The Veterinary Record, May 24, 2003 641

Papers & Articles

MOLECULAR characterisation of rinderpest viruses distin-guishes three viral lineages of clear epidemiological signifi-cance – a single Asian lineage and two distinct African lineages(Barrett and others 1998). From the limited material archivedin the past 40 years, lineage 1 (Africa) appears to have beengenerally restricted to eastern Africa. However, representativesof this clade spread from the Sudan into eastern Nigeria andwere present in Egypt in the early 1980s, during the last greatAfrican rinderpest pandemic. In contrast, lineage 2 (Africa)was formerly widespread in West and eastern Africa, and wasrepresented by viruses from Nigeria in 1958 and by surviv-ing strains of viruses circulating in Tanzania and Kenya in the1950s and 1960s (T. Barrett, personal communication).Notable members of this group are RBT1, isolated from cattlein northern Tanzania in 1960 (Plowright 1963) and RGK1, iso-lated from a reticulated giraffe in north-east Kenya in 1962(Liess and Plowright 1964). In the late 1970s and early 1980sthe virus that spread eastwards from Mauritania and Mali towestern Nigeria was probably lineage 2, whereas lineage 1virus crossed from Sudan to eastern Nigeria. Together theywere responsible for the rinderpest pandemic that engulfedmost of sub-Saharan Africa at that time (Nawathe andLamorde 1983).

From the 1960s to the 1990s, all the rinderpest virusesretained from outbreaks in eastern Africa belonged to lineage1, including those isolated in Ethiopia in the early 1990s.Virusof lineage 2 was not recognised again until 1994 and subse-quently, when it was detected causing epidemiologicallylinked outbreaks of rinderpest in wildlife, primarily buffaloes,lesser kudu and eland, in and around Tsavo and NairobiNational Parks in southern Kenya (Barrett and others 1998,Kock and others 1999).

African lineages of rinderpest virus can cause high mor-tality in certain wildlife species (primarily buffalo, lesserkudu and eland) and in cattle, but there is substantial evi-dence, going back to the early 1900s, for the existence ofstrains which either cause less severe disease or possibly evencause no clinical signs in cattle (Robson and others 1959,Plowright 1963, 1998, Branagan and Hammond 1965, Taylorand Watson 1967). Although the phenomenon of lower vir-ulence for cattle is not restricted to lineage 2 viruses it has

been most clearly demonstrated for viruses of the clade inEast Africa, as in the case of lineage 2 rinderpest virus iso-lated from an eland in Nairobi National Park, Kenya, in 1996(C. Dunn, cited by Barrett and others 1998). In the early1960s there were several outbreaks of rinderpest in the areanorth and east of the Tana River in eastern Kenya (Stewart1964) – an area which yielded the RGK1 lineage 2 virus froma giraffe in 1962, a virus fully virulent for cattle. However,rinderpest has not been confirmed in north-eastern Kenyasince then, although a retrospective analysis indicates thatthere is sufficient historical, clinical and serological evidenceto suggest that outbreaks of what could have been rinderpestof varying severity did occur in the livestock of ethnic Somaliherders. As a result, the authors postulated that the Africanlineage 2 rinderpest virus might have survived in Somaliherders’ cattle as a pathogen of lower virulence for cattle, butoccasionally causing severe, fatal epidemics in wildlife – ahypothesis which identified cattle as the reservoir host andwildlife as indicator hosts.

The mechanisms that allowed lineage 2 viruses to escapedetection in eastern Africa for more than 30 years need to beinvestigated in the context of the effort to eradicate rinderpestfrom Africa – the Pan African Rinderpest Campaign (PARC)and its successor the Pan African Programme for the Controlof Epizootics (PACE) – executed by the Organization ofAfrican Unity’s Inter-African Bureau for Animal Resources,with European Community funding. This paper describes theapplication of participatory disease search methods, based onestablished participatory rural appraisal (PRA) techniques(McCracken and others 1988), in combination with strate-gic laboratory diagnostic support, to clarifying the epidemi-ology of lineage 2 rinderpest in eastern Africa. The objectiveof the study was to collate historical and current epidemio-logical information about rinderpest-compatible diseaseevents, in order to test the hypothesis that lineage 2 rinder-pest virus persisted during that period in populations oftranshumant cattle in the Somali ethnic areas of easternAfrica. A rinderpest-compatible disease event is one in whichthe herd develops clinical and epidemiological signs consis-tent with rinderpest, either the severe classical disease ormilder forms.

Use of participatory epidemiology in studiesof the persistence of lineage 2 rinderpestvirus in East AfricaJ. C. Mariner, P. L. Roeder

In 1994, rinderpest virus of African lineage 2 was detected in East Africa after an apparent absence ofmore than 30 years. In 1996, a disease search, based on participatory epidemiological techniquessupplemented by serological and virological analyses, was undertaken in southern Somalia and north-eastern Kenya to collate past and current epidemiological information about rinderpest-compatibledisease events, and to test the hypothesis that African lineage 2 rinderpest virus persists in populationsof transhumant cattle in the Somali ethnic areas. The findings in Afmadu in Lower Juba led the search forrinderpest to the communities in the Bardera area and then on to the Kenya/Somalia border areasbetween Mandera and El Wak. The herders had a specific knowledge of the clinical signs of rinderpestand provided detailed and accurate descriptions of cases. They differentiated between classical acuterinderpest and a milder syndrome characterised by an ocular discharge and diarrhoea, few oral lesions,corneal opacity and occasional mortality. The studies provided evidence for the endemic occurrence ofrinderpest back to at least 1981, with a periodicity of five years in the incidence of the disease. After aperiod of high mortality in 1992 to 1993, around Afmadu, herders reported a mild disease, withoccasional increases in mortality, from other areas of Lower Juba and the Gedo Region. Reports byherders of a rinderpest-compatible disease in the El Wak area were pursued until active cases werelocated and rinderpest was confirmed.

