A Population-Based Survey of Tuberculosis Symptoms: How Atypical Are Atypical Presentations?
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
293
A Population-Based Survey of Tuberculosis Symptoms: How Atypical
Are Atypical Presentations?
Loren G. Miller,1,4 Steven M. Asch,2 Emily I. Yu,2 From the Divisions of 1Infectious Diseases and 2General Internal
Laura Knowles,5 Lillian Gelberg,3 and Paul Davidson5 Medicine, Veterans Affairs Greater Los Angeles Healthcare System;
3
Department of Family Medicine, and 4Division of Infectious Diseases,
UCLA Medical Center; and 5Los Angeles County Tuberculosis
Control, Los Angeles County Department of Health, California
There is scant information on tuberculosis symptoms from a population-based perspective.
We prospectively identified 526 tuberculosis cases reported in Los Angeles County over a 6-
month period. Of 313 persons who completed our questionnaire, 72.7% had cough, 48.2% for
12 weeks, and 52.3% had fever, 29.4% for 12 weeks. Among those with pulmonary disease,
only 52.4% had cough for 12 weeks. In a multivariate model, persons with significant symptoms
typical of tuberculosis disease (defined as cough or fever for 12 weeks, weight loss, or he-
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
moptysis) were associated with lack of medical insurance, negative tuberculin skin test, di-
agnosis during a process other than screening, and non-Asian race. In summary, classic symp-
toms of prolonged cough and fever are insensitive predictors of tuberculosis. Our data suggest
that Asians may need to be added to the list of persons who present with tuberculosis atypically.
We believe that the Infectious Diseases Society of America guidelines for community-acquired
pneumonia should emphasize demographic features in addition to clinical symptoms when
suggesting which patients require evaluation for Mycobacterium tuberculosis.
Nearly 20,000 cases of tuberculosis are diagnosed annually In the United States, foreign-born persons comprise an increas-
in the United States and 19 million worldwide [1, 2]. Despite ing proportion of cases [1, 5] and HIV commonly coinfects
the widespread nature of this infection, there is only scant pop- persons with tuberculosis, potentially altering disease presen-
ulation-based information concerning symptoms of active tu- tation [6–8].
berculosis. We believe a description of symptoms serves 2 im- A second purpose of our study was to contrast the clinical
portant purposes. First, it provides a contemporary description presentation of tuberculosis among subgroups. Previous anal-
of tuberculosis symptoms that may be valuable to clinicians, yses have suggested that older persons present more often with
public health officials, and agencies that use symptom-based atypical disease and that they may have less fever and less cough
screening surveys to screen for active tuberculosis. Second, a more often than younger persons [9–11]. Identification of other
subgroup analysis will provide insight as to groups that may subgroups that present atypically will help clinicians and screen-
present atypically with the disease. ing agencies target groups that may need more aggressive eval-
Most case series concerning tuberculosis symptoms are lim- uation to uncover possible tuberculosis disease.
ited by methodology. Most descriptions have been limited to To obtain information on the symptoms of tuberculosis, we
1 organ system (for example pulmonary or renal tuberculosis) chose Los Angeles County as our study population. The county
or cases from a single hospital [3, 4]. These limitations may has the second largest number of tuberculosis cases of any
have resulted in biased clinical descriptions of the disease that reporting jurisdiction in the country [12] and has a diverse im-
may not be representative of larger, more diverse, and less se- migrant population. The county also has several well-structured
verely affected populations. In addition, the epidemiology of tuberculosis screening programs that help identify infection in
tuberculosis has changed, raising the question as to whether high-risk groups, such as homeless persons, HIV-infected in-
tuberculosis symptoms described in the past remain valid today. dividuals, and contacts of infectious cases. This allows us to
capture data on persons with tuberculosis who may never have
been hospitalized, a group rarely included in previous case series
Received 21 April 1999; revised 17 September 1999; electronically pub-
of tuberculosis symptoms. To answer our research question, we
lished 11 February 2000.
Financial support: This study was supported in part by the Department designed a study instrument (a questionnaire) to elicit data on
of Veterans Affairs Office of Academic Affiliations, Special Fellowship in symptoms from all persons diagnosed with tuberculosis over a
Ambulatory Care.
6-month period.
Reprints or correspondence: Dr. Loren G. Miller, Division of Infectious
Diseases, Harbor-UCLA Medical Center, 1124 West Carson St., Box 466,
Torrance, CA 90502 (lgmiller@humc.edu).
