TRANSACVONS OFTHE ROYAL SOCIETY OFTROPICAL MEDICE’JEANtI HYGIENE (1997) 91,541-543 541

Epidemic dysentery in western Kenya*

M. A. Malakooti’, J. Alaii3, G. D. Shanks3,4 and l? A. l’hiips-Howard2,3 ‘Department of Preventive Medicine

and Biometrics, Uniformed Services University of the Health Sciences, Bethesda, Maryland, USA;2Division of Parasitic Dis-
eases,National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA;3Kenya
Medical Research Institute, Kisumu, Kenya; 4US Army Medical Research Unit, Kisumu, Kenya

Abstract
This paper describes the epidemiology of a probable Shigella dysenteriae type 1 dysentery epidemic in west-
ern Kenya. A retrospective record review over 2 years of all cases of dysentery, amoebiasis and diarrhoea
was carried out in 13 healthcare facilities in the Rarieda Division of Nyanza province. Of the 3301 cases
recorded, 2191 were dysentery, giving a cumulative 2 years incidence rate for dysentery of 4%.The epi-
demic began in December 1994 and peaked in February 1995, coinciding with the very dry season. One
location in the area had an overall attack rate of 9*3%, double that of other locations. Highest rates were
in children aged ~5 years and in persons >15 years old. S. dysenteriae type 1, with its increasing multi-
antibiotic resistance, is a continuing threat to the health of people in this region; this area may be suitable
for intensive, prospective surveillance as a prelude to a Shigella vaccine trial.
Keywords: dysentery, ShigeZZa
dysenteriue,epidemic, incidence, Kenya