Veterinary Record (2003)152, 641-647

J. C. Mariner, DVM,RDP Livestock Services,PO Box 523, 3700AM Zeist,The NetherlandsP. L. Roeder, BVetMed,MSc, PhD, MRCVS,GREP Secretary, FAO

Animal Health Service,Animal Production andHealth Division, Vialledelle Terme di Caracalla,00100 Rome, Italy

642 The Veterinary Record, May 24, 2003

MATERIALS AND METHODS

Preliminary investigationsVeterinarians from the International Committee of the RedCross (ICRC) assessing the crises in Somalia in 1992 con-ducted open-ended interviews to establish animal healthpriorities for relief efforts. Without introducing the subjectof rinderpest, livestock herders were asked to name anddescribe the animal health problems affecting their herds.After the results of the interviews had been analysed, sero-logical investigations for rinderpest were undertaken insouthern Somalia, especially the Lower Juba region, inunvaccinated yearling calves. The herders were asked to cat-egorise the time that had elapsed since the last rinderpestcases as either less than four months, more than fourmonths or more than one year. The positive predictive value,that is, the probability that an animal that tests positive istruly positive as measured by a standard test, and the asso-ciated 95 per cent confidence intervals for the herders’ diag-nosis of rinderpest compared with a serological standardwere calculated by the methods described by Thrusfield(1995).

Participatory epidemiological investigationsA more extensive search for evidence of current and histor-ical rinderpest-compatible events was undertaken in 1996by means of a method based upon participatory ruralappraisal and rapid rural appraisal techniques (McCrackenand others 1988, International Institute for Environmentand Development 1994, Catley 1997, 1999, Mariner andPaskin 2000). Over a period of four weeks, interviews wereconducted with individuals, with small groups and in twocases as part of a community meeting of elders. Key infor-mants were identified as veterinary staff, community lead-ers, technical assistants and decision makers. The specifictechniques used were: semi-structured interviews, usingchecklists of open-ended questions which asked the respon-dent to formulate an answer in his own words rather thanselect from a set of structured responses; probing, by ask-ing specific questions designed to obtain more detailedinformation related to a statement received in response toan open-ended question, both to collect more specific infor-mation and to test the information for internal consistency;timeline construction, that is, the ordering of key events inrelation to the local traditional calendar; triangulation(cross-checking of reports), that is, confirming informationby using several data sources (types of respondents and bio-logical samples) and several methods (interviews, timelines,maps, serology, and antigen detection) (Moris andCopestake 1993); and mapping, that is, asking the respon-dents to display their information in relation to known landmarks.

In addition to triangulation, the Autoregressive IntegratedMoving Average (ARIMA) time series model (SAS Institute1993), was used to test for significant patterns and periodic-ity in the aggregate timeline derived from the tabulated datafrom the three interview areas and the individual timelinesfor each of the three interview areas. The null hypothesis forthe ARIMA model was that the data constituted a collection ofrandom observations lacking a discernible pattern, referredto as ‘white noise’.

Sampling for diagnostic confirmationSerum samples were collected from cattle at sites in LowerJuba during the 1992 investigation. In 1996, when mildrinderpest-compatible field cases were encountered at Fino,Hashino, Alunga and Warengara, ocular and nasal swabs werealso collected from the cattle. In addition, serum samples werecollected from 80 goats in communities near the outbreaks,but not in direct contact with affected cattle.

Virological techniquesRinderpest virus detection Swabs were tested by the agargel immunodiffusion test (AGIDT) at the Kenya AgriculturalResearch Institute, National Veterinary Research Laboratory,Muguga, Kenya (the Muguga Laboratory).

Rinderpest serology The cattle sera collected in 1992 werescreened at the Muguga Laboratory in a microtitre virus neu-tralisation test (VNT) (Mariner and others 1990) at a finaldilution of 1/20 and positive samples were titrated in a two-fold dilution series. Goat sera from 1996 were tested at theWorld Reference Laboratory for Rinderpest by a competitiveELISA. All the tests were carried out as described by Andersonand others (1996).

RESULTS

The 1992 investigationsThe open-ended interviews conducted by the ICRC began withthe identification of the respondent and a general questionabout animal health problems in his cattle. In order to avoidbias, rinderpest was not mentioned until the subject wasintroduced by the respondents. Whenever a respondent intro-duced the subject of rinderpest he was asked to describe thedisease, and if he could not describe it accurately, the reportwas eliminated. The accounts were analysed for specificityand internal consistency, and cross-checked with other inde-pendent reports. Only specific reports which gave dates andlocations, and indicated contact herds by name or commu-nity, were used. Vague statements that were not supported byspecific accounts of contacts with affected animals and clin-ical descriptions were discarded.

Reports that met the above criteria were divided into fourcategories: (i) a first-hand report from a herder concerningmorbidity or mortality due to rinderpest in his own herd; (ii)a first-hand report of clinical cases observed by a veterinarianor veterinary assistant; (iii) a report of cases directly observedby a cattle herder; and (iv) a second-hand report or hearsayfrom veterinarians, herders, public officials or elders who didnot actually see the disease. This category was used to assistthe team in selecting locations for further study, but the infor-mation was not used in the construction of maps and time-lines.

The interviewers found that cattle herders in the LowerJuba locations of Afmadu and Kolbio consistently reportedrinderpest to be a major cause of mortality at the time of theinterviews. Serum samples were collected from 83, 10- to 24-month-old calves from herds recently affected by rinderpest.The herders reported that the calves had not been vaccinatedagainst rinderpest and no rinderpest vaccination markingswere evident. The VNT results are shown in Table 1; 23 of the83 calves were seropositive for rinderpest. They came fromfour locations: Afmadu, Kogani, Tapta and Badhaade.

