Download A study of infectious intestinal disease in England: plan and

D Sethi, JG Wheeler, JM Cowden, LC Rodrigues, PN Sockett, JA Roberts, P Cumberland,
DS Tompkins, PG Wall, MJ Hudson, PJ Roderick
original reports
A study of infectious intestinal disease in England:
plan and methods of data collection
Key words:
Summary: The Committee on the Microbiological Safety of Food, set up in 1989 by the
Department of Health in response to national epidemics of foodborne infection, considered case control studies
the available evidence and commissioned a study of infectious intestinal disease (IID) in
cohort studies
England. Seventy practices (with 489 500 patients overall) recruited from the Medical Research
compliance
Council’s General Practice Research Framework between August 1993 and January 1995
diarrhoea
collected data for one year. The practice populations were representative of practices in England
family practice
by area and urban/rural location, but with fewer small and affluent practices. There were five
gastrointestinal
main components. i) A population cohort of 9776 (40% of those eligible) were enrolled to
diseases
estimate the incidence and aetiology of IID in the community, and a large proportion were
incidence
followed up. A median of 10% of patients on practice age-sex registers had moved away or
intestinal diseases
died. ii) A nested case control component based on cases ascertained in the cohort was used to
methods
identify risk factors for IID in the community. iii) In a case control component used to
risk factors
identify risk factors and to estimate the incidence and aetiology of IID presenting in 34 general
practices 70% of the 4026 cases returned risk factor questionnaires, 75% submitted stools,
and matched controls were found for 75% of cases. iv) An enumeration component was used
to estimate the incidence of IID presenting to general practitioners (GPs) in 36 practices and
the proportion of specimens sent routinely for microbiological examination. v) In a
socioeconomic costs component used to estimate the burden of illness of IID in the community
and presenting to GPs 63% of those who returned a risk factor questionnaire also returned a
socioeconomic questionnaire and were representative by age, sex, and social class.
Despite variable enrolment and compliance the study sample had sufficient power for
the multivariable analysis. The characteristics associated with low enrolment and compliance
must be considered in the interpretation of the main study results.
Commun Dis Public Health 1999; 2: 101-7.
D Sethi, PJ Roderick
MRC Epidemiology and Medical Care Unit
JG Wheeler, LC Rodrigues, JA Roberts, P Cumberland
London School of Hygiene and Tropical Medicine
JM Cowden, PN Sockett, PG Wall
PHLS Communicable Disease Surveillance Centre
DS Tompkins
Leeds Public Health Laboratory
MJ Hudson
Centre for Applied Microbiology and Research
On behalf of the Infectious Intestinal Disease Study Executive:
JS Brazier, MM Brett, D Brennan, W Browne, PE Cook, JM Cowden,
P Cumberland, RP Eglin, N Fasey, S Gordon-Brown, P Hayes,
MJ Hudson, V King, JM Kramer, J Martin, C Olohan-Bramley, RJ Owen,
JA Roberts, PJ Roderick, LC Rodrigues, B Rowe, D Sethi, HR Smith,
MT Skinner, R Skinner, PN Sockett, DS Tompkins, PG Wall, JG Wheeler,
AL Wight.
Address for correspondence:
Dr Dinesh Sethi
Department of Public Health and Policy
London School of Hygiene and Tropical Medicine
Keppel Street
London WC1E 7HT
tel: 0171 927 2122
fax: 0171 637 5391
email: [email protected]
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Introduction
Increases in notifications of food poisoning and reports
of cases of campylobacter, salmonella, and listeria
infection, and national epidemics of foodborne
infections with organisms such as Salmonella enteritidis
led the Secretary of State for Health and Minister of
Agriculture, Fisheries and Food to set up the
Committee on the Microbiological Safety of Food in
1989 (‘Richmond Committee’) 1-4. The committee
reported that all indices of infectious intestinal disease
(IID) had increased between 1980 and 1990 2 and
recommended that a representative sample of all cases
of IID in England should be studied to identify causal
organisms, estimate the true incidence of disease and
its relationship to reported cases, and determine the
potential risk factors associated with the acquisition
of IID and the costs incurred by individuals and by
society. In response the Department of Health
commissioned a study of IID in England whose
methods are described in this paper.
A pilot study carried out in 1991 and 1992 tested
the feasibility of the design, established the basis for
the sample size calculations, and compared options
VOL 2 NO 2 JUNE 1999
101
original reports
for selection and follow up of subjects in general
practice5. The degree of underascertainment of cases
presenting with IID in this setting has been discussed6.
The main study began in August 1993 and data were
collected until January 1996. The following
organisations shared responsibility for the main study:
the PHLS, including its Communicable Disease
Surveillance Centre (CDSC), Leeds Public Health
Laboratory (PHL), and PHLS reference laboratories for
specific organisms, the Centre for Applied
Microbiology and Research (CAMR), the Medical
Research Council Epidemiology and Medical Care
Unit (EMCU) and the General Practice Research
Framework, and the London School of Hygiene and
Tropical Medicine (LSHTM).
FIGURE 1 Study design
Population cohort
component
70 practices
Enumeration
component
36 practices
Nested case control
component
70 practices
GP case control
component
34 practices
Socioeconomic costs
component
70 practices
Methods
Setting: stratification and selection of practices
The General Practice Research Framework is a
network of over 800 practices committed to research,
that care for about 10% of the registered national
population. Based on sample size calculations
(described below), a total of 70 volunteer practices
were selected from the framework. The country was
divided into three areas of similar population sizes
for comparison: i) North (former health regions of
Northern, Yorkshire, North Western, and Mersey), ii)
Midlands and South West (East Anglia, West
Midlands, Trent, South Western, and Wessex), and iii)
South East (Thames Regions). The number of practices
selected in each area was in proportion to the area’s
total population according to the 1981 Census.
Practices were selected to represent each area’s
socioeconomic and urban or rural characteristics.
Tertiles of the population distribution of ward-based
Jarman deprivation scores 7 were used to stratify by
socioeconomic characteristics and Office of Population
Censuses and Surveys area aggregates used to classify
locations as urban or rural.
Design of the main study and study components
The study included five main components (figure 1):
• a population cohort component to estimate the
incidence and aetiology of IID in the community
• a nested case control component based on cases
ascertained in the cohort to identify risk factors for
IID in the community
• a GP case control component to identify risk factors
and to estimate the incidence and aetiology of IID
presenting to GPs
• an enumeration component to estimate the
incidence of IID presenting to GPs and the
proportion of specimens sent routinely for
microbiological examination
• a socioeconomic costs component to estimate the
burden of illness of IID occurring in the community
and presenting to GPs.
Practice recruitment was staggered between August 1993
and January 1995. Each practice collected data for one
year from recruitment. Each practice dedicated a part
102
VOL 2 NO 2 JUNE 1999
time research nurse to the study. All 70 practices took
part in the cohort component and were randomised
within each stratum to take part in either the GP case
control or the enumeration component (see below).
Population cohort with nested case control
component
Two consecutive cohorts were each followed up for
six months. For each cohort, 200 people selected by
stratified random sampling by age and sex from the
age-sex registers of 70 practices were invited to take
part. The notes of people who the nurse could not
contact were searched at three months to determine
whether they had moved away or died. Participants
attended a briefing interview, completed a baseline
questionnaire, and were given a stool collection kit.
They posted diary cards each week to the nurse, to
declare that they had no symptoms of IID that week.
Those who developed symptoms (incident cases) were
asked to contact the nurse, fill in a risk factor
questionnaire, and submit a stool specimen. A control
matched for age and sex was selected systematically
from the cohort for the nested case control component
and asked to submit a stool specimen and fill in a risk
factor questionnaire 5. Two consecutive cohorts of six
months rather than one of 12 months were used to
increase the level of participation5. Age and sex strata
with poor responses in the first cohort were
oversampled in the second.
