The outbreak of Marburg haemorrhagic fever in Angola in 2004–2005 shows once again the devastating and rapid spread of viral haemorrhagic fevers in medical settings where hygiene practices are poorly applied or ignored. The legacy of years of war and poverty in Angola has resulted in very poor medical education and services. The initial high rate of infection among infants in Angola may have been related to poor hospital practices, possibly administration of vaccines. Though the outbreak in Angola was in a part of Africa not previously known to have filovirus infection, prior ecological modelling had predicted this location and many others. Prevention of future outbreaks will not be easy. The urgent need is dissemination of knowledge and the training, discipline and resources for good clinical practice. Educating the public to demand higher standards could be a powerful tool. Good practices are difficult to establish and maintain on the scale needed.
Viral haemorrhagic fevers (VHFs) are zoonoses able to cause dramatic and devastating local outbreaks with very high mortality. The geographic and ecological niches they occupy, however, are relatively small, confined to equatorial regions of Africa, and to somewhat more extensive semidesert regions of Asia, most of which are sparsely populated. For many VHFs, primary infection of humans is not a common event. With the exception of the mosquito-borne flaviviruses (dengue and yellow fever), which are not discussed in this review because of their entirely different epidemiology, the common feature among the viruses that cause the largest epidemics is their capacity for person-to-person spread, particularly in medical settings where hygiene practices are poorly applied or ignored. The current known geographic distribution of most important of these is shown in Figure 1. This review examines the ecology and epidemiology of haemorrhagic fever viruses that can be spread from person to person and the mechanisms through which they are able, often with our assistance, to cause epidemics. All this with a view to understanding the most important aspects of prevention and control of these outbreaks.
The major players in producing outbreaks of human-to-human transmission are the filoviruses, Ebola virus (EBOV) and Marburg virus (MARV) haemorrhagic fever viruses, Lassa virus and Crimean–Congo haemorrhagic fever (CCHF) virus.1 Filoviruses causing human disease are confined to equatorial Africa (Fig. 1). Direct human contact with the primary infections host is clearly uncommon because the number of primary human infections observed has been very small. Mortality is high, sometimes as much as 90%. Lassa virus is an arenavirus confined to West Africa (Fig. 1), but the host, Mastomys natalensis, is a common commensal rodent in village houses, and therefore primary human infections are common.2 Mortality is about 16% in untreated hospitalized cases, perhaps only about 2% in the community, but the number of people exposed in West Africa is very large; therefore, the actual number of cases is considerable. CCHF is the most widespread (Fig. 1), occurring over most of Africa and parts of Eastern Europe, as the name suggests, but also much of the interior of the Middle East and Asia. However, it occurs mainly in less densely populated areas. It is carried and transmitted transovarially by many species of hard ticks and can transiently infect asymptomatically a wide range of mammals including domestic herds and, possibly, birds. Mortality has been recorded at between 30 and 90%. Sporadic primary cases are probably common in some more remote areas but underreported. Recently, it has become clear that the infection is not uncommon in Iran, Turkey and Afghanistan and probably also in Iraq. Local outbreaks can be dramatic, often involving medical staff, particularly surgeons.