Papers & Articles

Number (%) of positiveLocation Number of sera Number of herds samples at 1:20 Final titres

Afmadu 22 9 10 (45) 1:32-1:512Abadjo 5 1 0Kogani 5 2 3 (60) 1:64Tapta 26 5 6 (23) 1:20-1:256Kudha 6 3 0Badhaade 9 2 4 (44) 1:20-1:256Bulo Hadji 5 1 0Abdulah Birole 2 1 0Janale 3 1 0Total 83 25 23 (28)

TABLE 1: Results of rinderpest virus neutralisation tests on sera collected in 1992 from cattleunder two years of age in the Lower Juba area

The Veterinary Record, May 24, 2003 643

Afmadu, the site with a seroprevalence of 45 per cent, is morethan 125 km from the Somali-Kenyan border (Fig 1). Therelationship between the herders’ reports and positive serol-ogy is shown in Table 2. The positive predictive value of theherders’ reports of rinderpest ranged between 31 and 60 percent, being higher the shorter the time that had elapsed sincethe reported outbreak; however, this trend was not statisticallysignificant. The aggregate positive predictive value of theherders’ reports was 39 per cent.

The 1996 disease searchGeneral findings The search began at Afmadu in the LowerJuba region which was the site of the most extensive rinder-pest-compatible event known to have occurred in cattle in theregion. Tracing the outbreak forward and backward in timefrom Afmadu led the search to the communities in theBardera area and border areas between Mandera and El Wak– all Somali ethnic areas. The sites of interviews are shown

in a sketch map of the Ethiopia, Kenya and Somalia borderarea (Fig 1). The individual reports in categories (i) to (iii)were tabulated and mapped to build a general view of the pastand recent incidence of rinderpest in the region. Building anagreed temporal and spatial view was a major component ofthe triangulation process, in that the construction of maps,tables and timelines made it easy to determine whether thereports were consistent or contradictory. The annual rinder-pest reports from 1980 to 1996 are shown in Table 3 for thethree areas Afmadu-Marere, Bardera and Mandera-El Wak,and the periodicity of the reports is illustrated in Fig 2. Table3 shows the numbers of respondents that accurately describedrinderpest and gave internally consistent accounts of havingpersonally observed severe rinderpest, mild rinderpest or arinderpest-like disease (a rinderpest-compatible event) ineach year. The ARIMA model rejected the null hypothesis thatthe aggregate timeline represented white noise (P<0·01) andfound that the disease occurred in an approximately five-yearcycle (Fig 2).

In total, 194 adult cattle herders, officials and veterinarystaff were interviewed in 66 interviews, giving an average of2·9 respondents per interview. Individuals often reported sev-eral events or an incident spanning several years, but 16 didnot report any occurrence of rinderpest. Fifteen of these werein three interview groups in the Bardera area and the otherwas a man in his early twenties interviewed with his fathernear Mandera. As these 16 did not make a report, they werenot requested to describe rinderpest and it was not knownwhether they could describe the disease. In the Afmadu-Marere group there were 12 interviews; 46 respondents wereinterviewed at 10 of them which were held in the area imme-diately surrounding Afmadu, and three respondents wereinterviewed at two interviews conducted at Marere on theJuba River about 90 km to the east of Afmadu. In theMandera-El Wak group there were 37 interviews; 97 respon-dents were interviewed at places where El Wak was the clos-est major town, and seven respondents were interviewed atfive interviews held at places closer to Mandera, along bothsides of the Kenya-Somalia border.

Almost all the livestock herders and other key informants,regardless of their age, could describe rinderpest accurately.Only two respondents who gave reports of rinderpest-compatible events gave contradictory or inaccurate descrip-tions of the disease. Of 66 interview groups, 65 (98·4 per cent)gave accurate descriptions of rinderpest. The clinical signsmentioned in order of precedence were an ocular discharge,diarrhoea, death, oral lesions, and in some cases skin lesions.Young animals were much more likely to be affected. Themost popular local name in Somali was ‘shifow’, which refersto the ocular discharge. Other names that specifically refer torinderpest were ‘madobeye’ (Somali) and ‘daadi’ (Gurreh).Non-specific Somali terms associated with rinderpest signswere ‘el waren’ and ‘hardik’ that refer to ‘a poke in the eye’ and‘bloody diarrhoea’, respectively.

Afmadu-Marere area The descriptions of rinderpest atAfmadu, and of classical, historic disease in other areasincluded death as a key component. In the Afmadu area, the46 respondents stated that rinderpest had been a cause of highmortality in 1992. Thirty-four of them stated that cases beganto appear in December 1991 and one reported that the out-break persisted into early 1994. They all stated that rinderpestwas not causing problems at the time of the interviews inFebruary 1996. The reports stated that the disease was trav-elling from the west to the east and was introduced throughcontact with cattle of the Mohamet Zubir clan coming fromthe west as a result of an extensive drought. Subsequently,three interviews were conducted at Marere on the Juba River.The respondents stated that they were aware of the rinderpestbetween Liboi and Afmadu, but that the disease had not

Papers & Articles

Time since the Number of Number of Positiveoutbreak of rinderpest Number of seropositive seronegative predictive valuein 1992 herds* herds herds (95 per cent CI)

More than one year 13 4 9 31 (6, 56)Within one year 5 2 3 40 (0, 83)Within four months 5 3 2 60 (20, 100)Total 23 9 14 39 (19, 59)

* Herders of two herds were not asked to state the time since the last outbreakCI Confidence interval

TABLE 2: Positive predictive value of herders’ reports of rinderpest in relation to the timesince the outbreak of rinderpest in 1992

FIG 1: Sketch map of the Kenya, Somalia and Ethiopia border area, showing the majortowns and selected locations referred to in herders’ reports. The years and areas reportedto be affected by rinderpest are indicated by dates; stars indicate the exact location ofoutbreaks of the disease

ETHIOPIA

SOMALIA

Mandera

Bardera

Afmadu

El Wak

BAKOOL

BAY

LOWER JUBA

KENYA

Fino

Hashino

Fafadun

Sidimo

Kismayo

Marere

Badhaade

Wajir

Liboi

Garissa

i

i

i

i

i

�

�

�

�

�

�

�

�

�

�

�� �

✪

✪✪

1991-96

1981-83

1994-96

1981-83 1986-88

1981, 1985-871991-93

Juba River

Shebelli River

Confirmed outbreaks

Linked outbreaks

Towns

Interview area centres

1981-83 Years of rinderpest reports

100 km

644 The Veterinary Record, May 24, 2003

reached their communities. The last outbreak of rinderpestmentioned at Marere was in 1987.