General practice case control component
Cases who presented in the surgery or on home visits
were ascertained in 34 practices over one year. Those
who fulfilled the case definition (box) were contacted
by the nurse and asked to fill in a risk factor
questionnaire and submit a stool specimen. Out-ofhours deputising agencies were contacted for details
of consultations for IID. For each case up to five sex
and age matched controls selected systematically from
the register were invited in sequence to attend until
one accepted. If the first potential control refused then
a second was invited, and so on. Controls were asked
to fill in a risk factor questionnaire and submit a stool
specimen.
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Definitions
Case definition
People of all ages with loose stools or significant vomiting (more than once in
24 hours, incapacitating, or accompanied by cramps or fever) lasting less than
two weeks, in the absence of a known non-infectious cause and preceded by a
symptom free period of three weeks. Exclusions: people with non-infectious
causes of diarrhoea - such as Crohn’s disease, ulcerative colitis, cystic fibrosis,
and coeliac disease - and non-infectious causes of vomiting such as surgical
obstruction, alcohol intoxication, morning sickness, infant regurgitation.
Control definition
People free of loose stools or significant vomiting for three weeks before the
matched case became ill. Controls were matched to cases by age (within age
group for children aged 0 to 5 months, within age group for children aged 6 to
11 months, within one year for children aged 1 to 4 years, but not below 11
months, within two years for those aged 5 to 19 years, but not below 4 years,
and within five years for adults, but not below 18 years). Controls for cases
over 5 years of age were also matched for sex.
General practice enumeration component
All cases presenting over one year in 36 practices were
eligible for inclusion. GPs were asked to follow their
normal clinical practice in requesting laboratory
investigations. The research nurse recorded personal
and clinical details, whether a stool specimen was
taken, and, if so, the results. Deputising agencies were
contacted for details of out-of-hours consultations.
Socioeconomic costs component
Three weeks after inclusion in the cohort, case control,
and enumeration components all cases received a
socioeconomic questionnaire. Cases who did not
return the questionnaire were sent a repeat
questionnaire.
Stool testing
Stool collection kits provided in the case control and
cohort components comprised a plastic ‘universal’
specimen pot, a sealable bag, a small plastic spoon, a
cardboard box, and an adhesive postage-paid label
addressed to Leeds PHL. Microbiological methods
and their prioritisation are described in the study
protocol8 and the accompanying paper (page 108)8a.
Specimens of at least 10 g were needed for complete
testing and archiving. Leeds PHL performed initial
tests and coordinated further tests at reference
laboratories. Clinically significant findings were
reported to the local nurse and consultant in
communicable disease control.
Questionnaires
The risk factor questionnaire sent to cases and controls
in the GP case control and nested case control
components
asked
for
sociodemographic
characteristics, clinical details, and known and
suspected risk factors for IID, both short and long term
(for example, consumption of foods, contact with pets,
travel, contact with cases of gastroenteritis, medicines
taken, type of accommodation, food handling, social
factors)5. The baseline cohort questionnaire sought
sociodemographic, accommodation, and food
handling details. Non-respondents were sent a
questionnaire that asked about family size, social
class, and reasons for refusal to take part. The
socioeconomic questionnaire asked about household
COMMUNICABLE DISEASE AND PUBLIC HEALTH
composition, income, the impact of the illness in
terms of consequent use of resources by individuals,
their families, the NHS, and absence from work, and
willingness to pay to avoid illness. Questionnaires
were modified for child cases and controls.
original reports
BOX
Data management
Staff of the EMCU and the LSHTM coordinated data
collection and management. Data were coded, double
entered using Epi-Info, validated, and monitored
(manually or by computers) at every stage for
irregularities9. Practice performance was monitored
and progress reports were returned to nurses at six
weeks and subsequently every three months. The
reports compared the actual number of cases and
controls monitored with the numbers predicted by the
pilot study5. Regional training nurses made scheduled
visits to each practice, investigating any
underperformance. Overall study performance was
monitored by the IID Executive Committee, which met
every three months.
Sample size calculations
The sample size was selected in order to estimate the
incidence of IID in England with a precision of 10% on
each side (95% confidence interval), to estimate the
incidence in each of the three geographical areas with a
precision of 20%, and to ascertain the organisms present
in at least 25% of cases8. This was based on the results
of the pilot study, which estimated a community
incidence of 13.8/100 person years, a presentation rate
of 2.13/100 person years, a positive stool rate of 0.108/
100 person years, and a compliance of about 75%5. The
sample size estimates suggested that 70 practices in the
cohort study with two six-month cohorts of 90 people
each would yield 6300 person years of follow up, and
that 35 practices in each of the case control and
enumeration components followed up for one year
would yield about 280 000 person years of follow up.
Within this sample it was calculated that the numbers
in the GP case control analysis would be large enough
to detect an odds ratio of 2 at the 1% level for exposures
with a prevalence of 10% in the population in both adults
and children. For reasons of logistics and cost, the GP
case control and enumeration components were run in
the same practices as the cohort.
Statistical analysis
The practice and cohort populations were compared
with those of England using the national registered
population and the Office of National Statistics (ONS)
1994 population estimate respectively10,11. Multiple
regression models were developed to predict the
practice characteristics associated with list inflation
(the retention on GP lists of patients who have moved
away or died) and factors associated with low
compliance, using Stata software12.
Ethics
Approval was obtained from the Royal College of
General Practitioners, LSHTM, and PHLS and from
VOL 2 NO 2 JUNE 1999
103
original reports
TABLE 1 Study and regional (mid 1994) population
distribution by area (North, Midlands and South West,
and the South East), and by Jarman score tertiles
Population (%)
Study practices
Country
North
Midlands and South West
South East
128 120 (29)
218 788 (44)
148 758 (30)
13 254 200 (27)
21 228 700 (44)
14 049 800 (29)
Total
495 666
48 532 700
<5 Low
-5 to 10 Mid
>10 High
110 172 (22)
201 813 (41)
183 681 (38)
14 223 630 (30)
14 870 582 (32)
17 960 992 (38)
Total
495 666
48 532 700
Area
Jarman
all 61 local research ethics committees. Written
informed consent was obtained from cases and
controls by each practice research nurse.
Results
Representativeness
GP practice population characteristics
The study practice population of 495 660 accounted
for 1% of the English population and was
representative of the three geographical areas (table
1). The lower tertile (least deprived) of Jarman scores
was slightly underrepresented in the study practice
population (22%) when compared with England as a
whole (30%). A lower proportion of the study
population attended GP practices with four or fewer
partners (43% ) than did the population of England
(58%). The study practice population and the English
population were similar as regards age and sex and
urban/rural location (62% urban, 61% rural).
Community cohort
Forty-five per cent of the study population enrolled
were male compared with 49% of the national
population (table 2, ONS mid-1994 estimate).
Enrolment was proportionately lowest in the 15 to 24
year age group when compared with the national
population. Fifty-six per cent of the cohort were from
TABLE 2 Age and sex distribution of community
population taking part in cohort study, compared with
population of England (mid 1994 estimate)
Cohort population
Age (yrs)
<1
1-4
5-9
10-14
15-24
23-34
35-44
45-54
55-64
65-74
75+
Totals
104
Male (%)
48
254
360
315
289
469
519
652
544
479
228
Female (%)
(1)
(6)
(9)
(8)
(7)
(11)
(12)
(16)
(13)
(12)
(5)
4157 (45)
47
277
382
298
396
719
718
741
618
503
291
(1)
(6)
(8)
(6)
(8)
(14)
(14)
(15)
(12)
(10)
(6)
4990 (55)
VOL 2 NO 2 JUNE 1999
Population of England
in thousands
Male (%)
325.6
1332.4
1628.9
1546.3
3206.7
4011.8
3286.1
3052.0
2373.4
1977.2
1142.6
(1)
(6)
(7)
(6)
(13)
(17)
(14)
(13)
(10)
(8)
(5)
Female (%)
308.8
1268.4
1545.7
1464.7
3039.0
3861.7
3259.3
3052.3
2444.3
2377.6
2203.5
(1)
(5)
(6)
(6)
(12)
(16)
(13)
(12)
(10)
(10)
(9)
23 882.1 (49) 24 825.4 (51)
manual classes compared with 65% in England (1991
census). The population was similar in ethnic
composition to the national population (1991 census).