diarrhoea from June 1994 to May 1996, in the Rarieda

A dysentery pandemic caused by Shigella dysenterike Division of Siaya District, Nyanza Province in western
type 1 has been spreading through Africa since 1979. Kenya. This area was chosen because detailed census
Starting in north-east Zaire (-RIGHT et al., 1984), it data were available from a large bed net study being
subsequently spread through Rwanda and Burundi done in the same Division, and because it was in this
(FROST et al., 1981), Tanzania (MHALU et al., 1984), area that SHANKS et al. (1996) reported epidemic dys-
Zambia in 1990 (TLTITLE et al., 1995), and Mozam- entery in their study volunteers.
bique in 1993 (ARP.GON et al., 1995), reaching South Rarieda Division is situated on the north-east shore of
Africa in 1994 (ROLL,INS et al., 1995). IIJIMA et aZ. Lake Victoria. It covers an area of 200 km2, with an es-
(1995) first reported isolation of Sh. dysenteriae type 1 timated population of 55000 people; most are subsist-
organisms in Kenya in 1994, from an area of the coast ence farmers and members of the Luo tribe. The main
near the Somali border. In 1995, in western Kenya, crops grown are maize, millet, sorghum and sugar. Fish-
SHANKS et al. (1996) found that a significant proportion ing is done along the lake shore. The climate is typically
of malaria study volunteers developed dysentery, and tropical with a long rainy season occurring in April-June
culture results confirmed Sh. dysenteriae type 1 as the ae- and a short rainy season in November-December. Ma-
tiologic agent. laria, diarrhoea and acute respiratory infections are the
This virulent organism is the cause of much morbidity prevalent diseases. Thirteen medical facilities were
and mortality, with epidemic case fatality rates up to identified, which serve the residents of this Division. All
20%. In addition to the concern that Sh. dysenteriae may medical facilities in Kenya are required by the Ministry
soon develop resistance to all known antibiotics, devel- of Health to keep a record of patient visits, recording at
oping countries are already faced with an equivalent least the date of the visit, patient’s name, age, sex, home
situation: the remaining effective antibiotics, the fluoro- address, diagnosis, and treatment. Numerator data were
quinolones and perhaps azithromycin, are not afforda- extracted by reviewing all patient entries from 1 June
ble for most people. In Kenya, a single 500 mg 1994 to 3 1 May 1996 in the record books of all 13 med-
ciprofloxacin tablet costs the same as 12 L of milk, and ical facilities. Data extraction forms were used to record
a 3 d course of treatment costs the equivalent of 2 the medical facility, date of visit, patient’s initials (to
weeks’ average salary. In addition, those affected by ep- avoid duplicate entries), address (i.e., village), and diag-
idemic dysentery are generally the poorest citizens and nosis as recorded in the record book. The data were
least able to afford any medication. compiled by the Centers for Disease Control and Pre-
These problems highlight the need for an effective vention (CDC)/Kenya Medical Research Institute
and long-lasting ShigeZZavaccine. Studies with type-spe- (KEMRI) bed net project health facility recorders at 11
cific live oral vaccines and parenteral polysaccharide sites and by one of the investigators at the remaining 2
conjugate vaccines thus far have given only short-term facilities. The only diagnoses that were recorded for this
protection against the homologous serotypes (DELLERT study were dysentery, amoebiasis or diarrhoea.
& COHEN, 1994; KOTLOFF et al., 1995). One difficulty Discussions with the healthcare providers at the facil-
in evaluating new vaccine candidates is finding suffi- ities revealed a general consensus regarding how diag-
cient numbers of volunteers in whom to demonstrate ef- noses were made: ‘dysentery’ was dia-posea in patients
ficacy. The purpose of this study was to investigate, who anDeared ill. were febrile. and had bloodv diarrhoea
retrospectively, a purported dysentery epidemic to as- with iGdomina1 1cramps; ‘amoebiasis’ in patients who
certain rates and duration of the epidemic, and to deter- were afebrile and had small amounts of blood in their di-
mine members of the population at greatest risk. These arrhoea, with or without abdominal cramping; and ‘di-
data might then be used in planning future intervention arrhoea’ in patients who had increased frequency of
strategies, including a ShigeZZavaccine trial. watery stools without visible blood or fever. Denomina-
tor data were provided from the current census of the
Methods CDC/KEMRI bed net project. The initial census was car-
ried out in March 1996, to enumerate all members of the
We undertook a retrospective record review over 2 population of Rarieda Division, by name, age and sex.
years of all cases diagnosed as dysentery, amoebiasis or Subsequently, follow-up visits were made twice yearly to
each house to ascertain any demographic change.
Address for correspondence: M. A. Malakooti, USUHS Pre- Daily rainfall measurements were made at 5 sites in
ventive Medicine, A1040N, 4301 Jones Bridge Road, Bethes- Rarieda Division for the duration of the study, and these
da, MD 208 14, USA. data were used to calculate monthly average rainfall fig-
*This paper is not subject to copyright. ures for comparison with the dysentery rates.
542 M. A. MALAKOOTI ETAL.

Results Table. Age-specific dysentery attack rates in

A total of 3301 cases was recorded, of which 2191 western Kenya, June 1994-May 1996
were diagnosed as dysentery, 670 as amoebiasis, and
440 as diarrhoea. The numbers of all cases plotted Dysentery attack rate (%)
against time are shown in Fig. 1, and clearly illustrate a Age Location 4 Location 8 Mean
(years) (highest rates) (lowest rates) (all locations)
10.0 ;:; 4.4
29 ;:i 1.2 2.3
1.9
10-14
15-19 10.2 2.4 5.1
220 11.3 2.9 5.1
All ages 9.3 1.9 3.8

