Meanwhile another study in Kenya
is seeking to validate WHO’s ICF guidelines (and screening tool) regionally.
“There
is some language in the WHO Guidelines, some caveats that bear mentioning,”
said Dr Sean Cavanaugh of the CDC. “First, performance of screening is contingent
upon prevalence of disease and other diseases that might present with similar
symptoms. The symptoms and reporting can be influenced by regional and cultural
factors such as smoking prevalence, which could certainly affect coughing, in
the way coughing is reported to clinicians. Regional temperatures might
influence the experience of fevers and night sweats, [as could] when patients in the
course of their disease register for HIV care and how well they communicate the
symptoms with the clinician, all of which could be culturally bound.”
“Screening
algorithms therefore should be validated in the region before they are to be
rolled out, ideally,” he said.
Dr
Cavanaugh presented preliminary
findings of a study that KEMRI and CDC performed attempting to validate
and compare three clinical screening methods ruling out TB in HIV-infected
persons while evaluating the performance of various diagnostic tests, including
new tools such as
GeneXpert, to determine the prevalence and incidence of TB among
HIV-infected persons in Kenya.2
The symptoms screens included:
The
ID-TB/HIV symptom screen
This was based on the ID-TB/HIV study that enrolled
approximately 2000 patients from 2004 to 2006 in South-East
Asia. Dr Cavanaugh said this screen is considered to be very
robust, as it was based upon the “gold standard” for TB diagnosis (including three
cultures of sputum; one of urine, one of blood and one of stool from every
patient). Yes to any of the following questions should identify the patient as
a TB suspect who should receive a full diagnostic work-up for TB, that would
include X-rays and smears):
- Any cough in the past
four weeks
- Any fever in the past
four weeks
- Drenching night sweats
lasting for more than three weeks
The
WHO 4 Symptom TB screen
This is based upon a broad review of ICF studies (as described
above).
- Current cough (in the past 24
hours);
- Any fever;
- Any night sweats; or
- Any weight loss
The
Kenya Ministry of Health (MoH) symptom screen
Like many other ministries of
health, Kenya’s
MOH have rolled out their own ICF case-finding algorithm that’s based mostly on
expert opinion of the disease epidemiology in their country, which was intended
to be very sensitive. It used seven elements:
- Cough for 2 or more
weeks: This symptom alone prompts an evaluation for TB, with microscopy
and X-ray if available
- Fevers for 2 or more
weeks
- Night sweats for 2 or
more weeks
- Noticeable weight loss
- Contact with known case
- Chest pain or
breathlessness
- Swelling in neck,
armpit, ABD, joints or groin
- Any of these above
symptoms would prompt the clinician to engage in a more detailed
interview, and physical exam.
The
study enrolled newly registered patients at 15 randomly selected HIV clinics in
three northern districts in Nyanza Province, who were seven years old or older
and had not received TB therapy in the past year, and who consented to
participate. All patients were screened using all three recommended ICF tools. Participants
are due to be followed for a full year to assess incidence of clinical disease.
At
enrolment, the participants provide three sputum specimens, regardless of
symptoms. The first specimen is sent for fluorescent microscopy and culture;
the second, a morning specimen, is sent for standard (ZN staining) microscopy,
fluorescent microscopy, GeneXpert and culture. Specimen 3, is the second spot
specimen – which was either collected on that first day or on the second day
when the morning specimen was brought in. All culture results are confirmed by
a Capilia TB test and line-probe assay. In addition, as a sub-study, stool and
lymph node aspirates are being collected from the subset of patients at the
larger hospitals.
A
TB case was defined as a patient with M. tuberculosis identified on any culture
or nucleic acid amplification test. A patient was defined as being negative for
TB if he or she had one sputum culture with no growth, plus another sputum
specimen and that was negative by culture or nucleic acid amplification.
Participants who didn’t meet either of these definitions were considered not
evaluable (these had missing samples or contaminated cultures).
Lab diagnosis results
Dr
Cavanaugh presented data from three of the study’s 15 sites (the first phase of
the study). 245 patients enrolled into the official study: 224 with known TB
status (positive or negative), 21 were not evaluable. At Kisumu District
Hospital, the largest
site (located in Kisumu city), 22 of the 139 enrolled patients had TB disease
with a prevalence of 16%; in Nyahera Health Centre, five out of the 40 enrolled
patients, (13%) and at Yala sub-district hospital, six out of 45 patients (13%)
were diagnosed with TB. Overall there were 33 cases out of 224 enrolled, for a
total prevalence of 15%, in this population.
The
diagnoses were made on the following basis:
- GeneXpert,
alone: 4 cases (12%)
- Sputum
Culture and GeneXpert: 22 cases (66%)
- Sputum
Culture alone: 7 cases (21%)
Dr
Cavanaugh then presented calculations to determine the incremental yield (IY) from adding culture or GeneXpert to the diagnostic
work-up — in other words, how many more cases would be diagnosed if these lab
tests were routinely available for diagnosis. For culture, the sample size was
30, as there were only 30 patients who had both two sputum smears examined by
fluorescent microscopy, and two evaluable (non-contaminated) cultures. Out of
those 30:
- Two fluorescent microscopy smears alone would diagnose only twelve cases
(40% sensitivity for fluorescent microscopy);
- If one culture is performed in
addition to those two fluorescent microscopy tests, twelve more cases can be
diagnosed for a total of 80% of the 30 cases and an IY of 12 or 40% (increased
sensitivity for adding one culture).
- Adding one more culture (two cultures
total), allowed for four more cases to be identified (for an incremental yield
of 13% - four out of 30.
- Together, a laboratory diagnostic work-up of two
fluorescent microscopy tests, plus two cultures would diagnose 93% of the
evaluable patients with TB; two cases would remain undiagnosed.