History of the filoviruses
Small but lethal epidemics of the African filoviruses were first reported in the 1960s and 1970s. The first recorded was not in Africa, rather in Marburg, Germany, where it killed seven of 32 people infected. All had handled monkeys newly imported from Uganda or their tissues.3 Filoviruses, specifically Ebola, in subsequent African outbreaks killed up to 88% of patients.4 Most of these early outbreaks were known to many in the infectious diseases community, but with the exception of the Marburg outbreak, relatively ignored by the press and thus the public, except locally. This was the period before the advent of acquired immune deficiency syndrome (AIDS), when the public and to a great extent the international medical community had concluded that infectious diseases had been conquered by antimicrobials and vaccines. This view changed with the advent of AIDS. Soon after, in 1990, an anomalous (outside Africa) epizootic occurred in Reston, Virginia, in recently imported primates from the Philippines.5 Even though this Ebola-like Asian virus turned out not to be pathogenic for humans, the media suddenly discovered the newsworthiness and income potential of stories about filoviruses. An explosion of press and public interest led to intense scrutiny of the VHF scene, and EBOV in particular. EBOV outbreaks, which scarcely need dramatizing, became intensely reported, featured and fictionalized in books and films, to the extent that during the 1994 Kikwit Ebola outbreak it was estimated that considerably more press attended the outbreak than medical personnel sent to control the disease. Indeed, the possible index case was claimed to have been uncovered not by a researcher but by an enterprising and well-informed newspaper reporter who won the Pulitzer Prize for her efforts.6
To some extent, press interest can be justified by the apparent and mysterious resurgence of filovirus outbreaks in the 1990s after many years of silence from the more dramatic players, particularly Ebola. Other viruses of equal or greater public health importance, such as Lassa fever and CCHF, remained ruthlessly endemic and epidemic in poor populations but never caught the public eye. Between the Ebola, Sudan, outbreak of 19797,8 and the large 1994 outbreak based in a hospital in Kikwit, The Democratic Republic of Congo (DRC),9 there were only very few isolated cases reported of Ebola or Marburg disease and no outbreaks. In contrast, since 1994 there have been multiple small and some ongoing Ebola haemorrhagic fever outbreaks of Ebola Zaire in Gabon and the neighbouring DRC involving both people and great apes10 and a large outbreak of Ebola Sudan haemorrhagic fever in 2000 in Uganda.11 There have also been two major Marburg haemorrhagic fever outbreaks: one in eastern DRC12 and an ongoing epidemic (2004–2005) in Angola. Most, but not all, were single-point source epidemics, with index cases often unidentified and subsequent spread in hospitals and in rural villages because of inadequate facilities and poor practices in caring for the sick.
The Marburg mystery
Late in 2004 and into 2005, an epidemic of Marburg haemorrhagic fever swept through Angola. At first sight, this was the wrong virus, causing the wrong disease in the wrong patients in the wrong place. Angola is south and west of the previously known distribution of MARV, which had previously been limited to Uganda, adjacent western Kenya and eastern DRC and possibly Zimbabwe (one case). The mortality rate in Angola is very high, currently 88% (329 of 374), compared with only 23% in the original 1967 outbreak in Germany. In the outbreak in Durba, DRC, in 1998–2000, mortality was also high (83%).12 Falsely elevated mortality particularly early in outbreaks is often due to inadequate case finding, especially missed milder or asymptomatic infections. In Durba, the high mortality could have been due to inadequate case finding, but subsequent serological studies found only 2% of the local population antibody positive, thus presumed survivors;12 therefore, the original estimates may have been reasonably accurate. Moreover, initial reports were that over 75% of the Marburg haemorrhagic fever cases in Angola were in children, mostly infants. This was also unexpected because Marburg has with one exception only been reported in adults, and reports of children with Ebola are also scarce.
Strain variation among filoviruses
Sequencing of filoviruses, both Ebola and Marburg, has shown remarkable conservation over time.13 Clearly, these viruses are not subject to immune pressure at least within the human population. Within epidemics, the strains are conserved. The only exception has been the Durba, Marburg haemorrhagic fever outbreak, where at least four strains with at least 20% nucleotide diversity were isolated, suggesting multiple introduction of viruses from an ecologic reservoir. However, within chains of transmission even in Durba, strains were conserved. Strain variation would be a primary candidate to explain the new observations in Angola. This is unlikely because preliminary genetic sequencing does not suggest that the Angola outbreak is due to a markedly different strain from those that caused previous outbreaks.14
Geographic distribution of primary filovirus infections
Put into the context of our limited understanding of filoviruses, we should probably not be surprised at all by the recent outbreak. The expected distribution of MARV is based on very few observations, essentially primates from Uganda (1967), twice in a cave in West Kenya, a single case possibly infected in Zimbabwe and the substantial outbreak in illegal gold miners in eastern DRC.12 More data points are available for EBOV infections, but with the exception of the Tai forest area in West Africa, Ebola outbreaks in Africa have been geographically more restricted. The real outliers were monkeys from the Philippines infected with Reston virus. Reston virus, however, does not cause human disease and has only been observed in epizootics in recently captured non-human primates from Southeast Asia.
Angola, among the poorest countries on earth, has been war torn for decades, with infant mortality of 192 per 1000 live births and a life expectancy at birth of only 36.8 years;15 Uige province, where the MARV outbreak originated in 2004, has numerous refugees from years of fighting. It borders DRC, one of the homes of EBOV, but is difficult to access, with multiple landmines, little or no infrastructure and poor health care. Refugees have been driven by these conditions into towns such as Uige where the outbreak was centred.