Bardera area There were first-hand reports of relativelyrecent rinderpest-compatible events in the El Wak area.Transhumant informants interviewed along the Juba riverstated that they had encountered rinderpest of moderatemortality at a well (Fafadun) in the direction of El Wak andin grazing areas to the west and north of it in Somalia in theyears 1994 and 1995. Eight former veterinary staff members,seven representing Bardera and one representing El Wak, wereinterviewed at Bardera as a group. They had been assignedto various parts of Somalia before the civil disturbances of1990. Seven of them expressed the opinion that rinderpesthad been present in Somalia in recent years, but the eighthstated categorically that classical severe rinderpest had lastoccurred in Somalia in 1974. When the seven respondentswere asked to substantiate their view, two gave first-handreports (category ii) of having observed rinderpest directly,the first in the Badhaade and Huluaqa area in the extremesouth in 1985, and the other at El Wak in 1994/95.

Sedentary and nomadic cattle herders along the JubaRiver 20 km to the north and south of Bardera gave first-hand reports of rinderpest for the years 1986, 1981 and 1974.These reports cross-checked with others from the area andprevious reports obtained at Afmadu. There were no first-hand reports of active or recent disease relating to theBardera area itself.

Three first-hand reports (two category i and one categoryii) of recent rinderpest-compatible events that occurred in theEl Wak area, and which cross-checked, were received duringinterviews conducted in the Bardera area. Other detailed second-hand reports of rinderpest in the El Wak area werereceived from herders and from the governor of El Wak,Somalia.

Mandera-El Wak area At Mandera and in its vicinity, anextensive outbreak of rinderpest with high mortality wasreported to have occurred between 1981 and 1983. All theMandera respondents who were of a mature age in 1981/83described outbreaks of severe rinderpest.

In both the Bardera and the Mandera-El Wak areas,descriptions of a recent disease syndrome highlighted diar-rhoea and an ocular discharge. Some livestock herdersreferred to this disease as ‘shifow’ (rinderpest) whereas otherscategorised the problem as ‘el waren’ or ‘hardik’. When thisissue was questioned at El Wak and the nearby towns ofGarsal and Warengara, the consensus view of elders, local vet-erinary personnel and livestock herders was that the presentdisease might be rinderpest (‘shifow’, ‘madobeye’ or ‘daadi’).However, because it did not kill the affected animals, theyfound it difficult to make a definitive statement. Some respon-dents described the disease in their own language as ‘mildrinderpest’ or ‘rinderpest-like’.

As the search was carried south on both sides of the bor-der reports were received of the recent occurrence of mildcases of rinderpest (‘shifow’ or ‘madobeye’) or other rinder-pest-compatible disease (‘el waren’ or ‘hardik’). At Fino, Kenya(N 3°22·888′, E 41°24·851′) on April 14, 1996, livestockherders reported an ocular discharge and diarrhoea in youngcalves less than one year old. Of three calves examined, allthree had marked epiphora, one had a unilateral cornealopacity, one had diarrhoea and one had slight oral erosions.Four more cases were encountered later at Hashino, a grazingarea about 10 km distant (N 3°27·196′, E 41°19·110′). All fourhad epiphora (Fig 3), three of them had oral erosions or ery-thematous lesions on the muzzle, lips and medial canthus,and one had bloody diarrhoea. Ocular swabs from one of theanimals at Fino and one of those at Hashino were positive forrinderpest antigen in the AGIDT.

Subsequently, animals were observed with marked oculardischarges, unilateral corneal opacity, sporadic oral erosions,and diarrhoea in youngstock and occasional adult animals, inone herd at Lafey, two at Fino, two at Alunga (N 3°02·371′, E41°02·982′) and two herds at Warengara (N 3°24·498′, E41°00·567′). The distribution of the clinically affected herdsis shown in Fig 1.

DISCUSSION

The identification of the cause of the epidemic of rinderpestin wildlife in Tsavo National Park in 1994 as African lineage2 rinderpest virus indicated that the source was not linked toknown foci of rinderpest persistence, and emphasised theneed for a specific search for rinderpest disease to define theorigin of the virus. This search needed a reappraisal of all theclinical events that could be suggestive of rinderpest, espe-cially the mild form. The results presented here represent theinitial findings of a continuing investigation. The extensiveand repeated reports of mild disease in cattle, with occasionalmoderate to severe epidemic flare-ups, linked to supporting

Papers & Articles

Year Mandera-El Wak Afmadu-Marere Bardera Total

1996 65 0 1 661995 6 0 4 101994 4 1 1 61993 4 41 1 461992 4 45 0 491991 2 34 0 361990 0 0 5 51989 5 0 0 51988 7 0 0 71987 12 34 3 491986 4 0 8 121985 0 0 1 11984 0 0 0 01983 3 0 1 41982 12 0 0 121981 16 1 3 201980 2 0 0 2Number of interviews 37 12 17 66Number of respondents 104 49 41 194

TABLE 3: Reports of the occurrence of rinderpest-compatible events by year and area

FIG 2: Annual numbers of respondents reporting rinderpest-compatible events in theBardera, Afmadu-Marere and Mandera-El Wak areas between 1980 and 1996

Num

ber

of r

espo

nden

ts

70

60

50

40

30

20

10

0

19801981 1982 19831984 19851986 1987 1988 1989 1990 19911992 1993 19941995 1996Year

BarderaAfmadu-MarereMandera-El Wak

The Veterinary Record, May 24, 2003 645

serology and antigen detection indicates that the Somali eth-nic areas visited have contributed to the maintenance ofrinderpest throughout at least the 1980s and 1990s. The moresevere waves of disease, such as that at Afmadu in 1992,appear to have been associated with periods of stress andincreased mobility due to drought. As no viral genetic mate-rial was isolated, it cannot be stated that the virus is of thesame lineage as that which caused the serious outbreaks inwildlife in Kenya. However, the apparently endemic nature ofthe mild, sometimes moderate, rinderpest in southern Somalilivestock, combined with the occurrence of more readilydetectable epidemics in wildlife should guide rinderpest con-trol strategies in East Africa.