Methods
Clinical Infectious Diseases 2000; 30:293–9
q 2000 by the Infectious Diseases Society of America. All rights reserved. Study population. The law requires health care professionals,
1058-4838/2000/3002-0010$03.00 laboratories, and governmental agencies to report all suspected and294 Miller et al. CID 2000;30 (February)
confirmed cases of tuberculosis in Los Angeles County (excluding Results
the municipalities of Pasadena and Long Beach) to the Los Angeles
County Tuberculosis Control Registry. From these records, we pro- Description of persons with tuberculosis. Of the 526 persons
spectively identified 735 consecutive, confirmed cases of tubercu- eligible to be surveyed, 159 could not be contacted and 54
losis reported from April through September 1993. We excluded refused to be interviewed. Our questionnaire was completed by
children and people who were incarcerated or who had died, leaving 313 persons. A previous comparison of demographic variables
536 patients. Of these, 10 did not speak 1 of the 5 languages of between respondents and nonrespondents using this registry
our survey and were excluded. We attempted to locate and survey found no significant differences [17].
the remaining 526 persons. Of the 313 persons surveyed, all but 5 had tuberculosis dis-
Data collection. We sent potential respondents a letter describ- ease at a site listed in the registry. By use of World Health
ing our study and provided them with a phone number to call if
Organization (WHO) definitions [23], pulmonary disease was
they did not wish to participate. Trained bilingual interviewers
found in 245 (79.5%), extrapulmonary disease in 54 (17.5%),
called all eligible patients using the phone numbers provided by
the registry, clinic, directory assistance, or friends or relatives listed and both pulmonary and extrapulmonary disease in 9 (2.9%).
in medical records. Because telephone surveys miss homeless per- Among the 54 persons with extrapulmonary disease, sites of
sons, interviewers sought them at soup kitchens, shelters, and a infection included the lymphatic system (22 patients), pleura
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
specialized county clinic for the homeless. Interviewers made an (9), miliary disease (9), bone or joint (7), genitourinary tract
average of 16 attempts to contact each potential respondent. (5), peritoneum (3), meninges (2), and other (6). For all but 7
After giving informed consent, subjects responded in their pre- patients, information on chest radiographs was listed in the
ferred language. Interviews were conducted in 5 languages: English, registry; radiographs demonstrated cavitary disease in 58
Spanish, Mandarin, Tagalog, and Vietnamese. The interview used (19%).
a structured form (a questionaire) to elicit information on clinical Our patient population was predominantly male (64.3%),
symptoms and demographic information. Items were derived from
nonwhite (89.6%), and foreign-born (71.4%). Demographic and
previously validated instruments [13–16]. We abstracted demo-
clinical information are presented in table 1. Additional data
graphic and clinical information—radiographic and microbiologic
not included in the table include: 53.7% had an annual income
results, skin test results, and site of tuberculosis infection from the
county tuberculosis registry. Previous studies on the same registry 1$5000 per year, 66.9% had 112 years of education, 38% spoke
population showed good agreement on assessment of HIV status, English as a preferred language, 3.9% had previous psychiatric
acid-fast bacilli (AFB) smear status, and culture results among data hospitalization, 10.5% had a normal chest radiograph, and
abstracted from the registry, medical records, and laboratories [17]. 71.5% had cultures positive for M. tuberculosis.
To check the reliability of the interview data and to ensure quality The diagnosis of tuberculosis was made in 70.1% of persons
control, we readministered the instrument to 23 subjects during the after the patient presented for medical care with symptoms of
data collection process and compared registry and interview values tuberculosis (such as cough, fever, etc.) or symptoms of other
for variables that could be derived from both sources (age, sex, illnesses. The other 29.9% were diagnosed during medical
ethnicity, and country of birth) and found good agreement (k 1 screening. This group includes persons who were diagnosed at
0.8). immigration or employment screening programs, and during
Statistical analyses. We defined persons with significant clinical
the screening of high risk groups such as HIV-infected persons,
symptoms as those who had had 1 of the following symptoms at
the homeless, contacts of persons with active tuberculosis, and
any time during the preceding 2 years: cough for 12 weeks, fever
persons diagnosed during routine physical check-ups.
for 12 weeks, weight loss, or hemoptysis. We chose these significant
symptoms on the basis of literature that emphasizes they are either Table 2 shows the percentage of patients who had each symp-
typical of tuberculosis disease [18] or symptoms that should prompt tom among all patients, among those with pulmonary disease,
clinicians to consider this disease [19, 20]. We considered including and among those with extrapulmonary disease. Significant
other symptoms as significant, for example night sweats and fa- symptoms were present in 70.6% patients. Among the persons
tigue, but did not because the literature suggested such symptoms in our population, 72.7% had cough, 48.2% for 12 weeks. Fever
were too nonspecific and shared by many disease processes [19, was seen in 52.3% and in 29.4% for 12 weeks.