A year before the Angola Marburg haemorrhagic fever outbreak, a report was published that used ecological niche modelling to predict potential geographic distribution of filoviruses.16 This model restricts Ebola haemorrhagic fever to the broadleaf tropical rainforest, very humid areas concentrated along the equator in Central Africa and parts of West Africa, where conditions are very similar. Angola is to the south and outside this niche. Very interesting in this respect is that further analyses by Peterson et al.16 of Southeast Asia predict suitable ecological niches for EBOV in the island of Mindanao in the Philippines, which is thought to be precisely the origin of some of the primates infected with the filovirus implicated in the Reston outbreak.17 Peterson et al.16 speculated that the EBOV strain from Sudan may occupy a slightly different ecological niche from the Zaire virus strain and that the Sudan strain has ecological similarities to the Reston virus.
The EBOVs are closely related antigenically and by sequence analysis, including the Reston virus, but MARV is genetically and antigenically distinct.18 Ecological niche modelling for MARV also differs from that for EBOVs, appearing to favour somewhat drier and more open areas and encompassing a larger geographic area, parts of which are further from the equator.16 The data for Marburg haemorrhagic fever modelling are based on only four distributional points. Nevertheless, reviewing the potential distribution maps the authors developed, it is extremely interesting to find that the northeastern province of Uige in Angola, where the 2004–2005 outbreak is only now taking place, had in fact been included in their prediction of potential MARV ecological niche distribution. Other sites where disease has not yet been reported were also predicted. It will be very interesting to see the review of their model by these authors if they now include the Angola outbreak in their analysis and also to see, with time, where other filovirus outbreaks may originate.
Potential reservoirs of filoviruses
Peterson et al.19 then used their ecological niche model and their expertise in natural history and mathematical modelling to draw up a list of potential primary filovirus hosts. Despite making some important assumptions to narrow the field, the list is still very long. Bats remain the primary suspects. Bats have been indirectly implicated in several outbreaks: both the Sudan outbreaks in the 1970s, both Kitum Cave outbreaks.13 Bats were also the species best able to maintain asymptomatic filovirus infection in laboratory experiments.20 Fruit bats congregate in remote caves and feed in the forest canopy with limited direct contact with people. It may be an intermediate host frequenting the canopy and the floor, such as a non-human primate is needed for human exposure. Peterson et al.19 go on to explain why the frustrating failure to identify infected primary hosts is no surprise to ecologists. Even if studies are focused correctly on species that are spatially distributed with the outbreak virus, and sampling correctly includes less common species, virus prevalence in the reservoir may be as low as 1%. If this is the case, more than 200 members of each species need to be sampled to have a 90% chance of detection. Specialized knowledge of each species will be needed, with appropriate equipment and techniques for capture. The process will be further complicated by limited populations and potentially endangered species on the list of suspects. If a species is detected with antibody to a filovirus or the virus detected in a specimen from that species, this does not prove that the species is the source, only that it was or is transiently infected by a filovirus. The difficulties in identifying the natural host have clearly been underappreciated and are a major technical challenge. The week this article went to press saw the definitive identification of three species of tropical rain forest fruit bats, long the primary suspects, as the reservoir of Ebola virus21. The range of these three species overlies with some precision the ecological niches proposed by Peterson for the Zaire strain of Ebola virus. We now need to identify the bat species (presumably different) which carry the Marburg virus which must, obviously, occupy a drier Savannah range.
Primary filovirus infections are infrequently reported. Overall, person-to-person spread has been the major source of infection in filovirus epidemics from the first Ebola haemorrhagic fever outbreak in Yambuku in 1976 to the present.4 In Yambuku, the prenatal clinic served many women but had limited equipment (needles and specula), which was constantly reused without adequate sterilization. Most of the cases became infected in the hospital, and the local community, despite their poverty and lack of education, quickly figured this out and abandoned the hospital, effectively terminating outbreaks.4 In Sudan, in 1976 and 1979, the same chain of events occurred, resulting in locals hiding their sick for fear of the hospital.7,8 In Kikwit in 1994, the hospital was the main site of transmission of the virus.22 In the large Uganda outbreak in 2000 caused by the Ebola Sudan strain, hospitals and clinics were again the foci.11 Close contact with patients ill with Ebola is the most important risk factor for illness. Other risk factors are infection from contaminated materials such as needles, contact with blood or secretions, preparation of a body for burial or, occasionally, sexual contact. Close contact with blood or tissues of infected monkeys is also an important factor. The virus enters through mucous membranes or skin lesions, and outbreaks have been abruptly terminated when blood and needle transmission were interrupted. There is still no treatment for filoviruses, therefore no candidates for prophylaxis. In later sporadic outbreaks, one secondary Marburg case nursed the index patient with no protective clothing and later assisted with resuscitation procedures. She also handled tissues wet with secretions from the companion of the index case, who was also infected.23 In 1980, a patient died 5 h after being admitted to a Nairobi hospital and the attending doctor subsequently became ill with, but survived, Marburg infection. Epidemiologic studies in Zaire and Sudan do not suggest spread through casual contact or by aerosol transmission. A formal study of risk factors for virus transmission in the Sudan epidemic in 1979 showed that caring for an ill patient carried a relative risk five times greater than persons with a lesser degree of physical contact, and no cases occurred in people who entered the room of an ill patient but had no physical contact.7 These data confirm that Ebola is not an airborne disease and depends on a close contact probably with infected blood or secretions for its propagation. The mode of acquisition of primary infection, however, is unknown. Though we await definitive, published evidence, with this knowledge it is reasonable to suggest that the high rate among children in Angola is related to poor hygiene on a hospital ward, where it has been suggested most of them were infected.
