal serum screening for Down syndrome in HIV-infected pregnant women. Among the few published studies, Gross et al10 found, in 49 HIV-infected pregnant women, a slight but positive correlation between AFP level and viral load. This result is contrary to our findings. Otherwise, they found a negative correlation between hCG level and CD4+ count. They suggested that the rate of false-positive screening results could be higher in HIV-infected women. This study was not designed to assess the usefulness of serum tests in this population and did not assess the impact on false-positive rate, nor the effect of anti-retroviral therapy. Einstein et al9 studied the impact of protease inhibitor treatment on marker levels of 39 HIV-infected pregnant women and found no difference in levels of AFP, hCG, or unconjugated estriol between treated and untreated patients. However, the authors observed lower levels of AFP with protease-inhibitor treatment compared with no protease-inhibitor treatment, but the group of patients with no protease-inhibitors was heterogeneous, combining treated and untreated patients. Therefore, no conclusion is possible in this study about the influence of protease inhibitors on marker levels.
In the only published matched case–control study comparing the false-positive rate of biochemical screening for Down syndrome between women infected and not infected with HIV, which included a relatively limited number of subjects (34 and 11) the authors found a significantly higher serum hCG MoM in HIV-positive women than in HIV-negative women. They also found a nonsignificant trend to a lower rate for AFP MoM in HIV-positive women. All their findings are consistent with ours. The authors found a difference in the false-positive rate in HIV-infected women compared with noninfected womenwhen using the quadruple screening test (AFP, hCG, unconjugated estriol and inhibin A) and no difference with triple screening. However, probably because of the limited number of patients, they did not separate treated and untreated women, so the study group was too heterogeneous to analyze the influence of HIV infection and treatment.
Our sample size is larger than that of previous studies (132 HIV-infected women), which thus implies a higher statistical power of our analysis. We tried to eliminate most analysis bias with a matching strategy. By emphasizing biochemical likelihood ratio rather than combined risk, we eliminated the bias of maternal age.
Because anti-retroviral therapy could be a factor in transplacental transfer of AFP and the hCG production by villous trophoblast cells, we thought that our HIV-positive group was too heterogeneous for simple comparison with an HIV-negative group, and a strength of our study is that we were able to separate treated from untreated infected women, for comparison with each other and with matched controls. Despite our retrospective design and the heterogeneous treatments with use of different antiretroviral molecules, we were able to assess the influence of HIV infection and anti-retroviral therapy on maternal serum marker levels and rates of false-positive of biochemical screening. By analyzing biochemical likelihood ratio, we were able to define modifications in biochemical-evidenced risk with HIV infection or anti-retroviral therapy and thus assess the effect on false-positive rates with biochemical screening.
We noted, when considering the combined risk for Down syndrome, a high rate of false positives in controls. The observed rate is more than twofold the expected rate (5%). Because the threshold of positivity (1:250) is designed for a 5% false-positive rate in the overall population, this high rate is probably due to the higher mean age of our controls (30.4 years) with respect to the overall population. A high false-positive rate (12%) in the control group was also reported by Yudin et al11 with triple screening. The authors’ explanation for the high rate was also mean age of patients.
As well, less than 50% of our HIV-infected sample underwent biochemical maternal serum screening for Down syndrome, even though 80% of women in the general population are screened. In our sample, social problems frequently associated with medical problems often lead to women seeking medical care late in the pregnancy. Furthermore, in addition to potential mother-to-child transmission, language barrier and cultural differences might explain different perception of aneuploidy and handicap and thus lead to low frequency of screening.
False-positive rate with this screening is a concern, especially in HIV-infected women in which added to the risk of very early preterm rupture of the amniotic membrane and fetal loss, exists as well the risk of mother-to-fetus transmission of infection. Even if the risk of vertical transmission seems to be reduced with anti-retroviral therapy and even if for some the risk is similar to that without the invasive procedure,21–23 further investigations with more cases are needed to precisely assess this risk. According to our data, HIV infection or anti-retroviral therapy doesn’t increase the false-positive rate of maternal biochemical screening, thus this test seems reliable in this population.
1. Centers for Disease Control and Prevention. HIV and AIDS–United States, 1981–2000. MMWR Morb Mortal Wkly Rep 2001;50:430–4.
2. Centers for Disease Control and Prevention. Epidemiology of HIV/AIDS–United States, 1981–2005. MMWR Morb Mortal Wkly Rep 2006;55:589–92.
3. Centers for Disease Control and Prevention. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. Department of Health and Human Services. October 10, 2006; 1–113. Available at: http://aidsinfo.nih.gov/contentfiles/PerinatalGL.pdf
. Accessed June 12, 2008.
