ARTICLE

TOP ABSTRACT ARTICLE REFERENCES

The State of Colorado recently passed controversial legislation which can be summarized as follows, "To prevent the emergence of multiple drug-resistant tuberculosis (TB), all patients with pulmonary TB will be treated with directly observed therapy (DOT) for the full course of treatment unless a variance is approved by designated public health authorities" (1). This law mandating "universal" DOT is generally consistent with, but goes beyond, the Centers for Disease Control recommendation that DOT should be considered for all patients (2).

"Universal" DOT is strongly opposed by Annas, a noted expert in medical ethics, who states it is "gratuitously annoying, inconvenient, inefficient, and wasteful" (3). Many in the medical and lay community are asking the question, Does the Colorado law go too far? A recent survey of health departments in the United States showed that pressing for universal DOT in settings that have already achieved very high rates of treatment completion may produce only small improvements in completion rates at very high marginal costs (4). Furthermore, despite vigorous promotion of DOT by the Centers for Disease Control, less than half, 46%, of TB patients were treated with DOT in major metropolitan health departments in the United States in 1996 (5). These findings and these concerns raise the question, Is there an alternative means for overcoming the compliance problem and achieving adequate treatment of all TB patients?

The usual justification for employing "universal" DOT is the well-known difficulty in differentiating patients who are reliable in taking medication from those who are not. Sackett and coworkers state that clinical judgment of a patient's compliance is no better than flipping a coin (6). Although the evidence presented subsequently indicates that this statement is not entirely correct, there is no question that clinical judgments in this area are fallible.

In the late 1960s, a compliance study was carried out using equipment, called a medication monitor, to determine the approximate time when each dose of medication was removed from a container (7, 8). The nurses and physicians who knew the patient best were asked to predict if the patient would take > 90%, 70 to 90%, or < 70% of the prescribed medication. Physicians were able to place the patients in the correct category 61% of the time and nurses 63% of the time (8). This degree of agreement between predictions and compliance is probably greater than what would be achieved today, because in the 1960s TB patients were usually hospitalized for many months during which the professional staff could assess their reliability. These findings support the position that all patients should be given DOT.

On the other hand, the same study showed that large numbers of patients are reliable or reasonably reliable in taking anti-TB drugs (7). Sixty-one percent of patients took greater than 90% of their prescribed medication, and 83% took more than 70%. Most of the patients in the study came from Denver General Hospital. The homeless, alcoholic, and psychotic patients were excluded and given DOT. Consequently, those who were in the study presumably had the motivation of the average public TB patient who is judged to be reliable. A few patients came from outside of Colorado to be treated as inpatients at National Jewish Hospital for drug-resistant TB, usually for 1 yr, and voluntarily decided to remain for a second year of outpatient treatment to ensure treatment success. This latter group most likely had a greater than average motivation which probably resulted in better compliance. However, all patients were treated for 18 to 24 mo, compared with today's regimens of 6 to 9 mo where better compliance can be anticipated. Therefore, the data derived from this early study are probably representative of results achieved in current programs that use DOT selectively. These data argue against the use of "universal" DOT and raise the issue: Could medication monitors be used to identify poorly compliant patients among those erroneously judged to be reliable?

An obvious limitation of medication monitors is that while patients might consistently remove medication on time, they could then discard it, and still create a monitor record that indicates regular drug ingestion. Most people believe that this type of deceptive behavior will occur very rarely, because those patients who fail to follow instructions will most likely fail to remove the medication on time. A study comparing the times recorded by the monitor when medication is removed and tests for the medication or metabolites in urine samples collected at surprise visits to the patient's home or place of work would establish how often this occurs.

Even if on-time removal without ingestion were to occur frequently and result in treatment failure, drug resistance is not likely to develop if the medications consist of combined preparations of isoniazid and rifampin (Rifamate) or isoniazid, rifampin, and pyrazinamide (Rifater) (Hoechst Marion Roussel, Kansas City, MO), because these preparations make it impossible for the patient to take a single drug (monotherapy), a major cause of drug resistance (9). Additional partial evidence supporting this assertion comes from the San Francisco health department, which had only one in 312 human immunodeficiency virus (HIV)-negative patients (0.32%) who developed drug resistance between 1990 and 1994 (10) despite the fact that only 22% of patients were on DOT (4) while using mostly combined preparations (Dr. Gisela Schecter, personal communication).