In the present study there were clusters of rinderpest-compatible events at four main periods and in two mainareas. These were 1981/83, 1985/88 and 1991/93 in theAfmadu area, and 1991/96 in the El Wak area. No reports werereceived from the areas visited for the year 1984, and 1985 wasmentioned only by one respondent. The year 1988 was men-tioned by two independent groups of respondents at onelocation: Garsal, Somalia, near Bardera. The focus of activityshifted from these southern areas in 1985/87 to the morenortherly area around El Wak in 1986/88. The peak year forreports from both the Afmadu and El Wak respondents was1987, as opposed to 1986 in Bardera. The ARIMA time-seriesmodel first determines whether there is a significant patternover time and then determines the periodicity of the pattern.It then corrects for the periodicity detected, and examines thedata for long-term trends such as an increase or decrease inprevalence. Assuming that the conditions remain essentiallyunchanged, the model can be used to predict future valuesin the data series on the basis of the previous pattern. Thetime-series analysis revealed a highly significant aggregatepattern of reports with an approximate periodicity of fiveyears, indicating that waves of rinderpest can be expected tooccur with this periodicity, provided that background epi-demiological factors such as partial or sporadic coverage byvaccination and climatic conditions remain unchanged. Thestatistical significance of the pattern of the rinderpest time-line and the pattern’s conformity to basic epidemiologicalconcepts indicates that the herders were providing genuinedescriptions of actual disease events.

That the 1991/93 cluster of events in the Lower Juba(Afmadu) area was truly rinderpest is supported by the preva-lence of antibodies in unvaccinated young cattle. No animalshad been vaccinated in this region of Somalia for two yearsbefore the samples were collected, owing to the severe civildisturbances. Residual maternal immunity is unlikely to have

accounted for more than a small fraction of the seropositiveanimals, because the youngest animals sampled were 10months old and the titres ranged up to 1/512. On the basisof contemporary herders’ reports and serological evidence, anaggressive rinderpest vaccination campaign was undertakenby the humanitarian relief effort in Somalia in 1993/94, afterwhich the respondents at Afmadu reported no further rinder-pest-compatible events.

The trail of events that led to the 1991/96 cluster of reportsin the El Wak area was pursued until active cases for labora-tory diagnosis were located. Informants from as far away asBardera indicated that the area north of El Wak was a focusof active rinderpest, and laboratory confirmation of rinder-pest in two affected herds substantiated the herders’ reports.The fact that all the 80 goats were seronegative is of no sig-nificance because small ruminants do not appear to be goodsentinels for mild rinderpest in cattle. The results agree withthe results of serosurveillance in northern Tanzania in 1997where less than 2 per cent of 1627 goats were seropositivewhereas 15 per cent of 1261 cattle in the same villages hadseroconverted to the field infection (Taylor and others 2002).

The results of the present study are fully applicable only tothe areas that the study team visited. Although the respon-dents gave some information about other areas where theyhad grazed or about herders with whom they had had con-tact, the evidential link is too weak to draw conclusions.However, the information suggested that there was a broaderarea of non-stable endemicity and provided ‘leads’ for furtherepidemiological research and investigation. The rinderpestdetected at Fino and Hashino was not the end of the trail.Further field studies should be carried out in adjacent areas,particularly to the south and between the Juba and ShebelliRivers.

This study was conducted to help to formulate eradica-tion strategies, and it is therefore important to relate theinformation obtained to other published and unpublishedrinderpest disease events of epidemiological significance. Atthe outset of Joint Project 15, a previous internationalrinderpest eradication effort, Atang and Plowright (1969)observed that three confirmed outbreaks of rinderpest hadoccurred in eastern and north-eastern provinces of Kenya in1967 and that the disease was ‘more or less continuously pre-sent’. Since then, the published record occasionally depictsrinderpest ‘footprints’ in the region. Before the 1994 KenyanTsavo outbreak, the last confirmed case of rinderpest in east-ern Kenya had been in an eland in Kinna, Eastern Provincein 1973 (Rossiter and others 1982). The same paper pre-sented evidence of rinderpest antibody in 35 of 425 samplesof serum collected from Kenyan wildlife during the 1970s,and suggested that the undetected circulation of virus wasthe most plausible explanation. Wafula and others (1982)reported that seven of 28 waterbuck and four of 10 impalafrom the Nakuru and Naivasha area of the central Rift Valleyof Kenya were seropositive, and that seven of 55 oryx fromGalana in eastern Kenya were seropositive. The paper doesnot give information about the age of the animals or col-lection dates, but postulates that there may have been anunrecognised morbillivirus circulating in the region.Frequent but limited vaccination programmes limit thevalue of cattle serology in the detection of the virus in theSomali ethnic areas of Kenya.

A major epizootic of rinderpest occurred in 1981/83 innorthern Tanzania (Rossiter and others 1983). The source ofthis outbreak has never been satisfactorily defined. Kenya hasbeen suggested as the source (Nyange and others 1985), butit is necessary to rule out the occult persistence of virus inTanzania during the period leading up to the 1981/83 out-break (Nyange and others 1985). Extensive serological evi-dence of the circulation of rinderpest virus in populations ofsmall ruminants and buffalo in northern Tanzania during the

Papers & Articles

FIG 3: Epiphora in anotherwise clinicallynormal calf infectedwith rinderpest virus

646 The Veterinary Record, May 24, 2003

mid and late 1980s was reported by Anderson and others(1990).