Seventy-two per cent of the cohort were married
compared with 58% nationally and 15% were single
compared with 26%. A smaller proportion of the
cohort were economically active (58% vs 61%) and a
larger proportion had retired from work (23% vs 19%).
A larger proportion were homeowners (80% vs 72%).
Enrolment in the community cohort
A total of 27 651 people were invited to take part in
the cohort study, 35% of whom (9776) enrolled and
24% (6686) declined (figure 2). Together these 16 462
people (59% of the total invited) constituted the total
known to be eligible for enrolment. Eight per cent
(2177) of the remaining 11 189 were known to be
ineligible because they were no longer registered with
the general practice, having moved away or died. The
nurse could not contact 3844 (14%) of the remaining
9012 (33%) people, and the reason for non-enrolment
was not obtained for the remaining 5168 (19%): (the
forms of 577 were lost and for 4591 the reason was
not recorded).
Ineligibility was higher among males and those
aged 15 to 34 years. Sixty-one per cent of refusers
provided sociodemographic details. Twenty-five per
cent of refusers were economically inactive, compared
with 9% of those who enrolled.
Estimate of practice list inflation
Practice records of 3007 of the 3844 people (figure 2)
not contacted were searched and 430 (14%) were found
to be ineligible, having moved or died. Ten per cent
of the 1238 of the 4591 people whose reason for nonenrolment was not recorded and whose notes were
searched were ineligible. We assumed that 10% of the
577 people whose forms were lost would also be
ineligible. A total of 3252 were estimated to be
ineligible (figure 2), a crude proportion of 12%.
The multiple regression model applied to the 3007
people not contacted showed that ineligibility was
commonest in areas with a mid-tertile Jarman score,
rural areas, and urban areas in the South East. This
was used to predict the proportion ineligible by
practice. A median of 10% of those invited were
ineligible (5th and 95th centiles 5.1 and 24.7). The
median adjustment used to correct for list size inflation
was therefore 90%. This corrected list size was then
used as the denominator to calculate the presentation
rates of IID in the case control and enumeration
components.
The overall enrolment after correcting for ineligible
patients on practice lists was 40%.
Compliance
GP case control component
A median of 70% of cases returned questionnaires and
a median of 75% submitted stool specimens (table 3).
Controls were found for a median of 75% of cases: 50%
of potential controls agreed at the first invitation but
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Cohort selected for invitation 27 651
Known to be eligible
16 462
Enrolled
9776
Refusers
6686
(35%)
(24%)
Not known if
eligible or not
Known to be
ineligible
No contact (3844) +
Reason not recorded (4591) +
Unknown* (577)
= 9012
(33%)
Known
to be
ineligible
2177
(8%)
Estimated
eligible
7937
Total estimated eligible
24 399
original reports
FIGURE 2 Numbers who enrolled, refused, and estimated to be eligible for cohort component practices
Estimated
ineligible
1075
Total estimated ineligible
3252
*Unknown: forms lost at two practices
the proportion declined to 37%, 27%, 18%, and 13%
for the second, third, fourth, and fifth invitations,
respectively. Higher proportions of controls than cases
completed questionnaires and submitted stool
specimens. Compliance of both cases and controls was
lowest among males and the 15 to 24 year age group.
Regression modelling showed that high Jarman score,
urban location, and location in the South East were
the practice characteristics associated with lower
questionnaire and stool compliance for cases and
control, and poorer case control matching.
Seventy-six per cent (5254/6897) of subjects in the
case control component submitted stools for analysis.
More than 90% of all specimens were tested to stage 5
(see table 1, page 109)8a. A third of specimens were
received for analysis within four days of onset of
illness and 63%, 77%, and 85% within 7, 10, and 14
days, respectively. Ninety-five per cent of specimens
were received within four days of voiding.
TABLE 3 Risk factor questionnaire, stool submission, and
socioeconomic questionnaire compliance in GP case
control component and nested case control component
Number
returned
Proportion (%) returned per practice
Median
5th centile 95th centile
GP case control component
Cases (of 4026)
Risk factor questionnaires
Stools submitted
Socioeconomic questionnaire
2642
2962
1652
70
75
43
40
54
17
81
85
61
Controls (of 2871)
Risk factor questionnaires
Stools submitted
2429
2292
88
85
68
56
100
96
Risk factor questionnaires
Stools submitted
Socioeconomic questionnaire
648
761
555
81
94
67
–
57
–
100
100
92
Controls (of 675)
Risk factor questionnaires
Stools submitted
613
555
100
88
67
50
100
100
Enumeration component (of 4744)
Socioeconomic questionnaire
2182
48
23
72
Nested case control component
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Baseline questionnaire return, follow up of
population cohort, and nested case control
component
Ninety-five per cent of the cohort component returned
baseline questionnaires. Eighty-two per cent
completed more than 23 weeks of follow up and 61%
completed the full 26 weeks. Overall, stool and
questionnaire compliance was higher in the nested
case control component (table 3) than in the GP case
control component.
Low compliance with
questionnaire return, stool submission, and case
control matching was associated with 10 to 24 year
age group, males, urban location, the South East, and
practices with high Jarman scores.
Returning the socioeconomic questionnaire
Compliance was higher in the cohort component
(median 67%) than the enumeration (48%) and GP case
control (43%) components. Overall compliance was
not high but 63% of those who returned risk factor
questionnaires also returned a socioeconomic
questionnaire. The age, sex, and social class
distribution was similar to that in the case control and
cohort studies, suggesting that the responses were
representative.
Discussion
Representativeness
The study practice population of nearly half a million
accounted for about 1% of the population of England.
Stratified recruitment of the 70 practices achieved a
study population representative of the national
population by age, sex, geographical area, and urban/
rural composition. The least deprived areas and
smaller practices were slightly underrepresented, but
this reflected the composition of the General Practice
Research Framework rather than the sampling
strategy. This is the first time IID has been studied in
such a large and representative practice population
sample, and the framework appeared to provide a
suitable sampling frame. Previous population based
studies in the United Kingdom (UK) were restricted
VOL 2 NO 2 JUNE 1999
105
original reports
to between one and seven practices, mainly in urban
areas, and were unlikely therefore to represent the
national population 13-16.
Despite a corrected enrolment rate of 40%, the
characteristics of the cohort population were very
similar to the population of England. Enrolment was
slightly lower in people aged 15 to 24 years, males,
and in manual classes than the structure of the national
population would imply. A larger proportion were
married, retired, and owned their own homes. Others
have found lower response rates in subjects who are
unmarried, unemployed, or poorly educated 17. A
population based community cohort study in the
Netherlands also found that people aged 19 to 35 years
and males were less likely to take part than the
reference population 18, but the Dutch sample also
underrepresented people over 65 years, a problem not
encountered in our study. Studies from the United
States have used highly selected cohort populations19-21.
One studied a cohort of 61 families with young
children, whose main wage earners had professional,
managerial, or sales occupations19. The results were
therefore less generalisable to the whole population
than ours. Previous larger scale community surveys
in the UK have used retrospective recall alone to detect
cases, which may be subject to bias16,22.
General practice list inflation
In the GP practice components it was assumed that
the total practice population would be followed for
the year of participation, but a correction was made
for practice list size inflation. This correction used
evidence collected during recruitment of the cohort,
which showed that a proportion of the names on
practice lists had moved out of the area or died. The
median list inflation was calculated to be 10%, within
the range of other estimates (5% to 30%)22,24-27. Our
correction accounted for known and estimated
exclusions for each practice and therefore reflected
variations between practices.