The original medication monitor was cumbersome to use (7). Since then the introduction of digital watches and computer chips has made practical "electronic" medication monitors feasible. The most widely used device, the MEMS cap (Aprex, Menlo Park, CA) records when a cap is removed from a medication bottle (11). Similar devices with five drawers for five different medications record when each drawer is opened (Poly Pharm, Northville, MI, and Informedix, Rockville, MD). Two devices record when all medication to be ingested at one point in time is removed from a compartment in a multiple compartment device (Anderson Clinical Technologies, Elmhurst, IL, and Compumed, Meeteetse, WY). One medication monitor, currently being developed with a National Institutes of Health (NIH) grant, determines when each pill is removed from a container (NBS Medical, Costa Mesa, CA). Although all of these devices have deficiencies, most of them could be used in their present form to monitor medication compliance. With further engineering more precise and less cumbersome equipment can be anticipated.

Medication monitors can be used for supervising the treatment of any illness. However, in contrast to tuberculosis where the number of different pills can be minimized by using the combined preparations of Rifamate and Rifater, many illnesses, like the acquired immunodeficiency syndrome (AIDS), are treated with a large number of different medications. If the MEMS cap or the device being developed by NBS Medical is used, one device would be needed for each medication. The devices produced by Poly Pharm and Informedix, with five drawers for five different medications, and the device from Compumed, which has multiple compartments each containing all pills to be taken at one point in time, are better adapted to deal with regimens that contain multiple different medications.

These increasingly practical medication monitors create an opportunity to make selective DOT a reasonable alternative to "universal" DOT. Patients who are homeless or judged to be unreliable such as alcoholics, substance abusers, and the mentally ill, along with HIV-positive patients and retreatment patients would usually if not always be treated with DOT. Patients judged to be reliable could be given self-administered medication in a medication monitor and kept on this form of treatment as long as their monitor record was adequate. If treatment failure occurred in the occasional patient, the chance that drug resistance would develop could be minimized by using combined preparations of anti-TB drugs (9, 10).

A study with a medication monitor for isoniazid and thiacetazone in Haiti found that good compliance in the first 11 wk of therapy correlated with good compliance and good clinic attendance in the following 9 mo of therapy (12), indicating that poorly compliant patients will usually be found early in the course of treatment allowing for an early intervention to correct the problem. In many cases this would be a change to DOT. In the original medication monitor study with 122 patients, a change to DOT was carried out on 9 occasions (7). It is recognized that some patients might interpret a change to DOT as punitive and refuse to cooperate, resulting in an excessive staff effort and unpleasant legal measures to implement DOT. On the other hand, placing everyone on DOT "punishes" reliable patients and most likely results in an even greater staff effort.

Special counseling of the patient and family based on the monitor record, which is often called "feed back," is another means of responding to a poor monitor record. This was carried out on 25 occasions in the original monitor study after which the patients' monitor records usually improved or remained at a level where the physician felt the chemotherapy would be adequate to control the disease (7). A study of patients taking isoniazid preventive treatment from a container with the MEMS cap showed that feed back from both the physician and pharmacist allowed drug adherence to be enhanced in noncompliant patients, but the effect was only transient, if this was not repeated every month (13). Sixty-nine percent of the patients in the study in Haiti who received feed back based on the monitor record took 90 to 100% of the medication, compared with 58% who did not, among patients completing a year of treatment (12). In addition, 70% of the patients in Haiti who received feed back attended the clinic regularly compared with 56% who did not (12). These findings suggest that monitor-directed counseling or feed back improves compliance modestly.

There are multiple other ways of using medication monitors to deal with the compliance problem. Beepers to remind and liquid crystal displays (LCD) to instruct the patient are available with most of the electronic devices. Moderately poor monitor records could be managed by extending the duration of treatment, if the disease is not too extensive or the poor compliance occurs late in the course of therapy. If monitored self-administered treatment from the beginning of therapy is thought to be unwise, patients could be started on DOT and changed to monitored self-administered treatment after several months, if they responded to treatment and cooperated with DOT.

While multiple investigations will be needed to confirm the findings in these early studies and to work out the optimal way of using medication monitors, these devices would appear to fill a void in our existing tools for dealing with the compliance problem. If, as allowed by the Colorado law, human judgment is used to choose patients who are given a variance to take self-administered medication, some treatment failures will sooner or later occur, because judgments regarding compliance have been shown to be very fallible (6, 8). If medication monitors are used to detect poor compliance among patients judged to be sufficiently reliable to take self-administered treatment, the occurrence of treatment failure can be avoided or the chance of it occurring minimized by corrective action taken early in the course of therapy. Of equal importance, treatment with monitored self-administered medication should provide less intrusive, less disruptive, and less expensive therapy for reliable TB patients than programs that are based on "universal" DOT.