This study indicates that traditional herbal medicine use may contribute to liver disease in Uganda. Use of traditional herbal medicines was independently associated with two to five fold increases in significant liver fibrosis. Herbs from the Asteraceae family were the most often used and showed the strongest association with significant liver fibrosis: a five-fold increase in all participants (p<0.001) and HIV-infected participants (p
Six of eight participants who took herbs in the Asteraceae family had significant liver fibrosis (see ). Many plants in the Asteraceae and Fabaceae families contain pyrrolizidine alkaloids, a known risk factor for veno-occlusive liver disease 
. Although none of the alkaloid-containing herbs used by participants in this study have been confirmed to contain pyrrolizidine alkaloids, ingestion of plants containing pyrrolizidine alkaloids caused outbreaks of veno-occlusive liver disease in Jamaica, India, Egypt, and South Africa 
. No outbreaks of veno-occlusive liver disease associated with pyrrolizidine alkaloids have been reported to our knowledge in East Africa. Pyrrolizidine alkaloids are inert until dehydrogenation by cytochrome P450 3A4 (CYP3A4) in the liver 
, where reactive toxic pyrrolic and N-oxide metabolites directly damage liver sinusoidal endothelial cells and hepatocytes (zone III of the liver acinus) 
. Pyrroles cause chromosomal damage in a dose-dependent manner, resulting in an inflammatory response that culminates in fibrin deposition 
Although plants in both the Asteraceae and Fabaceae families ingested by study participants may contain pyrrolizidine alkaloids, our data shows a strong association between significant liver fibrosis and use of herbs in the Asteraceae family but not the Fabaceae family. The literature about African traditional herbal medicines is limited and does not explain why this difference might exist. Traditional herbal medicine remedies used in Rakai and throughout Uganda are often mixtures containing multiple herbs 
. It is possible that herbs in the Asteraceae family are taken at high doses, or potentiate the toxicity of other herbs or hepatotoxins.
Two participants with fibrosis reported use of Vernonia amygdalina
in the Asteraceae family. This particular herb is commonly used in Africa is thought to have hepatoprotective properties 
. However, animal studies show that at higher doses, this member of the Asteracaae family may be hepatotoxic. In an in-vivo
liver injury model, low doses (250–500 mg/kg) of Vernonia amygdalina
were hepatoprotective, but a high dose (750 mg/kg) caused increased hepatoxicity 
Herbs from the Lamiaceae family were associated with a 3.4 fold increase in significant liver fibrosis among all participants in our study (p
0.017). Herbs in the Lamiaceae family have been associated with hepatoxicity in an in-vivo
rabbit model 
. In addition, Aloe, taken by two participants in our study, has been linked in case reports to severe hepatitis 
. However, data about the potential hepatotoxicity of many herbs used by participants in this study do not exist, or come from animal model studies only that should be interpreted cautiously.
The risk of significant fibrosis associated with herb use was similar in the overall and HIV- infected study populations. Data on herb use was limited in the HIV-infected population, and plant family specific analysis was only possible for the Asteraceae family. Only two HIV- infected participants reported using herbs in the Asteraceae family.
Despite the small number of HIV-infected participants in this study who reported herb use, it is important to note that ART may alter the toxicity profile of co-administered herbs. CYP3A4 is a major pathway for metabolism of a wide range of chemically distinct foreign compounds including phytochemicals and antiretroviral drugs 
. Antiretroviral drugs of the non- nucleoside reverse transcriptase inhibitor (NNRTI) and protease inhibitor (PI) classes are also inducers or inhibitors of CYP3A4 activity 
. Therefore, these drugs have potential to influence phytochemical toxification or detoxification pathways in the liver. For example, commonly used NNRTI in initial ART regimens in Uganda (efavirenz and nevirapine) are inducers of CYP3A4 and therefore have potential to increase generation of toxic metabolites of pyrrolizidine alkaloids 
. Inhibitors of CYP3A4 may lead to accumulation of phytochemicals or their metabolites in the liver which may also result in toxicity.
Conversely, herbs may potentiate ART toxicities by influencing antiretroviral drug disposition in the liver, kidney, and gut. Herbs may affect NNRTI and PI metabolism by CPY3A4 and alter activity of cellular drug transporters and glucuronidation pathways 
. Existing evidence from Africa about herb-ART interactions is limited to two herb families commonly used in South Africa: Hypoxis
(African potato) and Sutherlandia
, neither of which were taken by participants in this study. Hypoxis causes a dose-dependent inhibition of CYP3A4 up to 86% of the normal activity of CYP3A4 and 50% reduction of the expression of P-glycoprotein. Sutherlandia frutescens
also causes a dose dependent inhibition of CYP3A4 up to 96% of CYP3A4 activity 
. One participant in this study reported garlic use, which is known to significantly reduce concentrations of a PI (saquinavir), most likely by induction of CYP3A4 
. Since nevirapine and efavirenz are also eliminated by CYP3A4, garlic may reduce plasma levels of these drugs, but there are no clinical data on these interactions.
This study had limitations. The study was cross-sectional and only information about current herb use was available for analysis. Only 4% of participants in this study reported using herbs, compared to other studies in Uganda in which 60% of HIV-infected persons reported concurrent use of ART and herbs 
. Some misclassification of herb exposure could have occurred due to a social desirability or reporting bias, especially among HIV-infected persons on ART who are counseled to avoid herbs in the communities around Rakai. Only 2% of HIV- infected participants reported herb use. While this lower number of HIV-infected participants reporting herb use could represent effective counseling, the difference in herb use among those on ART and those not on ART was not significant (1% vs. 2%, p
0.42). The small number of participants reporting herb use limited many comparisons (e.g., herb-ART interactions) and suggests that our findings should be interpreted cautiously.
An important limitation of this study is the potential for reverse causality. Although the most frequently used families of herbs in this study contain known hepatotoxins (see ), it is possible that the association of fibrosis with herb use could represent reverse causality, or persons with symptomatic liver disease being more likely to use herbal medicines. According to consultations with local traditional practitioners, some of the herbs in Asteraceae and other families are sometimes prescribed for “fever with jaundice”. However, none of the study participants had been previously diagnosed with liver disease within the formal medical system or by traditional healers. Most herbs used in this study to treat fever are usually taken for general fever (“fever” or “malaria”), not fever with jaundice (“yellow fever”).