A second correction to the incidence in general
practice was made for suspected underascertainment
of cases – that is, failure to report a study case to the
coordinating centre at the EMCU 6. This was estimated
by using practices with computerised diagnosis to
identify patients who should have been ascertained
but were not. A median of 64% of cases were
ascertained: the proportion varied according to
practice characteristics: number of partners, urban or
rural location, study component, and previous
research experience 6. Estimates from the practices
selected were extrapolated to the remaining practices,
according to their individual characteristics. The
representativeness of the population and the
adjustments for list inflation and underascertainment
should have increased the accuracy of the corrected
IID presentation rate.
Compliance and follow up
Compliance in this study compared favourably with
other cohort studies of IID18. The corrected enrolment
106
VOL 2 NO 2 JUNE 1999
was 40%, a large proportion of baseline questionnaires
were returned, and 82% completed more than 23
weeks. A Dutch study described an enrolment of 36%,
86% of whom completed more than 14 of the 17 weeks
of total follow up18. Our follow up data are likely to
be highly accurate as they were derived using weekly
report cards. This is important as the person years of
follow up formed the denominator for calculating
incidence. Two consecutive cohorts of six months each
were used as opposed to one of 12 months, to increase
the level of participation5. Recruitment of controls and
compliance in the nested case control component was
very high, thus reducing potential bias.
In the GP case control component 70% of cases
returned questionnaires and 75% submitted stool
specimens. As presentation rather than completion
of the questionnaire was the means by which cases
were ascertained, no adjustment for presentation rates
had to be made for non-responders. In another study
88% of questionnaires were completed and 67% of the
cases provided stool specimens 15 . Although we
monitored the return of questionnaires as a regular
quality assurance check, the compliance we achieved
was not as high. Seventy-five per cent of controls were
matched to cases. Compliance among controls was
higher than among cases, probably because they had
chosen to take part in the study. For cases and
controls, compliance and case control matching were
both lowest in the 15 to 24 year age group and in males.
The practice characteristics associated with lower
compliance identified by regression modelling were
high Jarman score, urban location, and location in the
South East. Although lower compliance in the 15 to
24 year age group could have caused bias, information
collected showed that cases and controls were affected
equally. Matched analysis of risk factors would also
have helped to minimise confounding. Respondents
to the socioeconomic questionnaire had a similar
distribution of social class, age, and sex to that in the
case control and cohort studies, suggesting that the
sample was representative.
Implications of the study
The robustness of the overall design demonstrated
by the representativeness of the total sample
population and reasonable compliance with stools
and risk factor questionnaires suggests that efforts
made to avoid bias in this study were reasonably
successful and that its results will be applicable to
England as a whole 28 . Comparison of this study
w i t h o t h e r s t u d i e s o f g a s t ro e n t e r i t i s i n t h e
population supports this assertion, as discussed
above 13-16,19-22. The corrections made for practice list
size and for underascertainment will make
adjustments for presentation rates in general
practice more accurate 6 . Despite the strategy of
matched analysis potential biases the results may
not be entirely representative of urban practices
with high Jarman scores, and this may need to be
taken into account when interpreting the main
results. Researchers conducting observational
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Acknowledgements
We thank Professor TW Meade, the EMCU staff, and
M Goldsborough, A Williams, L Hands, E Marshall, P
Allen, F Symes, S Fox, and J Elwood for their
invaluable help.
We are grateful to the general practices in the
MRC’s General Practice Research Framework:
Abingdon, Ampleforth, Aston Clinton, Atherton,
Backwell, Barrow-in-Furness, Bassingham, Beckton,
Biggleswade, Bingley, Birmingham, Birstall, Bishop
Auckland, Bloomsbury, Bolton, Bournemouth,
Bracknell, Brightlingsea, Bromley, Camberley, Chapel
en le Frith, Chessington, Coleford, Cradley, Crediton,
Darlington, Durham, East Grinstead, Folkestone,
Fovant, Fremington, Guildford, Hampton Wick,
Heacham, Henleaze, Hull, Ipswich, Kidderminster,
Ledbury, Leeds, Limehouse, Liskeard, Luton,
Newcastle, Northampton, Ormskirk, Pinner,
Richmond, Scunthorpe, Settle, Sheffield, St Johns,
Stockport, Stratford upon Avon, Street, Sutton-inAshfield, Temple Fortune, Tewkesbury, Todmorden,
Tonbridge, Tooting, Torrington Park, Tunbridge Wells,
Walsall, Weymouth, Whitehaven, Widnes, Willesden,
Workington, and Yaxley.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
References
1. Editorial. Salmonella enteritidis phage type 4: chicken and
egg. Lancet 1988; ii: 720-2.
2. Richmond M (chairman). Report on the microbiological safety
of food. London: HMSO, 1990.
3. Hopper TJ, Mead GC, Rowe B. Poultry meat as a source of
human salmonellosis in England and Wales. Epidemiol Infect
1988; 100: 175-84.
4. Skirrow MB. Campylobacter enteritis: a new disease. BMJ
1977; ii: 9-11.
5. Roderick PJ, Wheeler JG, Cowden JM, Sockett PN,
Rodrigues LC. A pilot study of infectious intestinal disease
(IID) in England. Epidemiol Infect 1995; 114: 277-88.
6. Sethi D, Wheeler J, Rodrigues LC, Fox S, Roderick P. Investigation
of under-ascertainment in epidemiological studies based in
general practice. Int J Epidemiol 1999; 28: 106-12.
7. Jarman B. Identification of underprivileged areas. BMJ
1984; 286: 1705-9.
8. A study of infectious intestinal disease in England. Study
protocol. London: Department of Health, 1993.
8a. Tompkins DS, Hudson MJ, Smith HR, Eglin RP, Wheeler
JG, Brett MM, et al. A study of infectious disease in England:
microbiological findings in cases and controls. Commun Dis
Public Health 1999; 2: 108-13.
9. Dean AG, Dean JA, Burton AF, Dicker RC. Epi-Info Version 5: a
COMMUNICABLE DISEASE AND PUBLIC HEALTH
21.
22.
23.
24.
25.
26.
27.
28.
word processing, database and statistics program for epidemiology
on microcomputers. Stone Mountain GA: USD, 1990.
1994 Key population and vital statistics. Local and health
authorities. Series VS no.20. London: Office for National
Statistics, 1996.
General Medical Services Statistics 1994. London: Department
of Health, 1995.
Stata statistical software release 5.0. College Station, TX: Stata
Corporation, 1997.
Kendall EJC, Tanner EI. Campylobacter enteritis in general
practice. Journal of Hygiene 1982; 88: 155-63.
Tuckman E, Chapple PAL, Franklin LM, Manser IN, Woodall
JT, Randall KJ, et al. Acute gastrointestinal illness in general
practice. BMJ 1962; 135-41.
Nathwani D, Grimshaw J, Taylor RJ, Ritchie LD, Douglas
JG, Smith CC. Factors influencing general practitioner
referals to hospital of adults with presumed infective
diarrhoea. Br J Gen Pract 1994; 44: 17l-4.
Palmer S, Houston H, Lervy B, Rebiro D, Thomas P.
Problems in the diagnosis of foodborne infection in general
practice. Epidemiol Infect 1996; 117: 479-84.
Shahar E, Folsom R, Jackson R. The effect of nonresponse
on prevalence estimates for a referent population: insights
from a population-based cohort study. Ann Epidemiol 1996;
6: 498-506.
Hoogenboom-Verdegaal AMM, de Jong JC, During M,
Hoogenveen R, Hoekstra JA. Community-based study of
the incidence of gastrointestinal diseases in the Netherlands.
Epidemiol Infect 1994; 112: 481-7.
Dingle JH, Badger GF, Feller AE, Hodges RG, Jordan WS,
Rammelkamp CH. A study of illness in a group of Cleveland
families. I. Plan of investigation and certain general
observations. American Journal of Hygiene 1953; 58: 16-30.
Hodges RG, McCorkle LP, Badger GF, Curtiss C, Dingle JH,
Jordan WS. A study of illness in a group of Cleveland
families. xi: The occurrence of gastrointestinal symptoms.