Doing
the same analysis for GeneXpert yielded the following: again, the two
fluorescent microscopy alone diagnosed twelve cases, adding one GeneXpert test
which found 10 more cases, adding 33% incremental yield — ten out of 30;
running a second GeneXpert added four more cases (IY= 4/30 = 13%). Four cases
would be missed using just microscopy and the GeneXpert.
Since
the mid-morning specimen was sent for all four tests, it allowed a direct
comparison, an ‘intra-specimen comparison’ of test performance. There were 197
morning specimens with both one culture and one GeneXpert. A number of these
were either contaminated cultures (34) or non-tuberculosis mycobacteria (12) on
culture, leaving 151 comparable specimens — 16 of which were diagnosed with
TB. Of those, 13 cases were identified
by both tests. There were two patients that were positive on culture but negative
on GeneXpert, and one patient that was positive on GeneXpert but negative on
culture.
“There’s been some recent discussion
about the utility of doing a one-off test. It would certainly make things
cheaper for programmes, so we’re looking at the utility of a one-off test for
TB, using the more expensive technology of culture and nucleic acid
amplification,” said Dr Cavanaugh. The study’s early results are in the table
below, however, it should be noted that the analysis was restricted to
evaluable patients who had non-contaminated results — which would clearly affect
the utility of a one-off test in practice. Dr Cavanaugh noted that the
contamination rate was especially elevated at the beginning of this study but has since started coming down.
Performance of a single test
Test
|
Evaluable
participants
|
Sensitivity
(%)
|
Negative
predictive Value (NPV) (%)
|
Single
spot culture
|
206
|
25/29
(86%)
|
177/181
(98%)
|
Single
spot GeneXpert
|
220
|
22/32
(69%)
|
188/198
(95%)
|
Morning
culture
|
168
|
19/24
(79%)
|
144/149
(97%)
|
Morning
GeneXpert
|
185
|
17/25
(68%)
|
160/168
(95%)
|
As
for the performance of the screening tools? They all appeared to come up rather
short (see below), and would miss many cases that
would be picked up if all patients were simply providing specimens for culture
and GeneXpert.
Performance characteristics: recommended ICF screenings
Screening
Tool
|
Sensitivity
(%)
|
Specificity
(%)
|
NPV
(%)
|
PPV
(%)
|
WHO
(n
= 209)
|
21/31(68%)
|
100/178
(56%)
|
100/110
(91%)
|
21/99
(21%)
|
ID-TB/HIV
(n
= 211)
|
18/29 (62%)
|
111/182(61%)
|
111/122(91%)
|
18/89(20%)
|
Kenya
MoH
N
= 200
|
17/25
(68%)
|
102/175(58%
|
102/110(93%)
|
17/90
(19%)
|
Dr
Cavanaugh noted that the analysis had yet to be merged with data on important
clinical variables like chest radiographs, smear grade and CD4 counts.
It
is worth mentioning, however, that Kenya’s more complicated screen did perform
slightly better, suggesting that it may indeed be in the patients and the
programme’s best interests to adapt the symptom screen to regional
characteristics —so that the symptom screen reflects how local people living
with HIV describe or report their symptoms.
It
should be stressed that these are preliminary interim data with a small sample
size, so no firm conclusions should be drawn yet.
“But
this analysis suggests that the prevalence of TB in this population is quite
high – at least around 16%” said Dr Cavanaugh. “The proportion of patients
without symptoms is quite high – about 68% of the TB cases screened positive by
four-question screening questionnaire, so about 30% of the TB cases did not
report any of the four symptoms that we were investigating, and the negative
predictive value of all three guidelines is moderate, at best — though we need
to evaluate how chest radiography might influence results.”
Should
we move to laboratory screening of all people living with HIV or household
contacts?
“Although
it is too early to tell, a single sputum specimen
tested by culture, and then possibly by the
GeneXpert MTB/RIF assay, may have reasonable negative predictive value
– and that will be something that we’ll be looking at, as our data comes in,”
Dr Cavanaugh concluded.
He was not the only researcher at the
conference concerned about the lack of agreement of the symptom screens with sensitive diagnostic
tools such as GeneXpert and liquid culture diagnosed as TB. As noted in the
previous HATIP Dr
Adrienne Shapiro of Johns Hopkin’s University found
that the success of an active case-finding strategy was dependent upon the use
of TB culture.
“93% of the contacts were diagnosed solely on
the basis of TB culture, most were smear-negative. The contacts were largely
completely asymptomatic – only 11% had any symptoms. If we relied so
much on sputum smear, we would have found very little TB. So this also has
consequences for cost and feasibility, of future applications of active case finding or
household contact tracing,” she said.
One
potential implication regards the use of IPT, according to Dr Shapiro.
“The
results of this study suggest that there may be a fair amount of asymptomatic
culture-positive TB that would be inaccurately categorised as no TB. And these
patients are given IPT, which would not be appropriate since they would require
four-drug TB treatment. So, the screening policy may need to be adjusted for
extremely high incidence settings, or these results taken into account
when thinking about how to screen for IPT,” she said.
However,
this writer has noted that researchers consistently omit to mention one
critical aspect of the implementation of the WHO screening tool, which is that
screening is meant to be repeated on a monthly basis in anyone who is put on
IPT, so that any cases that escape detection before IPT is started should be
caught in one of the succeeding months if they are already in routine care. The
available clinical evidence suggests these cases respond to standard TB
treatment.
It
should be remembered that before the implementation of the ICF/IPT policy,
clinic staff rarely performed routine screening for HIV in PLHIV. Consequently,
there were many opportunities missed: many cases remained un-prevented,
undetected, undiagnosed, untreated — leading to increased transmission and
mortality.