Corresponding to the 1981/83 cluster of events in easternAfrica, there were reports from Oman in April 1983 to theOffice International des Epizooties (OIE) concerning thearrival of a dhow from Somalia with eight clinical cases ofrinderpest aboard, one of which died (OIE 1983a) and thesereports were followed by related, confirmed outbreaks on themainland (OIE 1983b, S. Routledge, personal communica-tion). Anecdotal reports which cannot now be substantiatedsuggest that there were two later instances (in 1988 and 1992)of clinically affected cattle arriving in the Yemeni ports ofHodaidah and Aden, purportedly originating from easternAfrica (M. Al Qadassi, personal communication), and initi-ating local outbreaks. It is perhaps significant that in thisstudy rinderpest-compatible events were detected in boththese years, and particularly in 1992. The most recent episodeof rinderpest in trade cattle was reported through the GulfCooperation Council in 1998 and involved the refusal of entryto Dubai (UAE) of clinically affected cattle described as orig-inating from Kismayo (Roeder 1998). Rinderpest viruses orig-inating from the Arabian Peninsula are poorly representedin the archive, but none of the five, not even the two fromYemen isolated in 1981 and 1995, was of African lineage.

The authors know of no published data from Somaliabetween the last reported outbreak in 1974 and the present.Government archives are almost certainly no longer available,and for the more northerly Somali communities in Ethiopiaand Somalia, information is very limited. However, all theavailable evidence suggests that rinderpest virus has not cir-culated there for many years. In this study, the respondentsclearly identified the Ogaden as a source of infection in 1981,but the north was not mentioned in any subsequent years. Inthe studies made by Catley and Mohammed (1995, 1996) inthe Sanaag region of Somalia, references to rinderpest werestrictly historical. One of the present authors (J. C. M.) madea survey in the Berbera and Hargeisa areas of Somalia inJanuary 1993 and found no evidence of active or recentrinderpest. In the last five years, the PARC has made a num-ber of local serological studies, collecting thousands of sam-ples in the Ethiopian Ogaden around Harer and as far southas Gode. These have been uniformly negative (A. Gopilo, per-sonal communication) as have the results of rinderpest sur-veillance activities in southern Ethiopia (A. Gopilo, G. van’tKlooster, personal communication). It therefore seems prob-able that these areas did not contribute to the rinderpest reser-voir at that time.

The importance of the Somali cattle population in themaintenance of lineage 2 virus was discussed by Barrett andothers (1998). Atang and Plowright (1969) divided rinderpestin East Africa into epidemiologically distinct northern andsouthern foci. The northern focus was characterised by ahighly fatal disease caused by a virus of the RGK1 strain(Plowright 1982) which causes about 60 per cent mortality(Liess and Plowright 1964); it was isolated during an epidemicin the Tana River area (Stewart 1964). The southern focusinvolved both cattle and wildlife but was characterised by amild disease in cattle with transitory diarrhoea and mouthlesions detected only with difficulty; the RBT1 strain was saidto be representative of the southern focus. Both the RBT1

strain and RGK1 strain are now known to be members of lin-eage 2 and this is further evidence of the variable patho-genicity of the virus. With respect to the mild forms ofrinderpest, Atang and Plowright (1969) stated that ‘it is hardlylikely that such a benign and immunising disease would bereported by the stockherders: it would have to be searched forby persistent and careful observers’.

Disease surveillance and the design of animal health inter-ventions is increasingly supported by the collection of animalhealth information through participatory appraisals (Inter-

national Institute for Environment and Development 1994,Catley 1999, Mariner and Paskin 2000). Southern Somalirespondents, some as young as 15 years old, gave detailed andaccurate descriptions of rinderpest, having developed thisknowledge through first-hand experience. The clinical term‘shifow’ in its strictest usage was compatible with classical vir-ulent rinderpest and included death as part of the description.The association of the syndrome including an ocular dis-charge and occasional or transitory oral lesions and diarrhoeawith ‘shifow’, combined with some individuals’ reluctance todraw definitive conclusions, indicated a sophisticated andprudent approach to clinical diagnosis. There was a relation-ship between herders’ reports of rinderpest collected in 1992and the serological findings on the associated sera. The pos-itive predictive value of herders’ reports in this endemic set-ting indicates that they are epidemiologically important toolsfor the detection of rinderpest. The positive predictive valuefor the aggregated data suggests that in the Somali setting fivereports of rinderpest from independent but epidemiologicallyrelated sources are diagnostic of rinderpest 90 per cent of thetime (P=1–[0·61]5). To put these values in perspective, stan-dard clinical case definitions for pertussis used by physicianshave a positive predictive value of about 70 per cent (Patriarcaand others 1988).

Unilateral corneal opacity in cattle was associated with theclinical syndrome detected in Mandera. Corneal opacity wasalso observed in affected cattle during investigations of theoutbreak of rinderpest in Tanzania in 1997, and ocular lesionshave been observed recently in African buffalo calves in Kenya(Kock and others 1999, Rossiter 2000). However, no causalrelationship has been established, and it is therefore notknown whether the corneal opacity represents a primaryrinderpest lesion or is the result of a secondary infection, pos-sibly an exacerbation of an intercurrent infection (Scott1964). It has long been known that rinderpest can cause bilat-eral corneal opacity and blindness as prominent clinical signsin kudus and giraffes, but it has been stated categorically notto occur in cattle (Plowright 1982). It would seem prudentto include corneal opacity in the case description of rinder-pest for surveillance programmes in East Africa until moredefinitive information is available; at worst it would slightlyincrease the size of the suspicious category.

