Marian Goble, Michael D. Iseman, Lorie A. Madsen, Dennis Waite, Lynn Ackerson, and C. Robert Horsburgh ABSTRACT Background and Methods The frequency of infection with multidrug-resistant
Mycobacterium tuberculosis is increasing. We reviewed the clinical
courses of 171 patients with pulmonary disease due to M. tuberculosis
resistant to rifampin and isoniazid who were referred to our
hospital between 1973 and 1983. The patients' records were analyzed
retrospectively. Their regimens were selected individually and
preferably included three medications that they had not been
given previously and to which the strain was fully susceptible.
Results The 171 patients (median age, 46 years) had previously
received a median of six drugs and shed bacilli that were resistant
to a median of six drugs. Thus, their regimens were frequently
not optimal. Of 134 patients with sufficient follow-up data,
87 (65 percent) responded to chemotherapy (as indicated by negative
sputum cultures for at least three consecutive months); 47 patients
(35 percent) had no response, as shown by continually positive
cultures. The median stay in the hospital was more than seven
months. In a multivariate analysis, an unfavorable response
was significantly associated with a greater number of drugs
received before the current course of therapy (odds ratio, 4.0;
95 percent confidence interval, 1.6 to 9.9; P<0.001) and
with male sex (odds ratio, 2.5; 95 percent confidence interval,
1.1 to 6.2; P<0.03). Twelve of the patients with responses
subsequently had relapses. The overall response rate was 56
percent over a mean period of 51 months. Of the 171 patients,
63 (37 percent) died, and 37 of these deaths were attributed
to tuberculosis.
Conclusions For patients with pulmonary tuberculosis that is
resistant to rifampin and isoniazid, even the best available
treatment is often unsuccessful. Only about half of such patients
eventually have negative sputum cultures despite carefully selected
regimens administered for extended periods. Failure to control
this resistant infection is associated with high mortality and
ominous implications for the public health.
Modern chemotherapy, appropriately prescribed and administered,
cures 98 to 99 percent of cases of previously untreated pulmonary
tuberculosis if the disease is caused by drug- susceptible Mycobacterium
tuberculosis 1 . Tubercle bacilli not previously exposed to antituberculosis
drugs may manifest drug resistance as a result of random mutation 2 .
Whereas such resistance can be overcome by appropriate multiple-drug
regimens, inadequate antimicrobial therapy can lead to the emergence
of drug-resistant strains. These resistant bacilli can in turn
be passed on to previously uninfected persons, who may subsequently
contract tuberculosis that is initially drug-resistant. Early
in the chemotherapeutic era, 1 to 3 percent of patients were
noted to have drug-resistant organisms, typically resistant
to a single drug 3,4 . During the period 1982 through 1986, however,
8.8 percent of previously untreated patients in the United States
were excreting M. tuberculosis resistant to at least one antituberculosis
drug, and 23 percent of previously treated patients were shedding
tubercle bacilli resistant to one or more medications 5 .
The majority of cases of resistance involve resistance to isoniazid
and streptomycin. Until recently, resistance to rifampin has
been relatively infrequent 6 . Introduced in the United States
in 1971, rifampin has been prescribed widely, and resistance
to it has emerged. Resistance to this drug was found in 0.6
percent of strains from previously untreated patients and 3.3
percent of strains from previously treated patients in the United
States during the period 1982 through 1986 5 . Some regions of
the country have a higher prevalence of resistance 7,8 .
Resistance to rifampin is of extraordinary importance in modern
chemotherapy. Rifampin and isoniazid are clearly the two most
active antituberculosis drugs. The availability of rifampin
allows successful and well-tolerated treatment in patients with
organisms resistant to isoniazid, streptomycin, or other agents 9,10,11,12 .
However, when resistance to rifampin occurs in the presence
of resistance to isoniazid, the prospects for successful chemotherapy
are greatly diminished 6 . In the United States, several outbreaks
of tuberculosis due to strains resistant to isoniazid, rifampin,
and other agents have recently been reported, in both patients
with and patients without infection with the human immunodeficiency
virus (HIV) 13,14,15 .
We present here our experience in treating a large series of
patients with pulmonary tuberculosis resistant to both rifampin
and isoniazid, and we discuss factors influencing outcomes and
the implications for both the individual patient and the future
of tuberculosis control.