21]. Bivariate and multivariate analyses. Table 1 summarizes
Based on the medical literature [6, 7, 9] and clinical judgement, bivariate analysis of significant symptoms and the predictor
we preselected 20 variables that we hypothesized to be predictors variables. Significant symptoms were associated with (P ! .05)
of significant symptoms. We examined the association between the
ethnicity, being US born, younger age, unemployment, home-
predictor variables and significant symptoms using bivariate anal-
lessness, absence of health insurance, HIV infection, alcohol-
ysis (x2 test for categorical variables and two-tailed t test for con-
ism, drug use, positive smear results, negative tuberculin skin
tinuous variables). A multivariate logistic regression model to iden-
tify independent predictors of significant symptoms was created by test (TST), and diagnosis because of symptomatic disease. Sig-
including all predictor variables with a P ! .20 on bivariate analysis nificant symptoms were not associated with having a positive
and excluding highly correlated predictor variables. All statistical culture for AFB (P = .41), normal chest radiograph (P = .30),
analyses were performed by use of the SAS statistical package [22]. previous psychiatric hospitalization (P = .32 ), education (P =
P < .05 was considered statistically significant. .44), and speaking English as a primary language (P = .11). ToTable 1. Demographic and clinical data for 313 persons diagnosed with tuberculosis (TB).
Bivariate analysis Multivariate model
No. (%)
No. (%) with significant
a b c
Data overall symptoms P OR (95% CI) P
Demographic
Sex
Male 198 (64.3) 142 (72.8) .16 .48
Female 110 (35.7) 71 (65.1)
Ethnicity
Asian 74 (23.9) 30 (42.2) !.001 0.17 (0.08–0.39) !.0001
Black 38 (12.3) 28 (73.7) .43
White 32 (10.4) 25 (78.1) .84
Hispanic 153 (49.5) 124 (81.0) .96
Other 12 (3.9) 8 (72.7)
US born
Y 86 (28.6) 67 (78.8) .05 .85
N 215 (71.4) 143 (67.4)
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
Age, y
!31 84 (26.9) 64 (77.1) .01 .27
31–60 169 (54.2) 121 (72.5)
160 59 (18.9) 32 (55.2)
Unemployed
Y 145 (48.0) 107 (75.4) .05 .66
N 157 (52.0) 101 (64.7)
Medical insurance
Y 140 (48.8) 91 (65.5) !.001
N 147 (51.2) 120 (82.8) 3.60 (1.65–7.83) .001
Homeless
Y 26 (8.6) 23 (88.5) .03 .71
N 278 (91.4) 188 (68.6)
Previous incarceration
Y 83 (27.3) 65 (78.3) .06 .74
N 221 (72.7) 146 (67.3)
Clinical
TB diagnosed because of
Symptoms 216 (70.1) 177 (81.9) !.001 5.89 (2.69–12.89) !.0001
Screening 92 (29.9) 37 (42.0)
HIV infection
Y 37 (12.1) 32 (86.5) .02 .53
N 268 (87.9) 180 (67.9)
Alcoholism
Y 83 (27.3) 66 (80.5) .02 .91
N 221 (72.7) 145 (66.5)
Drug use
Y 46 (15.2) 41 (89.1) .003 .26
N 257 (84.8) 170 (67.2)
Skin test
Positive 218 (83.5) 144 (67.0) !.001 6.40 (1.36–30.01) .02
Negative 43 (16.5) 39 (92.9)
Smear results
Positive 136 (44.9) 106 (78.5) .006 .37
Negative 167 (55.1) 105 (64.0)
Cavitary chest radiograph
Present 58 (19.1) 44 (78.6) .14 .92
Absent 246 (80.9) 167 (68.4)
NOTE. Not all patients answered each question; therefore, total no. of responses to each question
may not equal 313. Items not listed in this table, not associated with significant symptoms (P > .20),
and not included in the multivariate model were as follows: education (P = .44 ), normal chest radiography
(P = .30), previous psychiatric hospitalization (P = .32 ), and acid-fast bacilli (AFB) culture positivity
(P = .41). There was no statistically significant association with English speaking (P = .11 ), but this was
not included in the final multivariate model (see text).N, no; Y, yes.
a
Significant symptoms were defined as cough or fever for 12 weeks, weight loss, or hemoptysis.
