The potential toxicity of treatment is a major factor in the consideration of what to take to prevent mother to child transmission. Toxicity may be transient or longer term and may potentially affect both the child and the mother.

Overall, information about the safety of drugs in pregnancy is limited and there are even fewer data on the long-term safety of antiretroviral therapy, particularly combination antiretroviral therapy, in either women or children. What information on adverse events that is available also has to be put into the context of HIV infection and stage of disease in the pregnant women on treatment, since HIV disease itself can be associated with poorer pregnancy outcomes.

In addition, there tend to be more reports of toxicity with older drugs, primarily because there has been more experience using them. This does not mean they are less safe than new drugs. A drug with a known toxicity profile may represent much less of a risk than a drug that has only been used rarely in pregnancy.

Also, some toxicity is related to changes in the way that a womens body processes drugs when she is pregnant. Sometimes the way drugs are dosed may have to be changed because changes in absorption and metabolism.

Most of the side-effects that have been reported in pregnant women on NRTIs are similar to those reported in non-pregnant HIV-infected individuals. In some studies, these side effects overlap with complications normally seen in pregnancy or after childbirth. Typical side effects include anaemia, nausea, vomiting, raised liver enzymes, or elevated blood sugar levels (Lorenzi 1998). The most frequent side-effects in infants exposed to NRTIs in the womb are blood disorders such as neutropenia and anaemia (Mandelbrot 2001; Taha 2002).

As a class, the NRTIs have been associated with mitochondrial toxicity. This is damage to the mitochondria, the energy-producing organelles within cells. This can cause fat loss, peripheral neuropathy and lactic acidosis. ddC (zalcitabine, Hivid) is the most potent NRTI in terms of mitochondrial toxicity, followed by ddI, d4T, 3TC, AZT and abacavir (Ziagen). Animal studies have provided some evidence of mitochondrial toxicity, in monkey and mouse foetuses exposed to AZT and to 3TC (Ewings 2000; Olivero 2002; Walker 2004).

One of the most serious consequences of mitochonidrial toxicity is lactic acidosis, a dangerous elevation of lactate levels in the blood. The United States Food and Drug Administration and the manufacturer of d4T and ddI, Bristol-Myers Squibb, issued a warning to healthcare workers in January 2001 that pregnant women taking the two drugs may be at increased risk of lactic acidosis (Sarner 2001).

The amount of mitochondria in placental cells is lower in women taking NRTIs compared to HIV-negative women (Shiramizu 2003). Similarly, one-year-olds exposed to NRTIs in the womb have been shown to have fewer mitochondria than unexposed infants (Poirier 2003). It is not known whether this will translate into long-term health effects. So far, conflicting results have been reported, although French researchers have suggested that NRTIs taken by pregnant HIV-infected women may increase a child's risk of developing brain abnormalities. The risk associated with exposure to a combination of NRTIs was greater than that seen when a mother received only AZT monotherapy (Barret 2003). Several studies have provided further evidence to support the theory that NRTIs may cause clinical effects and abnormal metabolic measures in some exposed infants (Alimenti 2003; Divi 2004; Ekouevi 2003; Foster 2001; Giaqinto 2002; Noguera 2003).

However, many experts remain unconvinced that NRTI-induced toxicity is causing significant mitochondrial dysfunction in infants exposed to NRTIs, particularly when balanced against the health benefits of reduced HIV transmission (Morris 1999). In the United Kingdom, the Committee on Safety of Medicines issued guidance to National Health Service Trust medical directors in June 1999, concluding that a causal link between antiretroviral use in pregnancy and toxicity had not been proven, whilst the benefits of AZT in preventing transmission were well established. Furthermore, several large-scale reviews of children exposed to the NRTIs in the womb suggest that these drugs are not causing significant toxicity (Buchholz 2002; European Collaborative study 2003; Hanson 1999; Lipshultz 2000; Smith 1999).

AZT remains the only antiretroviral that has been thoroughly studied in pregnancy, and for which medium-term follow-up on exposed mothers and infants is available. Four-year follow-up on 234 uninfected children born to HIV-positive women who took the ACTG 076 regimen in pregnancy showed no evidence of illness or developmental abnormalities compared to untreated children (Culnane 1999). This has been confirmed in more recent prospective observational and randomised studies with up to 18 months' follow-up (Chotpitayasunondh 2002; Lipshultz 2000).

Respiratory distress and mild jaundice have been reported in several infants exposed to abacavir-based combinations (Puga 2002).

The effects of tenofovir exposure in the womb are still not well understood. A study in monkeys showed that the drug was associated with lower body weight and a small reduction in foetal bone porosity (Tarantal 2002).