American Journal of Hygiene 1956; 64: 349-356.
Monto AS, Koopman JS. The Tecumseh Study, xi: Occurrence
of acute enteric illness in the community. Am J Epidemiol
1980; 112: 323-33.
Feldman RA, Banatvala N. The frequency of culturing stools
from adults with diarrhoea in Great Britain. Epidemiol Infect
1994; 113: 41-4.
Fraser RC, Clayton DG. The accuracy of age-sex register,
practice medical records and family practitionar committee
registers. J Roy Coll Gen Pract 1981; 31: 416-9.
Roe L, Mant D, Coulter A. Screening: the inadequacy of
population registers. BMJ 1989; 298: 1100.
Fraser RC. The reliability and validity of the age sex register
as a population denominator in general practice. J Roy Coll
Gen Pract 1978; 28: 283-6.
Sheldon MG. The accuracy of age sex registers in general
practice. J Roy Coll Gen Pract 1984; 34: 269-71.
O’Mahoney PG, Thomson RG, Rodgers H, Dobson R, James OFW.
Accuracy of the family health services authority register in
Newcastle upon Type, UK as a sampling frame for population
studies. J Epidemiol Community Health 1997; 51: 206-7.
Doll R. Progress against cancer: an epidemiologic
assessment. The 1991 John C Cassel memorial lecture. Am J
Epidemiol 1991; 134: 675-88.
VOL 2 NO 2 JUNE 1999
original reports
studies need to be aware of these factors in order
to develop strategies to improve and allow for low
compliance.
107
original reports
A study of infectious intestinal disease in England:
microbiological findings in cases and controls
DS Tompkins, MJ Hudson, HR Smith, RP Eglin, JG Wheeler, MM Brett, RJ Owen, JS Brazier,
P Cumberland, V King, PE Cook
Summary: A study was undertaken to identify the microorganisms and toxins in
Key words:
stool specimens associated with infectious intestinal disease (IID) among cases in
case control studies
the community and presenting to general practitioners (GPs) and in asymptomatic
cohort studies
controls. Population based cohorts were recruited from practice lists in 70 practices
communicable diseases
and followed for 26 weeks (cohort component). Seven hundred and sixty-one cases
community-acquired
of IID identified from the cohorts, 2893 cases who presented to GPs in 34 of the
infections
practices (GP component), and age/sex matched control subjects (555 and 2264,
compliance
respectively) submitted stool specimens by post for comprehensive microbiological
diarrhoea
examination.
family practice
Campylobacter spp (12.2% of stools tested), rotavirus group A (7.7%), and
small round structured virus (SRSV) (6.5%) were the organisms most commonly gastrointestinal diseases
incidence
detected in the GP component. SRSV was identified in 7.0% of cases in the community
cohort. No target microorganisms or toxins were identified in 45.1% and 63.1% of
intestinal diseases
cases in the two components. Aeromonas spp, Yersinia spp, and some enterovirulent
risk factors
groups of Escherichia coli were detected as frequently in controls as in cases.
toxins
The higher frequency of detection of campylobacter, salmonella, and rotavirus
among cases who presented to GPs than among those in the community suggests
that those pathogens cause more severe illness. No enteropathogens were detected
from a large proportion of cases although comprehensive standard methods were
used to seek them.
Commun Dis Public Health 1999; 2: 108-13.
DS Tompkins, RP Eglin
Leeds Public Health Laboratory
MJ Hudson
Centre for Applied Microbiology and Research
HR Smith, MM Brett, RJ Owen
PHLS Central Public Health Laboratory
JG Wheeler, P Cumberland
London School of Hygiene and Tropical Medicine
JS Brazier
Cardiff Public Health Laboratory
V King, PE Cook
Department of Health
On behalf of the IID Study Executive, whose membership
included: JS Brazier, MM Brett, D Brennan, W Browne, PE Cook,
JM Cowden, P Cumberland, RP Eglin, N Fasey, S Gordon-Brown,
P Hayes, MJ Hudson, V King, JM Kramer, J Martin, C OlohanBramley, RJ Owen, JA Roberts, PJ Roderick, LC Rodrigues,
B Rowe, D Sethi, HR Smith, MT Skinner, R Skinner, PN Sockett,
DS Tompkins, PG Wall, JG Wheeler, AL Wight.
Address for correspondence:
David Tompkins
Leeds Public Health Laboratory
Bridle Path
York Road
Leeds LS15 7TR
tel: 0113 232 1300
fax: 0113 260 3655
email: [email protected]
108
VOL 2 NO 2 JUNE 1999
Introduction
Current data on the significance of enteropathogenic
microorganisms in England and Wales are derived
largely from routine voluntary reporting of specified
isolates by clinical microbiology laboratories to the
PHLS Communicable Disease Surveillance Centre
(CDSC). These reports are thought to represent a
fraction of the true incidence of infections as only a
proportion of cases seek medical attention, only a
subset of these submit a stool specimen for analysis1,
and there are also deficiencies and inaccuracies in
reporting by laboratories 2 .
Further limited
information is derived from CDSC’s surveillance
scheme for general outbreaks (those affecting
members of more than one household) of IID2.
Previous studies undertaken to define
microbiological causes of gastrointestinal disease have
sought to identify only a limited number of the
recognised or putative enteropathogens in faecal
specimens submitted from general practice, but have
not attempted to define the relative contributions of
enteropathogenic bacteria, protozoa, and viruses in
the community as a whole3-5.
Following a recommendation of the Richmond
Committee 6 , a complex collaborative study was
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Methods
Subjects
Cases were defined by symptoms of loose stools or
vomiting (details and exclusions page 103) 9. Single
stool specimens were obtained from cases and controls
in two study components – a population cohort
component (with a nested case control element) and a
GP case control component. Thirty-four out of 70 GP
surgeries throughout England selected to be
representative of the population as a whole took part
in the GP case control component 9 . As practice
recruitment was staggered, specimens were received
between August 1993 and January 1996.
Laboratory methods
Stool specimens were submitted by first class post to
Leeds Public Health Laboratory (PHL). Investigations
of small specimens were prioritised as about 9 grams
of faeces were needed for completion of investigations
for all target organisms and toxins (table 1). An aliquot
of each specimen of sufficient volume was sent to the
PHLS Laboratory of Enteric Pathogens (LEP) for
examination for the pathogenic or potentially
pathogenic (enterovirulent) groups of Escherichia coli
using DNA probes.10 A second aliquot was frozen at
-70°C in cryoprotective broth to be archived for further
studies 11.
TABLE 1 Target organisms sought and priority list for
microbiological investigations
Priority
of testing Procedure
Conventional selective and enrichment culture
techniques were used for the isolation of bacterial
enteropathogens.
Bacillus spp. spores and
Staphylococcus aureus were counted, as counts of >104/g
and >106/g of faeces, respectively, were considered
significant. Isolates were sent to PHLS reference
laboratories for confirmation of identity and typing.
Clostridium difficile cytotoxin (toxin B) was detected
using Vero cells. C. perfringens enterotoxin was
detected using an agglutination assay (PET-RPLA;
Oxoid, Basingstoke, Hampshire) and positive results
confirmed using an in-house enzyme immunoassay12
(EIA) at the Food Hygiene Laboratory of the PHLS
Central Public Health Laboratory (CPHL).
Conventional light microscopy of a wet film and
formol-ether concentrate was used to detect ova, cysts,
and parasites. An auramine-stained smear was
examined using fluorescence microscopy for cysts of
Cryptosporidium parvum and Cyclospora cayetanensis.
Transmission electron microscopy (EM) was used
to detect viruses, with EIAs for rotavirus group A
(Rotascreen; Microgen, Camberley, Surrey) and
adenovirus types 40/41 (Adenoclone type 40/41;
Cambridge Biotech, Worcester, MA, USA). All
astrovirus identifications were confirmed by culture
and specific fluorescent antibody testing (Oxford
PHL). Rotaviruses detected by EM but not by EIA
were subsequently confirmed as rotavirus group C by
the Enteric Virus Unit of CPHL using polyacrylamide
gel electrophoresis.