Significant symptoms were found in 70.6%.
b
Comparison by use of t test for continuous variables and x2 test for categorical variables.
c
Calculated with stepwise multivariate logistic regression analysis model.296 Miller et al. CID 2000;30 (February)
Table 2. Percentage of tuberculosis patients with specific symptoms. of tuberculosis symptoms, we sought to identify all other pub-
Pulmonary Extrapulmonary lished cohorts or cross-sectional studies of tuberculosis by
All patients disease disease
a searching MEDLINE citations for the period from 1966
(n = 313) (n = 254) (n = 54)
through February 1999. We made additional attempts to find
Significant symptoms 218 (70.6) 176 (70.4) 37 (68.5)
Cough 226 (72.7) 191 (75.8) 30 (55.6)
a population-based description of symptoms from the literature
Present for 12 w 150 (48.2) 132 (52.4) 15 (27.8) and a review of texts dating back to the late nineteenth century
Fever 162 (52.3) 127 (50.6) 32 (59.3) at the libraries of the Los Angeles County Department of
Present for 12 w 91 (29.4) 66 (29.3) 23 (42.6)
Fatigue 185 (59.7) 147 (58.6) 35 (64.8) Health and the University of California, Los Angeles. To our
Present for 12 w 127 (41.0) 105 (41.8) 20 (37.0) knowledge, our study is the only true population-based de-
Weight loss 138 (44.5) 108 (43.0) 27 (50.0)
scription of tuberculosis symptoms, with the exception of a
Sweats 148 (47.9) 116 (46.0) 28 (53.9)
Present for 12 w 91 (29.4) 72 (28.6) 19 (36.5) study by Gnaore et al. [7], and the first such description from
Anorexia 127 (40.6) 102 (40.2) 22 (40.7) the United States. The Gnaore study is of limited use because
Present for 12 w 81 (25.9) 66 (26.0) 15 (27.8)
Chest pain 128 (41.0) 105 (41.5) 22 (40.7)
it lacks a detailed description of the method used to collect
Present for 12 w 87 (27.9) 71 (28.1) 15 (27.8) information on tuberculosis symptoms. In addition, it was per-
Diarrhea 69 (22.0) 54 (21.3) 15 (27.8) formed primarily to contrast tuberculosis infection among
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
Present for 12 w 38 (12.1) 27 (10.6) 11 (20.4)
Hemoptysis 65 (20.9) 80 (23.8) 3 (5.6) those with and without HIV infection.
We found 3 additional studies that described tuberculosis
NOTE. Data are no. (%). Total responses to each question does not equal
313 because not all patients answered each question. However, the no. of non- symptoms in persons with both pulmonary and extrapulmon-
responders did not exceed 4 persons for any question about symptoms. For 5 ary disease [3, 4, 6]. These studies, as well as the study by
patients, registry information about site of tuberculosis disease was missing and
were not included in the subgroup analysis by site of disease.
Gnaore et al., are summarized in table 3. To contrast our results
a
Includes 9 patients with both pulmonary and extrapulmonary disease. in a more meaningful manner, we present the prevalence of
symptoms of tuberculosis normalized to our 12.1% rate of HIV
infection in the 2 studies that were undertaken to contrast symp-
determine whether significant symptoms were associated with
toms among HIV-positive and HIV-negative patients [6, 7].
the duration of time between disease diagnosis and interview
One interesting finding of our study was that the proportion
date, we compared this duration among persons with and with-
out significant symptoms and found no difference (58 days vs. of persons with tuberculosis who had the significant symptoms
63 days; P = .33, two-tailed t test). of prolonged (12 weeks) cough and fever was smaller than we
Results of multivariate analysis are summarized in table 1. anticipated and small in comparison to previously published
(The model excluded 1 explanatory variable, speaking English studies. In the other reports described in table 3, the prevalence
as a primary language, which was highly correlated with an- of prolonged (12 weeks) cough ranged from 42%–89% and
other explanatory variable, birth in the U.S.A., and which had prolonged fever from 23%–68%. In our study, only 48.2% and
a higher rate of missing values). In our model, independent 29.4% had prolonged cough and fever, respectively. Differences
predictors of significant symptoms were as follows: lack of in symptom prevalence noted in the table might reflect local
health insurance (OR, 3.60; 95% CI, 1.65–7.83; P = .001), neg- variation in disease. However, our rate of extrapulmonary dis-
ative TST (OR, 6.40; 95% CI, 1.36–30.01; P = .02), and diag- ease is similar to that of the studies in table 3 and that in other
nosis because of symptomatic disease (in contrast to diagnosis locales [3, 6, 7, 24–26]. The rate of cough in our study may
during some more general screening; OR, 5.89; 95% CI, also reflect the inclusion in our population of persons treated
2.69–12.89; P ! .0001) ). Asian ethnicity was a predictor of lack as outpatients, a group almost invariably ignored in previous
of significant symptoms (OR, 0.17; 95% CI, 0.08–0.39; P ! symptom studies. Our results suggest that atypical symptoms
.0001). With use of empirical imputation for missing values, of tuberculosis disease may be more common than previously
these variables maintained their statistical significance and held.