Protease inhibitors have generally been reported to be well tolerated in pregnancy. However, all of the protease inhibitors carry the risk of nausea during the first months of treatment, which may be exacerbate the nausea that occurs during the first twelve to 14 weeks of pregnancy. This is a particular problem for full-dose ritonavir (Norvir), which is poorly tolerated during pregnancy due to nausea and vomiting, as well as diarrhoea and taste changes (Limpongsangurak 1999).

More serious side-effects include raised blood sugars and the development or exacerbation of diabetes (Justman 2003; Stek 2004; Timmermans 2005). There is also evidence that women taking protease inhibitor-based combination regimens may be at increased risk of pre-eclampsia, a life-threatening condition associated with high blood pressure and protein in the urine which affects a small proportion of pregnant women (Suy 2004). It seems that HIV-infected pregnant women are at lower risk of pre-eclampsia than the general population but combination therapy restores the risk of HIV-infected women to normal (Wimalasundera 2002).

There are also concerns related to drug levels, since a number of studies have shown protease inhibitor levels below the therapeutic level in pregnant women, and two studies have also showed that negligible amounts of the prorease inhibitors reach the foetus (Marzolini 2002; Mirochnick 2002). Indinavir (Crixivan) is not recommended for use in pregnant women in the United States due to low blood levels. However, using low doses of ritonavir to boost levels of other protease inhibitors can counteract the effects of pregnancy on saquinavir levels (Acosta 2004; Zorrilla 2003).

There is evidence that nelfinavir does not cause an elevation in the risk of birth defects (Conner 2003).

The NNRTI nevirapine has not been associated with a significant increase in birth defects, while toxicity following the use of single dose nevirapine to prevent HIV transmission is exceedingly rare and tends to be limited to rash (Guay 1999; McIntyre 2000). However, long-term use of nevirapine is more risky, since it can cause severe rash or liver problems in women, and elevated liver enzymes and mild blood abnormalities in infants (Edwards 2001; Taha 2002; Thomas 2005). The risk of serious side-effects is elevated in women with CD4 cell counts over 250 cells/mm³ and in pregnant women (Timmermans 2005). Treatment guidelines now recommend that pregnant women with CD4 cell counts above this level should avoid nevirapine.

Efavirenz is classified by the Food and Drug Administration in Pregnancy Category D. This means that it may cause harm to the foetus when administered to a pregnant woman during the first trimester. This classification is based on studies in monkeys have shown that exposure to efavirenz can result in serious birth defects (neural tube defects), as well as a number of reports of isolated cases of congenital defects. These included Dandy-Walker central nervous system malformation in one foetus from a spontaneous abortion, a foetus with a neural tube defect in a pregnancy with elective termination in second trimester, and two cases of myelomeningocoele in infants born to women who were receiving efavirenz at the time of conception and during the first trimester. Although a causal relationship of these events to the use of efavirenz has not been established, these defects were considered to be similar to the observations in animal studies (De Santis 2002; Fundaro 2002).

During the first 14 weeks of pregnancy, the foetus is most vulnerable to any toxic effects of drugs. Taking anti-HIV drugs during this time may increase the risk of birth abnormalities, although there are very few data on the effects of antiretrovirals early in pregnancy. However, observational data are increasing as number of HIV-positive women who conceive on antiretroviral therapy.

For example, the European Collaborative Study has found no association between antiretroviral therapy in the first trimester or later in pregnancy and an elevated risk of congenital abnormalities. This study included 3740 infants, of whom almost 1770 were exposed to antiretrovirals in the womb. Nineteen women also took efavirenz in the first trimester, with no congenital abnormalities seen in their children (European Collaborative Study 2005).

One study reported that 23 of 100 HIV-infected pregnant women continued to take their antiretroviral regimens during the first trimester of pregnancy and no birth defects were reported (Bucceri 2002). Similarly, a retrospective British study looking at 195 mother-infant pairs found no malformations among infants exposed to antiretroviral therapy early in pregnancy (Jungmann 2001). Furthermore, a Brazilian study found that low bith weight, but not congenital abnormalities was more common in women starting antiretroviral therapy before conception than those starting treatment after the first trimester (Machado 2005).

Some women on treatment at the time of conception choose to interrupt therapy for the first three months of pregnancy due to concerns about the possible toxic effects of drugs. However this will likely result in a rebound in viral load within a few weeks. Some experts have argued that this rebound may in fact increase the risk of mother-to-baby transmission.

It is also worth noting that any of the toxicities associated with beginning a new treatment regimen may have to be endured all over again if drugs are re-started later in pregnancy, because the body may no longer be adjusted to particular blood levels of the drugs.