Other important viruses: Lassa and CCHF
Lassa fever ecology and epidemiology
This discussion of Marburg and Ebola outbreaks ignores the nosocomial haemorrhagic fever, which presents the greatest threat to public health. Lassa virus is an arenavirus and, like other arenaviruses, maintains itself as a lifelong infection of a rodent host M.natalensis, in which it is persistent and mostly silent.24 The rodent is well adapted to peridomestic life in village houses in West Africa. Accidental human exposure to the virus is therefore frequent. Among the VHFs, Lassa fever affects by far the largest number of people. Estimates projected from incidence measured prospectively in Sierra Leone and retrospectively in several West African countries conclude that in the 1980s there were over 200,000 infections a year across the region, with between 3000 and 5000 deaths.25 No estimates of similar accuracy of the current public health burden are available, but we can assume that this has increased greatly with increasing population, deteriorating housing, armed conflicts and as many as 2 million refugees. These refugees are mostly in the Mano River basin, which has always been the epicentre of Lassa fever.26 Lassa fever is an ever-present and likely increasing threat to native West Africans and to visitors, particularly relief workers in rural areas. It now affects larger communities in West Africa, outside its already broad area of rural endemicity.27 In 2000, at least four cases were imported into Europe. Several deaths in these workers were due in great part to lack of awareness of the risk, delay in diagnosis and delay in instituting effective antiviral therapy. Indeed, Lassa fever is an eminently treatable disease, using ribavirin, but treatment has to be instituted as early as possible to be effective. Post-exposure prophylaxis is now practised empirically using oral ribavirin, but its usefulness has not been systematically studied.
Despite Lassa fever being endemic over a large area for centuries or longer, the earliest cases of Lassa fever reported were associated with hospital transmission. Contact in households with persons ill, or recently ill with Lassa fever, as well as sexual contact with someone convalescent with Lassa fever is also an important risk factor for human-to-human transmission. Nosocomial transmission to hospital staff or other patients has been frequently recorded. It is well documented that person-to-person transmission in a hospital setting may be effectively prevented with simple barrier nursing techniques, available to most hospitals or clinics in many rural areas.28 These basic rules, however, may not be observed or even understood. Hospital outbreaks of Lassa fever have been consistently associated with inadequate disinfection, ill-advised surgery on patients with abdominal pain and fever, direct contact with infected blood and contaminated needles, indiscriminate use of needles for intravenous therapy or intramuscular injections along with inadequate needle and syringe sterilization (Fig. 2 A and B). These epidemics can be devastating, resulting in the deaths not only of patients but also of medical staff, surgeons, nurses and other scarce personnel.27
Geography and epidemiology of CCHF
Other hospital-based outbreaks have been associated with other haemorrhagic fever viruses, most notably CCHF. CCHF is a tick-borne viral disease known to occur from Eastern Europe to Asia, the Middle East and in all of Africa and People’s Republic of China. Humans are infected from ticks or by handling blood or secretions from infected people or domestic animals. CCHF virus infects at least 24 species of ixodid (hard) ticks, particularly Hyalomma species, which serve as both a reservoir and a vector of this agent. CCHF infection of humans has emerged as a sporadic but important disease particularly in arid livestock farming areas.29,30 A wide range of wild and domestic animals and birds may be infected.