younger age also became an independent predictor of signifi- Our study has several strengths. First, our population is a
cant symptoms (OR, 1.02; 95% CI, 1–1.04; P = .01). large, diverse group with a wide range of ages, ethnicities, coun-
tries of origin, routes of diagnosis, and clinical symptoms. These
factors allowed us to make group comparisons not previously
Discussion
reported by other studies. Virtually all studies on tuberculosis
We sought to describe tuberculosis symptomatology from a symptoms have compared symptoms among patients with and
population-based perspective. This is an important viewpoint without a single characteristic (e.g., age or HIV infection), with-
because many clinical descriptions of tuberculosis are case series out controlling for confounders; yet these studies have at-
of hospitalized patients and thus may be biased toward those tempted to draw conclusions about groups that present atyp-
with more severe disease. ically with tuberculosis [6, 7, 9, 27]. Our large and diverse
To contrast the results of our survey with other descriptions sample of 313 patients allowed us to control for multiple dem-CID 2000;30 (February) Tuberculosis and Atypical Presentations 297
Table 3. Comparison of clinical features of tuberculosis from the literature.
Current study Arrango [3] MacGregor [4] Elliott [6] Gnaore [7]
Year of study 1993 1970–1971 1971–1972 1989 1989–1990
Country US (Los Angeles) US (San Francisco) US (Philadelphia) Zambia Ivory Coast
Focus of study Population-based Urban hospital Postsanitory tuberculosis HIV HIV
No. of patients 313 95 41 249 4504
HIV 12.1 0 0 73.1 43.7
a c,d c
Cough 48.2 42 81 61.7 (67.8) 79.9 (80.7)
a b c,d c
Fever 29.4 37 44 36.7 (23.5) 69.2 (67.9)
c
Weight loss 44.5 7 75 — 74.3 (70.3)
Hemoptysis 20.9 9 24 — —
c
Lymphadenopathy 13.2 — — 51.4 (20.1) —
c
Diarrhea 12.1 — — — 8.1 (5.2)
NOTE. Data are percentages unless otherwise indicated.
a
Symptoms lasting at least 2 weeks.
b
Fever not distinguished from sweats.
c
Nos. in parentheses represent symptoms normalized for 12.1% HIV rate.
d
Symptoms lasting at least 1 month.
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
ographic and clinical characteristics and to identify independent populations [28]. Although we were concerned that we might
predictors of atypical disease presentation. have selected for a subpopulation of patients that was not rep-
An additional strength of our study is that we used a stan- resentative of tuberculosis patients in the county, a previous
dardized interview to collect information about symptoms. Al- comparison of demographic variables of respondents and non-
though obtaining clinical information from medical records has respondents from the county registry found no significant dif-
advantages (e.g., it describes patients’ complaints at the time ference [17]. Our data are also limited because we excluded all
of presentation for medical care), there are limitations to this persons who had died, which may have biased our study toward
method. Clinicians may obtain clinical information but not less symptomatic patients. However, this may have only a slight
document it. In addition, some clinical information may not impact on our results: in the 2 studies described in table 3 that
be pursued by clinicians. For example, a chart omitting infor- did not have high rates of HIV infection, mortality rates were
mation on weight loss may indicate either that the patient did low (!5%) [3, 4].
not lose weight or that this information was not collected. Our Our study has implications for clinicians. For patients with
data set contains information about cough, fever, and other pneumonia or an unexplained illness, clinicians should not as-
clinical symptoms on most consenting patients and appears sume tuberculosis is ruled out because prolonged cough or fever
valid when compared with that of previously published chart- is absent. The Infectious Diseases Society of America’s (IDSA)
based studies. For example, in the bivariate analysis, we found guidelines for community-acquired pneumonia states that test-
that persons with HIV were more symptomatic than those with- ing for M. tuberculosis should be done in “selected patients,
out and that older persons with tuberculosis were less likely to especially those with cough for 11 month, other common symp-
be symptomatic; these findings are similar to those of other toms, or suggestive radiographic changes” [20]. In our popu-
reports [6, 9, 27]. lation, the proportion of those with pulmonary tuberculosis
There are limitations to using information obtained from a who had cough for 12 weeks was only 52.4%, and almost cer-
standardized interview after a diagnosis of tuberculosis has been tainly the proportion of those with cough for 11 month would
established has limitations. In our study methodology may have be significantly lower. Although we were unable to find pop-
allowed patients to describe their symptoms differently than ulation-based studies of pulmonary tuberculosis, recent hos-
they would have at the time of disease presentation. Some mild pital-based studies have demonstrated cough rates of 50%–77%
fatigue initially unreported during disease presentation may [9, 27, 29]. This supports the hypothesis that a significant pro-
have taken on a larger significance later when the patient had portion of patients with tuberculosis, including those with pul-
had time to reflect that they had a serious infection. Conversely, monary disease, lack prolonged cough.