The severity of disease and the pathogenicity of specificviral strains in individual species observed during naturaloutbreaks and experimental infections are variable charac-teristics (Plowright 1963, Branagan and Hammond 1965,Anderson and others 1990). Furthermore, the rinderpestvirus does not spread between susceptible species in all fieldoutbreaks (Taylor and Watson 1967). These observations sug-gest that there are partial barriers to its transmission betweensusceptible species, and that a combination of circumstancesis required for at least some rinderpest strains to spreadbetween cattle and wildlife or between wildlife species. As aresult, field outbreaks of rinderpest vary in severity and ofteninvolve only a subset of the susceptible species.

Given the progress in the worldwide eradication of rinder-pest, it is essential that rinderpest surveillance techniquesshould be sensitive enough to detect all the foci involved inthe circulation of rinderpest virus, whether or not they areaccompanied by severe clinical disease. It is becoming clearthat remote and marginalised pastoral communities haveserved as the final reservoirs of rinderpest in Africa (Mariner1996). Owing to its high cost, the lack of access and slowtransport, there is scarcity of laboratory-based data on theepidemiology of rinderpest in remote areas. As a result, tra-ditional laboratory data do not give an adequate account ofthe epidemiological picture and may be misinterpreted. Theresults of this study illustrate the significant contribution thatparticipatory epidemiology, supported by laboratory testing,can make to the clarification of rinderpest epidemiology, by

Papers & Articles

The Veterinary Record, May 24, 2003 647(c) British Veterinary Association. All rights reserved

providing epidemiological information to bridge the gapsbetween events detected by more conventional means.

To be effective in the final stages of the disease’s eradica-tion rinderpest surveillance must be sensitive, specific andtimely. The surveillance system should incorporate tech-niques to meet these three requirements. Methods of search-ing for the disease are needed that detect a high percentageof rinderpest-compatible events (stomatitis-enteritis out-breaks) for further investigation (Mariner and others 1998).The participatory epidemiological approach is a powerful andefficient field technique that warrants adoption by the GlobalRinderpest Eradication Programme to compliment the con-ventional surveillance methods.

ACKNOWLEDGEMENTS

The authors thank the Director and staff of the KenyanVeterinary Service together with Dr Walter Masiga, thenDirector of OAU IBAR, and the staff of the Pan-AfricanRinderpest Campaign Coordination Unit, Nairobi, Kenya,who were partners in mounting this study which was fundedby the Food and Agriculture Organization of the UnitedNations. Invaluable information provided by Hans-PeterGiess, Sangeta Koenig and Giles Sandre of the InternationalCommittee of the Red Cross about early events is includedwith their kind permission. The valuable assistance of DrHenry Wamwayi of the Kenya Agricultural Research InstituteNational Veterinary Research Laboratory, Muguga, Kenya forlaboratory analyses is gratefully acknowledged. The assistanceof Willard C. Losinger with time-series analysis of the time-line data is highly appreciated as is Dr Vincent Martin’s valu-able assistance in preparing the sketch map.

ReferencesANDERSON, E. C., JAGO, M., MLENGEYA, T., TIMMS, C., PAYNE, A. &

HIRJI, K. (1990) A serological survey of rinderpest antibody in wildlife andsheep and goats in northern Tanzania. Epidemiology and Infection 105, 203-214

ANDERSON, J., BARRETT, T. & SCOTT, G. R. (1996) Manual on the Diagnosisof Rinderpest. FAO Animal Health Manual 1. Rome, Food and AgricultureOrganization

ATANG, P. & PLOWRIGHT, W. (1969) Extension of the JP-15 rinderpest con-trol campaign to Eastern Africa: the epizootiological background. Bulletinof Epizootic Diseases in Africa 17, 161-170

BARRETT, T., FORSYTH, M. A., INUI, K., WAMWAYI, H. M., KOCK, R.,WAMBUA, J., MWANZIA, J. & ROSSITER, P. B. (1998) Rediscovery of thesecond African lineage of rinderpest virus: its epidemiological significance.Veterinary Record 142, 669-671

BRANAGAN, D. & HAMMOND, J. A. (1965) Rinderpest in Tanzania: a review.Bulletin of Epizootic Diseases in Africa 13, 225-246

CATLEY, A. (1997) Adapting participatory appraisal (PA) for the veterinary epi-demiologist: PA tools for use in livestock disease data collection. Proceedingsof the Society for Veterinary Epidemiology and Preventive Medicine. Chester,UK, April 9 to 11, 1997. pp 246-257

CATLEY, A. (1999) Participatory Approaches to Veterinary Epidemiology:Methods on the Move. London, Sustainable Agriculture and RuralLivelihoods Programme, IIED

CATLEY, A. & MOHAMMED, A. A. (1995) Ethnoveterinary knowledge inSanaag region, Somaliland: notes on local descriptions of livestock diseasesand parasites. Nomadic Peoples 36/37, 3-16

CATLEY, A. P. & MOHAMMED, A. A. (1996) The use of livestock-disease scor-ing by a primary animal-health programme in Somaliland. PreventiveVeterinary Medicine 23, 175-186

INTERNATIONAL INSTITUTE FOR ENVIRONMENT AND DEVELOP-MENT (1994) Special issue on livestock. RRA Notes 20, 1-172

KOCK, R. A., WAMBUA, J. M., MWANZIA, J., WAMWAYI, H., NDUNGU,E. K., BARRETT, T., KOCK, N. D. & ROSS, P. B. (1999) Rinderpest epidemicsin wild ruminants in Kenya 1993-97. Veterinary Record 145, 275-283

LIESS, B. & PLOWRIGHT, W. (1964) Studies on the pathogenesis of rinderpestin experimental cattle. I. Correlation of clinical signs, viraemia and virusexcretion by various routes. Journal of Hygiene, Cambridge 62, 81-100

MARINER, J. C. (1996) The world without rinderpest: outreach to margin-

alised communities. In Food and Agriculture Organization AnimalProduction and Health Paper 129. Rome, Food and Agriculture Organization.pp 97-107