fall dropped from 25 cm in November 1994 to a mini-

mum of 3 cm in January 1995. As the rainy season
started in April 1995, the incidence of dysentery cases
diagnosed concomitantly fell.
Discussion
The results of this study confirmed the occurrence of
Fig. 1.Number of casesof dysentery,amoebiasisand diarrhoea epidemic dysentery in the Rarieda Division of western
in western Kenya, June 1994May 1996. Kenya during 1995. The number of diagnosed casesof
dysentery rose steeply in January 1995 (Fig. 1) and,
while they gradually fell again after February, they never
defined outbreak of dysentery starting in December returned to the previous low endemic levels. The data
1994, from a background rate of 20 cases/month, rising presented illustrate the potential value of routine sur-
to 230 cases/month. The total population of Rarieda Di- veillance data in the description of an epidemic of this
vision in June 1996 was 55530, giving a cumulative 2 nature. If monitored prospectively, such data could alert
years incidence rate for dysentery of 4%, of impending epidemics. The number of casesof amoe-
In order to determine the distribution and temporal biasis and diarrhoea diagnosed per month did not vary
and spatial clustering of the disease, denominators from in any obvious pattern, supporting the assertion of the
the census were taken, broken down by location and vil- occurrence of a dysentery epidemic. The low number of
lage, and then by age (age structure determined by cen- diarrhoea cases was striking, and undoubtedly largely
sus data). The dysentery age-specific 2 year attack rates due to under-reporting, rather than being an accurate
over the 2 years for the locations with the highest and reflection of the true incidence of uncomplicated diar-
lowest rates are shown in the Table. Location 4, Mem- rhoeal illness in the area. It may be that diarrhoea in
ba, had many more casesthan the others, increasing at children is regarded as unimportant by mothers unless
the same time in January 1995 but not peaking until the child is acutely ill with fever and visibly bloody
June. The overall attack rate for location 4 was 9.3% stools. It is unlikely that the increase in numbers of
over the course of the study, almost double that of the dysentery caseswas due only to reporting bias, as moth-
next hizhest area (location 1, 5.3%). The hitiest ers would have had the same criteria for reporting cases
monthl;incidence was 1% in location 4, follow&.l by before the epidemic. If news of a dysentery epidemic
0*8%, also in location 4. The dysentery caseswere 52% simply led to a shift, diagnosing dysentery instead of
males and 48% females. The age distribution showed a amoebiasis or diarrhoea, then one would expect a con-
lower incidence of dysentery in school-age children, and comitant fall in numbers of these 2 diagnoses while the
higher incidences in those <5 years old and in the 15-l 9 total number of all cases stayed the same: this did not
and >20 years age groups (Table), with a 2 years attack occur. There was no special incentive for healthcare
rate in location 4 of 11.3% in adults >20 years old. It is providers to diagnose casesas dysentery. Patients were
possible this was because of under-reporting of casesin not charged any more for having a more pernicious di-
school-age children, but the reasons are unclear. agnosis, and any medication prescribed, if available, did
The average monthly rainfall was plotted against the not result in income for the clinic or provider. It is
number of dysentery casesover the 2 years of the study unclear why the attack rate in location 4 was so much
(Fig. 2). The onset of the dysentery epidemic appeared higher than that in all the other locations. One possibil-
to coincide with the onset of the very dry season, as rain- ity is that it was due to better reporting, as one of the
clinics in this location was the site of several malaria
studies, and provided free care to study volunteers with
extensive community outreach. If this were even partial-
ly true, it suggests that the case numbers for the other
locations were underestimates due to under-reporting,
and that the calculated attack rates were also lower than

4%
actually occurred.
Stools cultured from 10 dysentery cases(18-55 years
old) in Tune 1995 (SHANKSet al., 1996). as the enidemic
wai waning, yieldid 4 significant isola&s: 2 Sh. -dysente-

IfIJLl
riae type 1, one Sh. dysenteriae type 2, and one Salmonel-
la enteriditis. Although only 3 cases of dysentery were

fj
cultured in June 1996 at the time of this study in the
same location, these were also found to be Sh. dysenteri-
aetype 1. This fits the pattern of epidemic dysentery be-
ing caused by Sh. dvsenteriaetype 1, and suggeststhat its
inuoduction to th; Kenya c&&t in 1994 Iti et al.,
Fig. 2. Number of cases of dysentery and rainfall in western 1995) was followed bv its snread inland and into west-
Kenya, June 1994-May 1996. em Kenya. Males anh fe&ales were equally affected,
DYSENTERY IN KENYA 543