some patients might have forgotten some symptoms over time. We believe that the IDSA guidelines should be restated. Our
However, the time from confirmation of tuberculosis disease data suggest that cough for 11 month is too insensitive a marker
and questionnaire completion did not differ among those with to trigger an evaluation for pulmonary tuberculosis. We suggest
and without significant symptoms. that demographic data may help identify pulmonary tubercu-
Our study has additional limitations. We were unable to in- losis more sensitively. For example, our population has pro-
terview all 526 eligible patients. However, we made concerted portions of patients who are homeless (8.6%) and infected with
efforts to locate homeless persons with tuberculosis and to con- HIV (12.1%) that far exceed those in the general county pop-
duct interviews with persons in their native language, and our ulation [30]. We suspect these and other demographic factors
level of response (60%) is common in studies of vulnerable (e.g., recent incarceration or foreign birth in an area of endem-298 Miller et al. CID 2000;30 (February)
icity) may be useful markers to trigger an evaluation for tu- non-Asians were found (P 1 .05 for each comparison, data not
berculosis in a patient with pneumonia. Unfortunately our data shown).
set cannot address the specificity of these demographic markers. The atypical presentation of tuberculosis among Asians
The surprisingly low rate of persons lacking typical significant might be explained by physiologic or cultural differences among
symptoms of tuberculosis may in part explain the robust lit- ethnic groups. It has been demonstrated that Asians have dis-
erature on delays in suspicion of tuberculosis disease in hos- tinctive responses to specific stimuli or infections [37, 38]. In
pitalized patients [31–33]. addition, there is evidence suggesting that some Asians report
Our data also provide some insight as to the sensitivity of pain differently than other ethnic groups [39] and so might
symptom screening for tuberculosis. The Centers for Disease report symptoms later. Furthermore, cultural differences in def-
Control and Prevention (CDC) recommends that screening pro- initions of fever and cough may have explained our observa-
grams that evaluate persons at risk for tuberculosis should in- tions. Whether the symptomatic differences in Asians with tu-
clude questioning for symptoms suggestive of tuberculosis dis- berculosis represent physiologic or cultural differences (or both)
ease [34, 35]. Many organizations incorporate symptom-based is unclear, although we suspect the latter plays a more important
questionnaires into their screening programs. For example, the role.
US Federal Bureau of Prisons screens new inmates for tuber- The nature of the Asian population included in our study
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
culosis disease by performing skin testing and conducting in- may have also affected our results. This population was largely
terviews that ask about symptoms suggestive of tuberculosis foreign born (195%), and most were from the Philippines, Vi-
disease (N. A. Kendig, Federal Bureau of Prisons, Washington etnam, Korea, China, or Taiwan. Therefore, this group may
D.C., personal communication). Symptom screening is also im- not be representative of Asian populations originating or living
portant for other high-risk settings, such as homeless shelters, elsewhere, and extrapolating atypical tuberculosis presentations
drug-treatment centers, emergency rooms, and immigration to American-born Asians and those other locations may be
processing centers. The validity of such symptom-based screen- dubious. In addition, the relationship between Asian ethnicity
ing techniques has never been examined. If we had administered and atypical symptoms may be confounded by clinical infor-
the tuberculosis symptom screener used by Los Angeles County mation we did not have access to, such as previous Bacille
[36] to this population, 75% would have had a positive survey Calmette-Guérin (BCG) administration. Our observations sug-
and 99% would have had either a positive symptom screener gest that persons of Asian origin may need to be added to the
or TST (data not shown). We can provide insight neither on list of groups that present with tuberculosis atypically, although
the screener’s specificity nor on its sensitivity for those pre- this assertion would be strengthened by similar results in other
senting earlier with the disease. Further research on high-risk multiethnic populations with a significant proportion of Asians.