4. US Department of Health and Human Services. Public Health Service Task Force recommendations for use of antiretroviral drugs in pregnant HIV-1 infected women for maternal health and Interventions to reduce perinatal HIV-1 transmission in the United States. Rockville, MD: HIV/AIDS Treatment Information Service. Available at http://AIDSinfo.nih.gov
. Retrieved June 4, 2008.
5. Wald NJ, Kennard A, Hackshaw A, McGuire A. Antenatal screening for Down’s syndrome. J Med Screen 1997;4:181–246.
6. Cuckle H, Sehmi I. Calculating correct Down’s syndrome risks. Br J Obstet Gynaecol 1999;106:371–2.
7. Papantoniou NE, Daskalakis GJ, Tziotis JG, Kitmirides SJ, Mesogitis SA, Antsaklis AJ. Risk factors predisposing to fetal loss following a second trimester amniocentesis. BJOG 2001;108:1053–6.
8. Shapiro DE, Sperling RS, Mandelbrot L, Britto P, Cunningham BE. Risk factors for perinatal human immunodeficiency virus transmission in patients receiving zidovudine prophylaxis. Pediatric AIDS Clinical Trials Group protocol 076 Study Group. Obstet Gynecol 1999;94:897–908.
9. Einstein FH, Wright RL, Trentacoste S, Gross S, Merkatz IR, Bernstein PS. The impact of protease inhibitors on maternal serum screening analyte levels in pregnant women who are HIV positive. Am J Obstet Gynecol 2004;191:1004–8.
10. Gross S, Castillo W, Crane M, Espinosa B, Carter S, DeVeaux R, et al. Maternal serum alpha-fetoprotein and human chorionic gonadotropin levels in women with human immunodeficiency virus. Am J Obstet Gynecol 2003;188:1052–6.
11. Yudin MH, Prosen TL, Landers DV. Multiple-marker screening in human immunodeficiency virus-positive pregnant women: screen positivity rates with the triple and quad screens. Am J Obstet Gynecol 2003;189:973–6.
12. O’Brien JE, Dvorin E, Drugan A, Johnson MP, Yaron Y, Evans MI. Race-ethnicity-specific variation in multiple-marker biochemical screening: alpha-fetoprotein, hCG, and estriol. Obstet Gynecol 1997;89:355–8.
13. Reynolds TM, Vranken G, Van Nueten J. Weight correction of MoM values: which method? J Clin Pathol 2006;59:753–8.
14. Spencer K. The influence of fetal sex in screening for Down syndrome in the second trimester using AFP and free beta-hCG. Prenat Diagn 2000;20:648–51.
15. Wald NJ, Cuckle HS, Densem JW, et al. Maternal serum screening for Down’s syndrome in early pregnancy. BMJ 1988;297:883–7.
16. Harrell FE. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. Springer series in statistics. New York (NY): Springer; 2001.
17. Mikolajczyk RT, DiSilvesto A, Zhang J. Evaluation of logistic regression reporting in current obstetrics and gynecology literature. Obstet Gynecol 2008;111:413–9.
18. Duc-Goiran P, Mignot TM, Robert B, Machavoine F, Mondon F, Hagnére AM, et al. Expression and localization of alpha-fetoprotein mRNA and protein in human early villous trophoblasts. Placenta 2006;27:812–21.
19. Newby D, Dalgliesh G, Lyall F, Aitken DA. Alphafetoprotein and alphafetoprotein receptor expression in the normal human placenta at term. Placenta 2005;26:190–200.
20. Muller F, Bussieres L. Maternal serum markers for fetal trisomy 21 screening. Eur J Obstet Gynecol Reprod Biol 1996;65:3–6.
21. Somigliana E, Bucceri A, Tibaldi C, Alberico S, Ravizza M, Savasi V, et al. Early invasive diagnostic techniques in pregnant women who are infected with the HIV: a multicenter case series. Am J Obstet Gynecol 2005;193:437–42.
22. Coll O, Suy A, Hernandez S, Pisa S, Lonca M, Thorne C, et al. Prenatal diagnosis in human immunodeficiency virus-infected women: a new screening program for chromosomal anomalies. Am J Obstet Gynecol 2006;194:192–8.
23. Maiques V, Garcia-Tejedor A, Perales A, Cordoba J, Esteban RJ. HIV detection in amniotic fluid samples. Amniocentesis can be performed in HIV pregnant women? Eur J Obstet Gynecol Reprod Biol 2003;108:137–41.