The high risk of nosocomial outbreaks of CCHF was first recognized in 1976, when a laparotomy was performed on a patient in Pakistan with abdominal pain, haematemesis and melena. Eleven secondary cases in hospital staff resulted in three deaths, including deaths of a surgeon and an operating-theatre attendant.31 Since then, similar nosocomial outbreaks have been reported in many countries from South Africa to countries of the former Yugoslavia, Iran, Iraq, United Arab Emirates, Afghanistan, China and Russia.32
The wider impact
Potential for virus transmission in medical care settings other than hospitals
The scale of the problem of transmission of VHF viruses in poor medical settings is dwarfed by the public health impact with the spread of more common but initially silent viruses. To illustrate the enormity of the problem, in the early 1990s we investigated what was described to us by locals as ‘an outbreak of jaundice’ in a town in central Pakistan.33 On questioning, it appeared that this ‘outbreak’ had been going on for at least 24 months. Because this was not the usual pattern of hepatitis A or E outbreaks, the usual causes of jaundice outbreaks, we performed a randomized serological survey of the town to determine what might be happening. We found that 6.5% of the population had antibody to hepatitis C, and when we conducted a nested case–control study, the major risk factor was repeated injections for a wide variety of complaints at medical care providers, several of whom had minimal or no formal training (Fig. 3). Reuse of needles and syringes was common, and conditions were very poor (Fig. 4). Injections ranged from antibiotics to vitamins to ‘coloured’ water. A major precedent for the spread of hepatitis C by use of inadequately sterilized intravenous equipment comes from Egypt, primarily during treatment campaigns for schistosomiasis.34 With the knowledge of the practices and consequences of poor practice in Africa, and now in Asia, we have to conclude that transmission of blood-borne viruses in medical facilities of all kinds is probably common within the endemic area of the haemorrhagic fever viruses. Indeed, hepatitis C virus (HCV) and human immunodeficiency virus (HIV) may be the viruses most commonly spread by this method. The difference with the haemorrhagic fever viruses is that the consequences of haemorrhagic fever viruses are immediately noticeable, whereas with HCV and HIV it takes years, even decades, for the transmission to be appreciated.
To return to the recent Marburg haemorrhagic fever outbreak in Angola, there should have been no surprises because all the elements were in place. Though the virus was not known there before 2004, ecological modelling shows that Angola, particularly northern Uige where the outbreak occurred, should be considered within the Marburg range. The legacy of years of war and poverty led to very poor medical education and services, providing just the conditions needed for spread within medical facilities. The initial high rate of infection among children in Angola was most probably related to needle use, perhaps for vaccines, perhaps from multidose vials. The high mortality, given the recorded history of this virus, may also be related to parenteral exposure (particularly needle exposure), which is known to carry the highest mortality.4
Prevention of future outbreaks, given the economic situation in many parts of Africa, will not be easy. Availability of vaccines would clearly be desirable. Vaccines for all these viruses can technically be made, but sources of funding for development costs and then the cost of delivery to the people most at need are not available.35 Vaccination, were a vaccine available, is a practical solution for Lassa fever, given the size of the exposed population. For filoviruses and CCHF, it would be hard to identify the target population to vaccinate and even harder to reach them. Treatment, principally ribavirin, is available only for Lassa fever and CCHF, including after exposure.
What is needed now is to concentrate on disseminating and implementing the understanding of the risks of blood-borne viruses and the discipline of good training and good clinical practice. Indeed, training of medical staff is the critical component. With improved education of health personnel and application of basic barrier nursing, hospital personnel, patients and their families in endemic areas can be protected. Education of the public, the consumer, can be a powerful tool. In our study of HCV infection in Pakistan, we implemented prevention by teaching the community to insist on new needles and syringes, often buying their own and taking them to the medical facility. This works well when the public is aware and concerned about the threat but would be a considerable challenge in remoter areas. Nevertheless, the AIDS epidemic gives us a powerful educational tool, which should be used to try to get the message to as many of the public as possible, even in the poorest regions, so that pressure from the public for higher standards acts in concert with better training of medical staff. This is possible. Poor people are uneducated, not stupid. Even in the remotest settings, the community grasps very quickly that the hospital is where people become infected with VHFs, so they immediately desert the hospital, and even hide their sick from medical personnel. This is one of the major reasons why case finding can be so difficult. Indeed, over the years, this simple action by the community has been the most frequent and effective means by which filovirus outbreaks have been terminated. Nevertheless, however simple and cheap these measures may seem, good practices are difficult to establish and maintain on the scale needed in areas where education and training is limited and where basic hospital supplies are not readily available.