persons with and without tuberculosis needs to be done to In summary, we have described contemporary clinical pres-
answer these questions. entations and demographics of persons with tuberculosis in an
Our subgroup analysis provided several interesting obser- urban American city. Cough and fever were surprisingly less
vations on groups that present differently. Lacking health in- common than expected, suggesting that atypical tuberculosis
surance was an independent predictor of having significant presentation may be more common than previously held. This
symptoms. This may reflect that lack of insurance acts as a finding has implications for clinicians who we believe should
barrier to care and delays the diagnosis until the infected person not be dissuaded from dismissing tuberculosis as a possibility
becomes even more symptomatic. We also observed that those in those who lack more “classical” symptoms of the disease
with a negative skin test were more likely to be symptomatic, (e.g., prolonged cough or fever). Our data also suggest that
perhaps reflecting either preexisting impaired cell-mediated im- screening programs that use symptom surveys in combination
munity (CMI) or waning CMI from overwhelming infection with TST are very sensitive for finding tuberculosis, but that
[28]. without adding the TST these surveys may miss significant clin-
The association between Asian ethnicity and lack of signif- ical disease. Further research needs to be performed to evaluate
icant symptoms was surprising. Asians were less likely to have the sensitivity and specificity of symptom-based screening for
significant symptoms of tuberculosis even when controlling for tuberculosis in persons at high-risk for the disease. The strong
confounders in our multivariate model. Two possible expla- association of Asian ethnicity with lack of significant symptoms
suggests that Asians may need to be added to the list of groups
nations for this phenomenon are that recent Asian immigrants
that present atypically, although this result needs to be con-
are more likely to be diagnosed early, during screening at im-
firmed in other populations of patients with tuberculosis.
migration, and that Asians have a lower rate of HIV infection
than other large ethnic groupings [40]. Although our model Acknowledgments
controlled for HIV as well as tuberculosis diagnosed during
screening programs, such as those performed by immigration We are grateful for the advice and support of the following persons
authorities, we still found significant differences. No differences for their contributions to the study and feedback on early drafts of
in age or rates of extrapulmonary disease between Asians and the manuscript: the staff at Los Angeles County Tuberculosis Control,CID 2000;30 (February) Tuberculosis and Atypical Presentations 299
Barbara Leake, Mike Wada, Lorraine Chun, Ron Anderson, Matthew mocystis carinii pneumonia, bacterial pneumonia and tuberculosis in HIV-
Goetz, Martin Shapiro, and the fellows in the Health Services Research infected patients. AIDS 1998; 12:885–93.
community at UCLA and its affiliated hospitals. 20. Bartlett JG, Breiman RF, Mandell LA, File TM. Community-acquired pneu-
monia in adults: guidelines for management. The Infectious Diseases So-
ciety of America. Clin Infect Dis 1998; 26:811–38.
References 21. Smetana GW. Diagnosis of night sweats. JAMA 1993; 370:2502–3.
22. SAS users guide—basics 1E. Cary, NC: SAS Institute, 1990.
1. Centers for Disease Control and Prevention. Tuberculosis morbidity: United 23. World Health Organization. Framework for effective tuberculosis control.
States 1997. MMWR Morb Mortal Wkly Rep 1998; 47:253–7. Publication no. WHO/TB/94.17. World Health Organization: Geneva,
2. World Health Organization. Treatment of tuberculosis: guidelines for national 1994.
programmes. Publication no. WHO/TB/97.220. World Health Organiza- 24. Calder L, Priest P. One year of tuberculosis in Auckland. N Z Med J 1996;109:
tion: Geneva, 1997. 164–7.
3. Arango L, Brewin AW, Murray JF. The spectrum of tuberculosis as currently 25. Byram D, Hatton P, Williams S, Pearson SB. The form and presentation of
seen in a metropolitan hospital. Am Rev Respir Dis 1973; 108:805–12. tuberculosis over a 10-year interval in Leeds. Br J Dis Chest 1985; 79:
4. MacGregor RR. A years’ experience with tuberculosis in a private urban 152–60.
teaching hospital in the postsanatorium era. Am J Med 1975; 58:221–8. 26. Page MI, Lunn JS. Experience with tuberculosis in a public teaching hospital.
5. Centers for Disease Control and Prevention. Tuberculosis morbidity: United Am J Med 1984; 77:667–70.
States 1996. MMWR Morb Mortal Wkly Rep 1997; 46:695–700. 27. Katz PR, Reichman W, Dube D, Feather J. Clinical features of pulmonary
Downloaded from http://cid.oxfordjournals.org/ by guest on October 12, 2015
6. Elliott AM, Halwiindi B, Hayes RJ, et al. The impact of human immuno- tuberculosis in young and old veterans. J Am Geriatr Soc 1987; 35:512–5.