and the basic age distribution of cases showed more demic shigellosis by the World Health Organization
adults affected than children. This also fits the euidemi- (WHO, 1995).
ology of shigellosis, which in its endemic form is caused In summary, epidemic dysentery has affected this
by the other .ShigeZZuspecies and affects mostly children area of western Kenya since January 1995, and by May
under the age of 10 years, but in epidemics is usually 1996 the attack rate had not returned to the pre-epi-
caused by Sh. dysenteriue type 1, with a much broader demic level. The disease was almost certainly caused by
age distribution. Sh. dysentetiue type 1, and the multi-antibiotic resistance
Of great concern is the widespread and rapidly in- of this organism is a serious threat to the health of the
creasing antibiotic resistance of S/z..dysenteriae type 1. In local population. This area may be a suitable site for
1976 the strain that caused the epidemic in Central more intensive, prospective dysentery surveillance in
America was found to be resistant to tetracycline, sul- preparation for future intervention such as a ShigeZlu
phathiazole, chloramphenicol and ampicillin (OLARTE vaccine trial.
et al., 1976). The central African strain initially had the
same resistance pattern, but within 2 years had also de- Acknowledaements
veloped resistance to trimethoprim/sulphamethoxazole. We are g&eful to Caleb Ouma for his invaluable assistance
By 1982 resistance to nalidixic acid had been docu- in the field. This paper is published with the permission of the
mented, and 3 1% of strains in Rwanda were resistant to Director, Kenya Medical Research Institute.
this antibiotic only 3 years later (MUTWEWINGABO &
METS, 1987). In their 1990 Burundi study, I&IS et al. Disclaimer
(1994) found 100% of their Sh. dysenteriue type 1 iso- The views of the authors do not purport to reflect those of
the US Navy, US Public Health Service, or US Army.
lates were resistant to ampicillin, tetracycline, chloram-
phenicol, streptomycin,- sulphafurazole (=sulfisox- References
azole), trimethoprim/sulphamethoxazole, and nalidixic Araeon. M.. Barreto. A.. Chambule. T.. Nova. A. & Tallarico.
acid. In Kenya, nalidixic acid and fluoroquinolone anti- M’. (i99$. Shigeliosii in Mozambiqbe: & 1993 outbreak
biotics have been effective, but a recent hospital treat- rehabilitation-a follow-up study. Tropical Doctor, 25,
ment study appeared to demonstrate emerging nalidixic 159-162.
acid resistance (Shanks, G. D., unpublished observa- Dellert, S. F. & Cohen, M. B. (1994). Diarrhea1 disease. Es-
tions), and the cost of the fluoroquinolones is prohibi- tablished pathogens, new pathogens, and progress in vaccine
tive for most local patients. The common practice of development. Gastroenterolopy - Clinics in North America, 23,
empirical multi-antibiotic therapy of any patient who 637-654.
gives a history of bloody diarrhoea probably contributes Ebright, J. R., Moore, E. C., Sanbor?, W. R., Schaberg, D.,
Kyle, J. & Ishida, K. (1984). Epidemic Shiga bacillus dysen-
to this spreading resistance. terv in Central Africa. American 7ournal of Troaical Medicine
While this epidemic was not caused by a change in the and Hvtiene. 33. 1292-1297. I . ’
people’s general living standards, the effect of the water Frost, J.-i., Rbwe; B., Vandepitte, J. & Threlfall, E. J. (1981).
shortage associated with the drastic decrease in rainfall Plasmid characterization in the investigation of an enidemic
at the time of the onset of the epidemic may be relevant. caused by multiply resistant Shigellaldysenteriae tie 1 in
In this area of Kenya there is plenty of water during the Central Africa. Lance&ii, 1074-1076.
rainy season to meet all the people’s needs, but this is Iiiima. Y., Oundo. 1. 0.. Taea. K.. Saidi. S. M. & Honda. T.
not true during the dry season. They do not have access (1995): Simult&eous ouib;eak due to Vibrio chokrae and
to piped water, and very few have access to deep, pro- ShigeUa dysenteriae in Kenya. Lancet, 345,69.