Methods The records of 231 patients hospitalized at our institution
between January 1, 1973, and December 31, 1983, with pulmonary
tuberculosis due to drug-resistant M. tuberculosis were reviewed
retrospectively. All 171 patients shedding M. tuberculosis resistant
to rifampin are discussed in this report. One of us was actively
involved in the direct care of these patients at the center
during the entire study period.
Susceptibility testing of all M. tuberculosis strains was performed
in our laboratory on 7H11 agar according to standard proportionality
techniques 2 . Resistance was indicated by the growth of more
than 1 percent of the colonies on drug-containing medium, as
compared with growth on drug-free (control) medium (Table 1).
A drug was designated as "previously unused" if given for less
than three months. Individually tailored regimens 16,17 were
selected for the patients on the basis of the results of in
vitro susceptibility tests and the limited previous use or the
nonuse of the drugs. The drugs and the dosages used are listed
in Table 1. When possible, we administered three drugs not given
previously to which the tubercle bacilli were fully susceptible
at the lowest concentration tested in vitro, including one parenteral
agent (an aminoglycoside or polypeptide) and two oral agents
that met these criteria. Since many patients did not have such
a regimen available to them, we also used drugs to which the
organisms were at least partially susceptible (no growth at
the higher concentrations tested, or 2 to 33 percent of control
value for growth at the lowest concentration) or drugs previously
given for a relatively short time; because of a lack of confidence
in the full efficacy of drugs so selected, more than three medications
were included in these regimens. The 171 patients received a
median of four drugs; 32 patients received six or more drugs.
View this table: Table 1. Concentrations of Drugs Tested and Dosages Administered in 171 Patients with Tuberculosis. [in this window] [in a new window]
To ensure compliance, nurses observed the patients while they
took their medications. Ethionamide, pyrazinamide, aminosalicylic
acid (para-amino salicylate), and cycloserine were given in
divided doses to increase drug tolerance. Great efforts were
made to continue therapy even if adverse drug reactions other
than those considered life-threatening occurred. Patients usually
remained in the hospital until a series of sputum cultures remained
negative for at least three months or until treatment failure
was evident after at least three months of therapy, as shown
by continually positive sputum cultures. Success was indicated
by three consecutive months of consistently negative cultures.
While the patients were hospitalized, spontaneous or induced
sputum specimens were examined every one to two weeks by fluorochrome
staining of the concentrated specimen and culture on 7H11 agar
and Lowenstein-Jensen medium. If patients resumed shedding tubercle
bacilli after their cultures had been negative for three consecutive
months, they were considered to have relapsed, whether or not
they were still receiving therapy. After hospitalization, they
returned to the care of the referring physician or agency, with
the recommendation that oral antimycobacterial therapy should
continue for at least two years after the last positive culture
and that injectable antibiotics be given for four to six months
after the initial date of culture conversion.
For this analysis, data on all 171 patients were reviewed, including
(2), and gastrectomy (1). During their initial medical encounter
before referral to our center, three patients presented with
organisms already resistant to rifampin, including two nurses
who presumably acquired these organisms from patients for whom
they were caring. Factors relating to the emergence of resistance
to rifampin could not be identified in 29 patients.
Duration of Hospitalization The median stay in the hospital was 7.3 months for patients
whose treatment was successful and 7.8 months for those whose
treatment failed.
Adverse Reactions Fifty-one (30 percent) of 171 patients had adverse reactions
that led to the discontinuation of one or more antimycobacterial
medications (Table 2). Twenty of these patients had serious
toxic reactions to one or more drugs during the first three
months of therapy.
View this table: [in this window] [in a new window]
Table 2. Drug Toxicity Leading to Withdrawal of Treatment.
Adjunctive Surgery Nine patients underwent surgical procedures as an adjunct to
their antituberculosis chemotherapy. Three patients underwent
right upper lobectomy, two left upper lobectomy, and two left
pneumonectomy. The sputum cultures of all of these seven subsequently
became negative. (Two patients had consistently negative cultures
before surgery and were considered to have responded to chemotherapy,
three had decreases in the number of tubercle bacilli in their
sputum, and two had no bacteriologic response to chemotherapy
alone.) The cultures of the two other patients remained positive;
one patient underwent surgery to collapse a left-upper-lobe
cavity with paraffin plombage, and in the other patient an Eloesser
flap was created for drainage. The only surgical complication
was a bronchopleural fistula, which resolved spontaneously by
two months after the operation.