Written clinical reports were posted to
participating surgeries and all clinically significant
findings were reported by telephone. Data on
microbiological findings were analysed at the London
School of Hygiene and Tropical Medicine.
Statistical methods
The frequency with which organisms were identified
in controls was standardised by age to reflect the age
distribution of the general population rather than that
of the cases. The direct standardisation method was
applied, using the mid-1994 population age
distribution of England.
Target organism sought
Results
All 6743 stools received were cultured for all bacterial
target organisms and 90% of them were of sufficient
volume for investigation to stage 5 (table 1).
Specimens were obtained between 0 and 73 days after
the onset of symptoms (median 3 days, mode 1 day).
Specimens were received by Leeds PHL between 0 and
45 days after being obtained (median 2 days, mode 1
day).
Target organisms or toxins were identified in 54.9%
of cases in the GP component and 36.9% in the cohort
component (tables 2 and 3). Small round structured
virus (SRSV) was the most commonly identified target
organism in cases of IID in the community cohort (7%
of cases positive). Campylobacter spp. and rotavirus
group A were commoner than SRSV in cases in the
GP component. The distribution of microorganisms
Stage 1
Bacteriological culture
Campylobacter sp
Stage 2
Bacteriological culture
Aeromonas sp, Bacillus sp,
Clostridium difficile, Salmonella sp,
Shigella sp, Staphylococcus aureus,
Vibro sp, Yersinia sp
Stage 3
Bacterial culture
Direct microscopy
Escherichia coli O157,
Giardia intestinalis
Stage 4
1-2g faeces to the PHLS
Laboratory of Enteric
Pathogens (DNA probes)
Direct microscopy
Enterovirulent E. coli,
Cryptosporidium parvum,
Cyclospora cayetanensis
Stage 5
Virology (electron
microscopy and enzyme
immunoassay
Adenovirus, astrovirus, calicivirus
rotavirus, SRSV (Norwalk-like)
Stage 6
Toxin tests
culture counts for
vegetative cells and spores
C. difficile, C. perfringens,
B. cereus, S. aureus
Stage 7
Concentration and
Protozoa and helminths
for ova, cysts, and parasites
Stage 8
20% frozen suspension
Archiving at CAMR
COMMUNICABLE DISEASE AND PUBLIC HEALTH
original reports
undertaken to determine the true incidence of
infectious intestinal disease (IID) in England and to
estimate the incidence of gastrointestinal disease in
the community attributable to a microbiological
cause 7. The feasibility of the methodology adopted
was confirmed in a pilot study 8. This paper presents
the main microbiological results of the study. Further
details of the methods are included in an
accompanying article9, and other results including
incidence, socioeconomic costs, and risk factors will
be published in a detailed report 7.
VOL 2 NO 2 JUNE 1999
109
original reports
TABLE 2 Target organisms identified in the GP case control component
Cases
Number
identified
Bacteria
Aeromonas spp
Arcobacter cryaerophilus
Bacillus spp (>104/g)
Campylobacter spp
Clostridium difficile cytotoxin
C. perfringens enterotoxin
Escherichia coli O157
Enterovirulent E. coli (DNA probes):
Attaching and effacing
Diffusely adherent
Enteroaggregative
Enteropathogenic
Enterotoxigenic
Vero cytotoxigenic (non-O157)
Plesiomonas shigelloides
Salmonella spp
Shigella spp
Staphylococcus aureus (>10 6/g)
Vibrio spp
Yersinia spp
Protozoa
Cryptosporidium parvum
Giardia intestinalis
Viruses
Adenovirus types 40, 41
Astrovirus
Calicivirus
Rotavirus group A
Rotavirus group C
SRSV
Organism or toxin not detected
Number
tested
Controls
Percentage
identified
Number
identified
Number
tested
Percentage
identified
Standardised
percentage*
164
1
4
353
38
114
3
2893
2893
2571
2893
2259
2871
2893
5.7
<0.1
0.2
12.2
1.7
4.0
0.1
96
–
8
16
41
15
–
2264
2264
2176
2264
2039
2256
2264
4.2
–
0.4
0.7
2.0
0.7
–
3.2
–
0.3
0.5
0.4
0.5
–
119
103
141
4
52
6
1
146
23
10
1
51
2774
2774
2774
2774
2774
2774
2893
2893
2893
2568
2893
2893
4.2
3.7
5.1
0.1
1.9
0.2
<0.1
5.0
0.8
0.4
<0.1
1.8
67
93
43
6
–
9
–
10
–
5
–
56
2230
2230
2230
2230
2230
2230
2264
2264
2264
2172
2264
2264
3.0
4.2
1.9
0.3
–
0.4
–
0.4
–
0.3
–
2.5
1.5
3.8
1.8
0.2
–
0.5
–
0.4
–
0.1
–
2.4
39
28
2892
2893
1.3
1.0
2
10
2264
2264
0.1
0.4
–
0.3
81
77
40
208
6
169
2612
2612
2612
2709
2709
2612
3.1
3.0
1.5
7.7
0.2
6.5
3
5
4
9
–
6
2210
2210
2210
2211
2211
2210
0.1
0.2
0.2
0.4
–
0.3
–
–
–
0.1
–
0.1
1305
2893
45.1
1834
2264
81.0
85.2
* percentage standardised by the population age distribution (England, mid 1994)
SRSV: small round structured virus (Norwalk-like)
Organisms not detected in specimens in this study include; Arcobacter butzleri, A. skirrowii, Balantidium coli, Cyclospora cayetanensis, Enteroinvasive Escherichia
coli, Entamoeba histolytica, Helicobacter cinaedi, H. fennelliae, nematodes, Vibrio parahaemolyticus
in controls, when standardised by age, illustrates the
carriage of enteropathogenic microorganisms by
asymptomatic individuals in the general population
of England (tables 2 and 3).
The relative proportions of frequencies of detection
in cases and controls in the GP component (where the
numbers were larger) are compared in table 4,
highlighting the differences in proportions between
different microorganisms. High case:control ratios
were seen with recognised enteropathogens such as
campylobacter, salmonella, C. parvum, and the enteric
viruses. More than one target organism or toxin was
identified in 11.3% of cases presenting in the GP
component, 6.4% in the cohort component, and in less
than 2% of all controls (table 5). Many combinations
were observed but none predominated.
The percentages of isolates identified only after
enrichment procedures in cases and controls,
respectively, were 31.5% and 58.3% for salmonella,
4.5% and 0% for campylobacter, 78.5% and 84.7% for
aeromonas, and 64.9% and 83.1% for yersinia. The
effect of enrichment on isolation rates became more
pronounced as time between the onset of symptoms
110
VOL 2 NO 2 JUNE 1999
and the receipt of the specimen in the laboratory
increased for both salmonella and campylobacter (data
not shown).
The speciation and typing of bacterial isolates were
consistent with data from national surveillance, with
C. jejuni and C. coli being the most frequently identified
campylobacters (88% and 9% respectively) and S.
enteritidis and S. typhimurium the commonest
salmonellas (55% and 20%, respectively).
The percentages of faecal specimens positive for
selected target organisms and toxins in different age
groups (table 6) confirmed previous observations4,13,14.
Viral infections were commoner in infants and young
children and salmonella and campylobacter infections
commoner in older children and adults.