Kotloff, K. L., Losonsky, G. A., Nataro, J. P., Wasserman, S.
tected wells. Instead, water is collected from the few un- S., Hale, T. L., Taylor, D. N., Newland, J. W., Sadoff, J. C.,
protected ponds remaining from the rains, where Formal, S. B. & Levme, M. M. (1995). Evaluation of the
clothes are washed, bathing done and livestock watered. safety, immunogenicity, and efficacy in healthy adults of four
The collected water is then carried in containers often doses of live oral hybrid Escherichiu coli-Shigella flexneri 2a
for long distances back to the houses, where it becomes vaccine strain EcSf2a-2. &tine. 13.495-502.
a precious commodity used for drinking and cooking, Mhalu, F. S., Moshi, W. K. & Gbaga, I. (1984). A bacillary
and usually not for hand-washing. In addition, during dysentery epidemic in Dar es Salaam, Tanzania. Journal of
the dry season people also use water from the shore of Diarrhoeal Diseases Research, 2,2 17-222.
Lake Victoria, which is contaminated by human and an- Mutwewingabo, A. & Mets,T. (1987). Increase in multiresist-
imal wastes, and may have been the source of this epi- ante of a Shipella dwenteriae tvDe 1 strain in Rwanda. East
African Medic& Jo&al, 64,8 fs-8 15.
demic. Olarte, J., Filloy, L. & Gallindo, E. (1976). Resistance of Shig-
This simple retrospective study has several obvious elk dysenteriaetype 1 to ampicillin and other antimicrobial
limitations. First, there was no way to determine the ac- agents: strains isolated during a dysentery outbreak in a hos-
curacy of the diagnoses that were made in the peripheral pital in Mexico City. Journal of Infectious Diseases, 133,
health facilities. All but one were simple outpatient clin- 572-575.
ics, with patients clinically diagnosed without laboratory Ries, A. A., Wells, J. G., Olivola, D., Ntakibirora, M., Ivey, C.
facilities. The criteria for the diagnosis of dysentery were B., Greene, K. B., Tenover, E C., Wahlquist, F. B., Griffin,
quite broad, and it could be argued that cases of dysen- P. M. & Tauxe, R. V. (1994). Epidemic ShigeUa dysenteriue
type 1 in Burundi: panresistance and implications in preven-
tery and amoebiasis were possibly often misdiagnosed. tion. your& of Infectious Diseases, 169, 1035-1041.
Nevertheless, this would have been true before and Rollins, N. C., Wittenberg, D. F., Coovadia, H. M., Pillay, D.
throughout the study period, and so simple diagnosis G., Karas, A. J. & Sturm, A. W. (1995). Epidemic Shigella
misclassification would not explain the marked rise in dysenteriue type 1 in Natal. 3ournal of Tropical Pediatrics, 41,
dysentery cases seen in this epidemic. Also, it cannot be 281-284.
assumed that all cases of dysentery in the Rarieda pop- Shanks, G. D., Ragama, 0. B., Aleman, G. M., Andersen, S.
ulation were reported to the 13 health facilities. More L. & Gordon, D. M. (1996). Azithromvcin pronhvlaxis ure-
affluent patients may have visited a private clinic, and vents epidemic dysentery. fiansactions if the R&ai Society of
others may have sought help from traditional healers. Trobical Medicine and Hvkne. 90.3 16.
The extent of the underestimate is unknown, and thus Tuttle, J., Ries, A. A., Chzba, k M., Perera, C. U., Bean, N.
H. & Griffin, P. M. (1995). Antimicrobial-resistant enidemic
the rates depicted are a minimum. However, it is prob- Shigella dyse&ri~ @pe 1 ‘in Zambia: modes of transmission.
able that the true incidence was proportionally well rep- Journal of Infectious Diseases, 171,371-375.
resented over time. We were able to control for an WHO (1995). Guidelines for the Control of Epidemics due to Shig-
overestimate of cases by recording the home village of all ella dysenteriae type 1, 2nd edition. Geneva: World Health
the dysentery patients and excluding non-Rarieda pa- Organization.
tients. Finally, the argument for assuming that the dys-
entery cases were caused by Sh. dysenteriue type 1 has Received 26 March 1997; revised 12 May 1997; accepted
been discussed above, and fits with the review of epi- for publication 13 May 1997