MARINER, J. C. & PASKIN, R. (2000) Manual on Participatory Epidemiology.FAO Animal Health Manual 10. Rome, Food and Agriculture Organization

MARINER, J. C., VAN DEN ENDE, M. C., HOUSE, J. A., MEBUS, C. A.,SALIFOU, S. & STEM, C. (1990) The serological response to a thermostableVero cell-adapted vaccine under field conditions in Niger. VeterinaryMicrobiology 22, 119-127

MARINER, J. C., VAN’T KLOOSTER, G., GEIGER, R. & JEGGO, M. H. (1998)Rinderpest surveillance performance indicators. Proceedings of the TechnicalConsultation Meeting on the Global Rinderpest Eradication Programme, FAO

Rome, September 28-30. pp 130-140MCCRACKEN, J., PRETTY, J. & CONWAY, G. (1988) An Introduction to Rapid

Rural Appraisal for Agricultural Development. London, Institute forInternational Environment and Development

MORIS, J. & COPESTAKE, J. (1993) Qualitative Enquiry for RuralDevelopment. London, Intermediate Technology Publications

NAWATHE, D. R. & LAMORDE, A. G. (1983) Towards global eradication ofrinderpest. Revue Scientifique et Technique de l’Office International desEpizooties 2, 997-1011

NYANGE, J. F. C., OTARU, M. M. M. & MBISE, A. N. (1985) Incidences ofabnormal wild animal mortality in northern Tanzania 1979-82. Bulletin ofAnimal Health and Production in Africa 33, 55-57

OIE (1983a) Confirmed outbreak rinderpest on dhow ‘Yaqoobi’ ex Somalia.Epizootiological Information Number OM 83/2/88. Paris, OfficeInternational des Epizooties

OIE (1983b) Rinderpest outbreaks reported at Sumail. EpizootiologicalInformation Number OM 83/3/113. Paris, Office International des Epizooties

PATRIARCA, P. A., BIELLIK, R. J., SANDEN, G., BURSTYN, D. G., MITCHELL,P. D., SILVERMAN, P. R., DAVIS, J. P. & MANCLARK, C. R. (1988) Sensitivityand specificity of clinical case definitions for pertussis. American Journal ofPublic Health 78, 833-836

PLOWRIGHT, W. (1963) Some properties of strains of rinderpest virus recentlyisolated in E Africa. Research in Veterinary Science 4, 96-108

PLOWRIGHT, W. (1982) The effect of rinderpest and rinderpest control onwildlife in Africa. Animal diseases in relation to animal conservation.Symposium of the Zoological Society of London 50, 1-28

PLOWRIGHT, W. (1998) Professional commitment to the success of the GlobalRinderpest Eradication Programme. Proceedings of the FAO TechnicalConsultation on the Global Rinderpest Eradication Programme. FAO, Rome,September 28 to 30, 1998. pp 7-22

ROBSON, J., ARNOLD, R. M., PLOWRIGHT, W. & SCOTT, G. R. (1959) Theisolation from an eland of a strain of rinderpest virus attenuated for cattle.Bulletin of Epizootic Diseases in Africa 7, 97-102

ROEDER, P. L. (1998) Rinderpest problem areas: how to remove the virus fromits last strongholds: Lineage Asia. Proceedings of the FAO TechnicalConsultation on the Global Rinderpest Eradication Programme. FAO, Rome,September 28 to 30, 1998. pp 37-50

ROSSITER, P. (2000) Morbilliviral diseases: rinderpest. In Infectious Diseasesof Wild Mammals. Eds E. S. Williams, I. K. Barker. Ames, Iowa StateUniversity Press. pp 37-45

ROSSITER, P. B., JESSETT, D. M., WAFULA, J. S., KARSTAD, L., CHEMA, S.,TAYLOR, W. P., ROWE, L., NYANGE, H. C., OTARU, M., MUMBALA, M. &SCOTT, G. R. (1983) Re-emergence of rinderpest as a threat in East Africasince 1979. Veterinary Record 113, 459-461

ROSSITER, P. B., KARSTAD, L., JESSETT, D. M., YAMAMOTO, T., DARDIRI,A. H. & MUSHI, E. Z. (1982) Neutralising antibodies to rinderpest virus inwild animal sera collected in Kenya between 1970 and 1981. PreventiveVeterinary Medicine 1, 257-264

SAS INSTITUTE (1993) SAS/ETS User’s Guide, Version 6. 2nd edn. Cary, SAS

InstituteSCOTT, G. R. (1964) Rinderpest. Advances in Veterinary Science 9, 113-224STEWART, D. R. M. (1964) Rinderpest among wild animals in Kenya, 1960-2.

Bulletin of Epizootic Diseases in Africa 12, 39-42TAYLOR, W. P. & WATSON, R. M. (1967) Studies on the epizootiology of

rinderpest in blue wildebeest and other game species of Northern Tanzaniaand Southern Kenya, 1965-7. Journal of Hygiene, Cambridge 65, 537-545

TAYLOR, W. P., ROEDER, P. L., RWEYEMAMU, M. M., MELEWAS, J. N.,MAJUVA, P., KIMARO, R. T., MOLLEL, J. N., MTEI, B. J., WAMBURA, P.,ANDERSON, J., ROSSITER, P. B., KOCK, R., MELENGEYA, T. & VAN DENENDE, R. (2002) The control of rinderpest in Tanzania between 1997 and1998. Tropical Animal Health and Production 34, 471-487

THRUSFIELD, M. (1995) Veterinary Epidemiology. 2nd edn. London,Blackwell Science. pp 277-280

WAFULA, J. S., MUSHI, E. Z. & KARSTAD, L. (1982) Antibodies to rinderpestvirus in the sera of some wildlife in Kenya. Bulletin of Animal Health andProduction in Africa 30, 363-365

Papers & Articles