Discussion
This is the first British investigation to use such a wide
range of microbiological methods to study IID in a
large and representative population. Similar
physician-based and population cohort studies with
comprehensive microbiological investigations are in
progress in the Netherlands 15,16. Isolation rates for
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Cases
Bacteria
Aeromonas spp
Arcobacter cryaerophilus
Bacillus spp (>104/g)
Campylobacter spp
Clostridium difficile cytotoxin
C. perfringens enterotoxin
Escherichia coli O157
Enterovirulent E. coli (DNA probes):
Attaching and effacing
Diffusely adherent
Enteroaggregative
Enteropathogenic
Enterotoxigenic
Vero cytotoxigenic (non-O157)
Plesiomonas shigelloides
Salmonella spp
Shigella spp
Staphylococcus aureus (>10 6/g)
Vibrio spp
Yersinia spp
Protozoa
Cryptosporidium parvum
Giardia intestinalis
Viruses
Adenovirus types 40, 41
Astrovirus
Calicivirus
Rotavirus group A
Rotavirus group C
SRSV
Organism or toxin not detected
Controls
Number
identified
Number
tested
Percentage
identified
Number
identified
Number
tested
Percentage
identified
Standardised
percentage*
46
–
–
32
9
9
–
761
761
684
761
614
756
761
6.0
–
–
4.2
1.5
1.2
–
28
–
2
4
5
3
–
555
555
526
555
487
551
555
5.0
–
0.4
0.7
1.0
0.5
–
5.2
–
0.4
0.9
0.3
1.1
–
23
23
21
1
12
3
–
8
1
1
–
26
732
732
732
732
732
732
761
761
761
683
761
761
3.1
3.1
2.9
0.1
1.6
0.4
–
1.1
0.1
0.1
–
3.4
10
13
4
2
–
6
–
2
–
1
–
16
542
542
542
542
542
542
555
555
555
524
555
555
1.9
2.4
0.7
0.4
–
1.1
–
0.4
–
0.2
–
2.9
1.0
2.0
0.5
0.2
–
1.0
–
0.2
–
–
–
3.1
3
3
761
761
0.4
0.4
–
3
555
555
–
0.5
–
0.5
13
14
8
29
2
50
715
715
715
718
718
715
1.8
2.0
1.1
4.0
0.3
7.0
1
1
1
–
–
3
535
535
535
535
535
535
0.2
0.2
0.2
–
–
0.6
–
–
–
–
–
0.5
480
761
63.1
462
555
83.2
83.9
original reports
TABLE 3 Target organisms and toxins identified in the cohort component
* per cent standardised by the population age distribution (England, mid 1994)
SRSV: small round structured virus (Norwalk-like)
Organisms not detected in specimens in this study include; Arcobacter butzleri, A. skirrowii, Balantidium coli, Cyclospora cayetanensis, Enteroinvasive E. coli,
Entamoeba histolytica, Helicobacter cinaedi, H. fennelliae, nematodes, Vibrio parahaemolyticus
campylobacter and salmonella were higher in this
study than have been reported previously 3-5. No
pathogens were detected in 45% of cases of IID in the
GP component and 63% of cases in the community
cohort component, however, and there are several
possible reasons for this. Some microorganisms
identified as causes of IID (for example, Listeria
monocytogenes, microsporidia) were not sought in this
study and other unrecognised pathogens may exist.
Non-infective causes of intestinal disease can evoke
symptoms similar to those of IID, although our case
definition sought to exclude these causes. There were
some technical reasons for potential under-recovery:
a single specimen only was examined from each case,
30% of specimens were too small for complete
analysis, and delays between onset of symptoms and
receipt of specimens could have impaired viability and
therefore detection by the methods used. EM detection
of virus particles is relatively insensitive. Use of
sensitive molecular methods such as polymerase chain
reaction (PCR) amplification would probably have
increased detection. The potential remains for further
investigation of the archived material using such
COMMUNICABLE DISEASE AND PUBLIC HEALTH
molecular techniques.
Viruses were confirmed as the commonest causes
of IID in the community, particularly in children, with
SRSV the most frequently identified target organism
in the population cohort component. Evidence from
outbreaks, studies of children admitted to hospital
with diarrhoea, and failure to identify bacterial
pathogens in the majority of cases all suggest that
viruses are responsible for many cases of community
acquired IID13,14. The contribution of SRSV and other
viruses to widespread morbidity has now been
confirmed. In contrast to the other viruses, SRSV were
identified more frequently in cases in the population
cohort than in the GP component. This is consistent
with the finding that SRSV infection generally causes
less severe illness than infection with other recognised
pathogens such as rotavirus, salmonella, and
campylobacter7. Rotavirus group A was the target
organism most frequently identified in children up to
the age of 5 years in the GP component. It has been
estimated that about 18000 children under 5 years of
age are admitted to hospital in England and Wales
each year for rotavirus infection17. A rotavirus vaccine
VOL 2 NO 2 JUNE 1999
111
original reports
TABLE 5 Frequency of multiple organisms in faecal
specimens (percentage)
TABLE 4 Relative proportions of target organisms in
cases and controls in GP case control component
Bacteria
Aeromonas spp
Bacillus spp (>104/g)
Campylobacter spp
Clostridium difficile cytotoxin*
C. perfringens enterotoxin
Escherichia coli O157
Enterovirulent E. coli (DNA probes):
Attaching and effacing
Diffusely adherent
Enteroaggregative
Enteropathogenic
Enterotoxigenic
Vero cytotoxigenic (non-O157)
Salmonella spp
Shigella spp
Staphylococcus aureus (>10 6/g)
Vibrio spp
Yersinia spp
Protozoa
Cryptosporidium parvum
Giardia intestinalis
Viruses
Adenovirus types 40, 41
Astrovirus
Calicivirus
Rotavirus group A
Rotavirus group C
SRSV
Relative
proportion
identified
(case/control)
95%
confidence
interval
1.35
0.42
17.31
6.05
5.96
NA
1.05-1.72
0.13-1.40
10.52-28.49
1.81-20.26
3.49-10.17
–
1.43
0.89
2.64
0.54
NA
0.54
11.43
NA
1.69
NA
0.71
1.06-1.92
0.68-1.17
1.88-3.69
0.15-1.90
–
0.19-1.50
6.03-21.63
–
0.58-4.94
–
0.49-1.04
15.27
2.54
3.69-63.15
1.24-5.21
22.84
13.03
8.46
18.86
NA
23.83
7.23-72.23
5.28-32.14
3.03-23.16
9.70-36.68
–
10.58-53.69
Organisms
identified
NA: not applicable (no positive control)
SRSV: small round structured virus (Norwalk-like)
* aged over 1 year
was licensed for use in the United States in August
199818 and approval for use in the United Kingdom is
awaited.
C. jejuni was the most frequently identified
target organism in the GP component, with group
A rotavirus the runner-up, reflecting patterns seen
in national surveillance data 2 , although the two
sources cannot be compared directly. In routine
diagnostic practice many of the target organisms
0
1
2
3
4
GP component Community
cases (%)
cohort (%)
(n=2893)
cases (n=761)
Controls (%)
(n=2819)
1305 (45.1)
1261 (43.6)
276 (9.5)
48 (1.7)
3 (0.1)
2296
478
41
4
–
480 (63.1)
232 (30.5)
48 (6.3)
1 (0.1)
–
Total (%)
(n=6473)
(81.4) 4081 (63.0)
(17.0) 1971 (30.5)
(1.5)
53 (0.8)
(0.1)
53 (0.8)
3 (0.1)
and toxins are sought only when travel history, food
history, or clinical features indicate a need for
specific tests. The relationship between the results
of this study, including results of other aspects of
this study, and national surveillance data are
presented elsewhere 7,19.
Salmonellas were isolated more frequently from
cases in the GP component (5%) than from cases in
the community cohort component (1.1%), reflecting
the severity of disease associated with this pathogen.
VTEC O157 was isolated and also detected by DNA
probes in three cases only in the GP component (0.1%),
indicating that cases occur rarely, although affected
individuals
may
have
severe
disease.
Enteroaggregative E.coli, a heterogeneous group
associated with diarrhoea in children and adults20, was
the most commonly detected enterovirulent E.coli in
GP component cases.
The rarity of recognised enteropathogens in control
subjects was expected, but the relatively high rates of
detection of aeromonas, yersinia, diffusely adherent
E. coli, and attaching and effacing E.coli in controls in
both components of the study suggests that their
detection in cases was not always related to IID.