Exclusions from the Outcome Analysis Of the 171 patients, 37 were excluded from the analysis of the
outcome of chemotherapy for the following reasons: 8 died after
receiving therapy for two months or less; 22 were lost to follow-up
before they were observed long enough during chemotherapy; 3
did not meet the criterion of three months of negative sputum
cultures before surgical resection, although they had had favorable
responses; 1 deferred definitive therapy because of pregnancy;
and 3 had had no positive cultures within the month before therapy
started.
Outcomes of Chemotherapy One hundred thirty-four patients were eligible for the analysis
of the outcomes of chemotherapy. The sputum cultures of 87 of
these patients (65 percent) became negative. The interval from
the beginning of therapy to the first of a series of negative
cultures ranged from one to eight months (median, two). Treatment
failure, defined as the inability of chemotherapy to sterilize
cultures, occurred in 47 patients (35 percent). Of the 87 patients
who originally had negative cultures, 12 (14 percent) had relapses
5 to 62 months after the initiation of therapy, and 9 of these
12 had relapses within the first 2 years. Overall, 75 patients
(56 percent) remained free of disease throughout a mean follow-up
period of 51 months (range, 10 to 167), and 59 patients (44
percent) had unfavorable outcomes.
Deaths Death due to tuberculosis occurred in 27 of the 59 patients
(46 percent) with treatment failure or a relapse. The mean interval
from the initial diagnosis to death was 11 years (median, 13),
and the mean age at death was 52 years (range, 26 to 74). Four
other patients with treatment failure died of causes unrelated
to tuberculosis. Of the 87 patients with successful treatment,
15 died during the follow-up period; 13 died of causes not related
to tuberculosis, and 2 died of undetermined causes. The estimated
survival of the patients who could be evaluated and for whom
the cause of death was known is shown in Figure 2. Among the
37 patients excluded from the analysis of the outcomes of chemotherapy,
10 (27 percent) died of tuberculosis and 7 (19 percent) died
of other causes during the follow-up period. Thus, a total of
63 of the 171 patients died.
View larger version (13K): [in this window] [in a new window]
Figure 2. Survival of 134 Patients with Tuberculosis Resistant to Isoniazid and Rifampin, from the Initiation of Study Therapy to Death or the Most Recent Follow- up Examination. Successful treatment was indicated by consistently negative sputum cultures for three consecutive months, and relapse by the resumption of shedding of tubercle bacilli; patients with relapses are included among those with successful treatment. Treatment failure was indicated by an inadequate response to at least three months of therapy. The events represented are deaths due to tuberculosis.
Factors Influencing Outcome In univariate analyses, four factors were associated with the
failure of drug therapy to sterilize sputum cultures: previous
administration of more drugs, regimens including fewer previously
unused drugs, in vitro resistance to more drugs, and male sex
(this factor was marginally associated with failure) (Table 3).
When all the factors were included in a multivariate model,
only sex (odds ratio, 2.5; 95 percent confidence interval, 1.1
to 6.2) and the number of drugs received previously (odds ratio,
4.0; 95 percent confidence interval, 1.6 to 9.9) were significantly
associated with treatment failure (P<0.03 and P<0.001,
respectively).
View this Table 3. Predictors of Outcome of Chemotherapy in 134 Patients with table: [in this window] [in a new window]
Pulmonary Tuberculosis Resistant to Isoniazid and Rifampin.
Many regimens were employed. The most frequently used regimens
consisted of combinations of pyrazinamide, ethionamide, cycloserine,
and aminosalicylic acid with one of the injectable aminoglycoside
or polypeptide antimicrobial agents. Because of the small numbers
of patients in each group, the efficacy of individual regimens
could not be analyzed.
Discussion Treatment of patients with pulmonary tuberculosis due to M.
tuberculosis resistant to rifampin, isoniazid, and other medications
is complicated, risky, and of limited efficacy. Fully 46 percent
of our patients with treatment failure or relapses died. Physicians
and society must recognize that multidrug-resistant tuberculosis
is an ominous, deadly disease.
Patients who are treated unsuccessfully but remain alive pose
a major public health problem. They must be isolated because
of the risk of transmitting virtually untreatable drug-resistant
disease. Drug-resistant tubercle bacilli were once believed
less likely to be spread to contacts because the virulence of
these organisms was diminished. However, a recent report 20
demonstrated
that rates of infection (as detected with the tuberculin skin
test) among contacts of patients with drug-resistant tuberculosis
were comparable overall to those among susceptible persons.