deficiency virus on presentation and diagnosis of tuberculosis in a cohort 28. Bindman AB, Grumbach K, Keane D, Lurie N. Collecting data to evaluate
study in Zambia. J Trop Med Hyg 1993; 96:1–11. the effect of health policies on vulnerable populations. Fam Med 1993;
7. Gnaore E, Sassan-Morokro M, Kassim S, et al. A comparison of clinical 25:114–9.
features in tuberculosis associated with infection with human immuno- 29. Herer B, Kuaban C, Papillion F, Durieux P, Chretien J, G.Huchon G. Fea-
deficiency viruses 1 and 2. Trans Royal Soc Trop Med Hyg 1993; 87:57–9. tures in hospitalized patients with symptom-detected or radiologicially-
8. Schoch OD, Reider HL. Characteristics of sputum smear-positive tubercu- detected pulmonary tuberculosis. Eur Respir J 1989; 2:3–6.
losis patients with and without HIV infection in a hospital in Zimbabwe. 30. Field Research Corporation. 1997 Los Angeles County health survey: survey
Eur Respir J 1996; 9:284–7. methodology. Los Angeles: Field Research Corporation, 1997.
9. Korzeniewska-Kosela M, Krysl J, Muller N, Black W, Allen E, Fitzgerald 31. Mathur P, Sacks L, Auten G, Sall R, Levy C, Gordin F. Delayed diagnosis
JM. Tuberculosis in young adults and the elderly: a prospective compar- of pulmonary tuberculosis in city hospitals. Arch Intern Med 1994; 154:
ison study. Chest 1994; 106:28–32. 306–10.
10. Morris CD. Pulmonary tuberculosis in the elderly: a different disease? Thorax 32. Enarson DA, Grzybowski S, Dorken E. Failure of diagnosis as a factor in
1990; 45:912–3. tuberculosis mortality. Can Med Assoc J 1978; 118:1520–2.
11. Nagami P, Yoshikawa TT. Aging and tuberculosis. Gerontology 1984; 30: 33. Counsell SR, Tan JS, Dittus RS. Unsuspected pulmonary tuberculosis in a
308–15. community teaching hospital. Arch Intern Med 1989; 149:1274–8.
12. Belanaon A. Los Angeles County tuberculosis control: trends in tuberculosis 34. Centers for Disease Control and Prevention. Guidelines for preventing the
morbidity in Los Angeles County. 1980–91. Los Angeles: Department of transmission of Mycobacterium tuberculosis in health-care facilities, 1994.
Health Services, 1992. MMWR Morb Mortal Wkly Rep 1994; 43(RR-13):1–132.
13. Andersen R, Aday LA. Access to medical care in the US: realized and po- 35. Centers for Disease Control and Prevention. Prevention and control of tu-
tential. Med Care 1978; 7:533–46. berculosis in correctional facilities: recommendations of the advisory
14. Bloom B, Jack SS. Persons with and without a regular source of medical council for the elimination of tuberculosis. MMWR Morb Mortal Wkly
care. Vital Health Stat 10 1985; 151:1–46. Rep 1996; 45(RR-8):1–27.
15. Verneaz G, Burnam MA, McGlynn EA, et al. Review of California’s program 36. Los Angeles County Tuberculosis Control. Document TB-35. Los Angeles,
for the homeless and mentally disabled. Santa Monica: Rand Corporation, 1995.
1988. 37. Stern RM, Hu S, Uijtdehaage SH, Muth ER, Xu LH, Koch KL. Asian
16. Blendon RJ, Aiken LH, Freeman HE, Corey CR. Access to medical care hypersusceptibility to motion sickness. Human Heredity 1996; 46:7–14.
for black and white Americans: a matter of continuing concern. JAMA 38. Gray GC, Fogle EF, Albright KL. Risk factors for primary pulmonary coc-
1989; 261:278–81. cidioidomycosis hospitalizations among United States Navy and Marine
17. Asch S, Leake B, Gelberg L. Does fear of immigration authorities deter Corps personnel, 1981–1994. Am J Trop Med Hyg 1998; 58:309–12.
tuberculosis patients from seeking care? West J Med 1994; 16:373–6. 39. Faucett J, Gordon N, Levine J. Differences in postoperative pain severity
18. Danil TM. Tuberculosis. In: Isselbacher KK, Braunwald E, Wilson JD, et among four ethnic groups. J Pain Symptom Manage 1994; 9:383–9.
al., eds. Harrison’s principles of internal medicine. 13th ed. New York: 40. Haverkos HW, Turner JF Jr, Moolchan ET, Cadet JL. Relative rates of AIDS
McGraw-Hill, 1994:71–8. among racial/ethnic groups by exposure categories. J Natl Med Assoc
19. Selwyn PA, Pumerantz AS, Durante A, et al. Clinical predictors of Pneu- 1999; 91:17–24.You can also read