Typing of these organisms did not help to identify
pathogenic subsets. Aeromonas and yersinia in most
cases and controls were isolated only after enrichment,
indicating that these microorganisms were present in
small numbers in stool specimens. Further studies
on archived isolates may help to identify differences
in virulence characteristics and improve our
understanding of the role of these microorganisms in
IID.
TABLE 6 Percentage of stools positive, by age group, for cases with selected target organisms in the GP case control
(GPCC) and population cohort (PC) components
< 1 year
Number of cases submitting specimen:
Campylobacter spp
Clostridium difficile cytotoxin
Clostridium perfringens enterotoxin
Salmonella spp
Cryptosporidium parvum
Adenovirus types 40, 41
Astrovirus
Rotavirus group a
SRSV and calicivirus
1-4 years
5-14 years
15-74 years
> 74 years
GPCC
302
PC
31
GPCC
606
PC
156
GPCC
221
PC
123
GPCC
1664
PC
427
GPCC
94
PC
24
2.0
7.2*
4.0
2.3
0.7
6.9
1.9
21.3
14.7
6.4
28.6*
0.0
0.0
0.0
7.1
7.1
10.3
17.8
5.4
1.7
5.6
2.5
2.8
10.3
6.7
17.3
14.8
6.4
0.9
1.9
1.9
1.3
6.1
4.0
8.8
13.5
11.3
0.6
4.5
5.9
5.0
1.0
3.0
6.7
6.0
3.2
0.0
1.6
2.4
0.8
1.8
0.9
4.5
9.9
16.9
0.7
3.2
6.7
0.5
0.2
1.8
2.3
6.0
3.5
0.3
0.9
0.5
0.0
0.0
1.2
2.0
5.2
9.6
7.6
5.4
0.0
0.0
1.1
1.1
4.4
10.1
4.2
5.3
0.0
0.0
0.0
0.0
0.0
0.0
4.5
* percentage in controls GPCC – 16.6, PC – 21.0
SRSV: Small round structured virus (Norwalk-like)
112
VOL 2 NO 2 JUNE 1999
COMMUNICABLE DISEASE AND PUBLIC HEALTH
Acknowledgements
We thank the staff of the laboratories that took part in
the study, the many individuals who provided helpful
comments and guidance, and the surgeries in the
MRC’s General Practice Research Framework (listed
on page 107) 9.
References
1. Feldman RA, Banatvala N. The frequency of culturing stools
from adults with diarrhoea in Great Britain. Epidemiol Infect
1994; 113: 41-4.
2. Wall PG, de Louvois J, Gilbert RJ, Rowe B. Food poisoning:
Notifications, laboratory reports and outbreaks – where do
the statistics come from and what do they mean? Commun
Dis Rep CDR Rev 1996; 6: R93-100.
3. Hunt DA, Shannon R, Palmer SR, Jephcott AE. Cryptosporidium
in an urban community. BMJ 1984; 289: 814-6.
4. Skirrow MB. A demographic survey of campylobacter,
salmonella and shigella infections in England. Epidemiol
Infect 1987; 99: 647-57.
5. PHLS Study Group: Cryptosporidium in England and
Wales; prevalence and clinical and epidemiological features.
BMJ 1990; 300: 774-7.
6. The Committee on the Microbiological Safety of Food. The
microbiological safety of food: part 1. London; HMSO, 1990.
7. Report of the study of infectious intestinal disease in
COMMUNICABLE DISEASE AND PUBLIC HEALTH
England. London: HMSO, 1999 (in press).
8. Roderick P, Wheeler J, Cowden J, Sockett P, Skinner R,
Mortimer P, et al. A pilot study of infectious intestinal
disease in England. Epidemiol Infect 1995; 114: 277-88.
9. Sethi D, Wheeler JG, Cowden JM, Rodrigues LC, Sockett
PN, Roberts JA, et al. A study of infectious intestinal disease
in England: plan and methods of data collection. Commun
Dis Public Health 1999; 2: 101-7.
10. Smith HR, Scotland SM. Recent developments in laboratory
techniques for the detection of diarrhoeagenic Escherichia
coli. PHLS Microbiology Digest 1994; 11: 7-12.
11. Hudson MJ, Hill MJ, Elliott PR, Berghouse L, Burnham WR,
Lennard-Jones JE. The microbial flora of the rectal mucosa
and faeces of patients with Crohn’s disease before and during
antimicrobial therapy. J Med Microbiol 1984; 18: 335-45.
12. Bartholomew BA, Stringer MF, Watson GN, Gilbert RJ.
Development and application of an enzyme linked
immunosorbent assay for Clostridium perfringens type A
enterotoxin. J Clin Pathol 1985; 38: 222-8.
13. Caul EO. Viral gastroenteritis: small round structured
viruses, caliciviruses and astroviruses. Part II. The
epidemiological perspective. J Clin Pathol 1996; 49: 959.
14. Kapikian AZ. Overview of viral gastroenteritis. Arch Virol
1996; 12 [suppl]: 7-19.
15. DeWit MAS, et al. Gastroenteritis in sentinel practices in
the Netherlands. In: Proceedings of the 4th world congress on
foodborne infections and intoxications. Berlin: Federal Institute
for Health Protection of Consumers and Veterinary
Medicine, 1998: vol 1: 226-31.
16. Abbink F, et al. A population cohort study with a nested casecontrol study; a study design to estimate the incidence and
aetiology of gastroenteritis in the Netherlands. In: Proceedings
of the 4th world congress on foodborne infections and intoxications.
Berlin: Federal Institute for Health Protection of Consumers
and Veterinary Medicine, 1998: vol 2: 766-9.
17. Ryan MJ, Ramsay M, Brown D, Gay NJ, Farrington CP, Wall
PG. Hospital admissions attributable to rotavirus infection
in England and Wales. J Infect Dis 1996; 174(suppl 1): S12-8.
18. Centers for Disease Control and Prevention, Rotavirus
vaccine for the prevention of rotavirus gastroenteritis among
children: recommendations of the Advisory Committee on
Immunization Practices (ACIP). MMWR Morb Mortal Wkly
Rep 1999; 48 (RR-2): 1-23.
19. Wheeler JG, Cowden JM, Sethi D, Wall PG, Rodrigues LC,
Tompkins DS, et al. Study of infectious intestinal disease in
England: rates in the community, presenting to GPs, and reported
to national surveillance. BMJ 1999; 318: 1046-50.
20. Law D, Chart H. Enteroaggregative Escherichia coli. J Appl
Microbiol 1998; 84: 685-97.
21. Joint DH/PHLS Working Group. The prevention and
management of Clostridium difficile infection. London,
Department of Health/PHLS, 1994.
22. Standing Medical Advisory Committee, Sub-Group on
Antimicrobial Resistance. The path of least resistance.
London: Department of Health, 1998.
23. Goosens H, Sprenger MJW. Community acquired infections
and bacterial resistance. BMJ 1998; 317: 654-7.
VOL 2 NO 2 JUNE 1999
original reports
C. perfringens enterotoxin was confirmed as an
important cause of IID in the community and was
commoner in GP patients (4%) than in the population
cohort (1.2%). C. difficile cytotoxin was most frequently
identified in both cases and controls under 2 years of
age, when the organism is a common feature of normal
faecal flora21, but was also found in 21 older cases who
presented to GPs, associated with prior use of
antibiotics 7. Current efforts to reduce unnecessary
prescribing of antibiotics may reduce the frequency
of associated episodes of diarrhoea 22,23.
This study has helped to improve understanding
of the microbial causes of IID in England and
highlighted the differences between the spectrum of
disease in the community as a whole and among those
who present to GPs. Results of this study will help
medical microbiologists to optimise laboratory
protocols for the investigation of stool specimens
submitted by GPs. The relative numerical importance
of different microorganisms has been clarified but this
should be considered alongside the severity of illness
that they cause7. Work on archived stool specimens
and bacterial isolates will help to provide further
useful information.
113