In fact, because of their prolonged shedding of tubercle bacilli,
previously treated patients with drug resistance had higher
numbers of tuberculin-positive contacts than patients without
resistance. Epidemics have been reported in which disease, including
fatal meningitis and highly destructive lung disease, has evolved
among persons infected with these drug-resistant microbes 13,14,15,21,22,23 .
Quarantine of patients with treatment failure in their homes
is a potential remedy but is seldom achievable. Confinement
of patients may protect others, but they must then bear the
additional burdens of isolation and stigmatization.
Because the number of drugs available was limited, our regimens
for retreatment frequently entailed the use of agents known
to be toxic or difficult to tolerate. When side effects occurred
that were not potentially life-threatening and substitution
of another medication was not feasible, we made great efforts
to coach the patient through the drug intolerance, using both
symptomatic palliation and psychological support. This required
familiarity with each of the medications, as well as a great
deal of time in giving care. Nonetheless, our success rate (65
percent) in treating disease due to multidrug-resistant isolates
was less favorable than the rates in some earlier series of
patients given retreatment in the pre-rifampin era 24,25,26 .
The poorer results in the current study may be due to the larger
number of drugs to which the isolates from our patients were
resistant.
In our study an unsuccessful response to the therapy prescribed
was strongly associated with a greater number of drugs received
previously and with male sex. The reason for the association
with male sex is not readily apparent, but it may be related
to behavioral as well as biologic factors. The association with
previous treatment with a greater number of drugs emphasizes
the importance of thoroughly studying the patient's treatment
history. Suboptimal previous therapy leads to the development
of in vitro resistance. Thus, if the patient has already received
a large number of drugs, few agents will still be available
to which the isolate will be susceptible, and the chances for
successful outcome decline.
However, some patients received drugs to which their organisms
were susceptible according to conventional testing, yet they
did not respond clinically. This suggests that improved methods
of susceptibility testing are needed to provide more useful
laboratory guidelines for retreatment regimens. We are currently
exploring the use of the rapid radiometric system (Bactec) to
determine minimal inhibitory concentrations 27,28 -- values that
we anticipate will replace the traditional "critical concentration"
determined in solid agar.
Surgical resection as an adjunct to antituberculosis therapy
was performed in patients who had amenable lesions and who were
being treated with medical regimens deemed unlikely to effect
permanent control of their disease. We attempted to schedule
surgery so that it could be performed when the mycobacterial
count in sputum reached its expected nadir. Three patients with
indeterminate responses to medical therapy and two without bacteriologic
responses to chemotherapy were clearly helped by resection.
In the other two patients, whose sputum cultures had already
become negative, the benefit of surgery was less clear.
When rifampin was introduced in 1971, it represented a major
advance in antituberculosis treatment, leading to both short-course
chemotherapy 1 and highly effective, well-tolerated management
of tuberculosis due to isoniazid-resistant organisms 10 . However,
as anticipated, 29
resistance to rifampin has emerged in substantial
numbers of patients, frequently in association with the previous
use of rifampin in an inadequate regimen and with noncompliance.
Noncompliance contributed to the development of resistance to
rifampin in 74 of our patients (43 percent of this series of
171 patients), and was suspected in many more. Noncompliance
with prescribed treatment is extremely common 30 . The most effective
safeguard is the directly observed administration of medications.
The treatment of patients with multidrug-resistant tuberculosis
entails five to seven visits per week. The cost of retreatment
of patients in whom treatment fails and resistance to rifampin
emerges should be considered when the economics of a tuberculosis-control
program are being evaluated 31 . Our very limited rates of success
in treating multidrug-resistant disease raise serious concerns
about the future of tuberculosis control. If we, with all our
resources, can successfully treat less than 60 percent of such
patients, what will happen in less ideal circumstances? To avoid
squandering the potential effectiveness of rifampin because
of noncompliance or inept prescribing, systematic efforts are
required to preserve the potency of this crucial drug against
tuberculosis.
This report includes patients first seen between 1973 and 1983.
Thus, it is likely that few, if any, of these patients were
infected with HIV. However, many patients who have contracted
multidrug-resistant tuberculosis have HIV infection 14,15 ; their
response to treatment appears to be very poor, with high rates
of treatment failure and mortality 32 . Their response to antituberculosis
therapy may be suboptimal, even when the organisms are susceptible
to the drugs used 33,34 . Thus, although we recommend similar
strategies for the therapy of drug- resistant tuberculosis whether
or not it accompanies HIV infection, the efficacy of such therapy
in patients with HIV remains to be determined. New antimycobacterial
agents are urgently needed to expand our therapeutic options