Key Issues In Methadone Maintenance Treatment. Chapter 2 "Treatment Effectiveness I: Randomised Controlled Trials"

I. The Randomised Controlled Trial

The gold standard for establishing the effectiveness of any treatment in modern medicine is a reproducible demonstration in a randomised controlled trial that the treatment produces a superior outcome to a relevant comparison treatment, such as no treatment or minimal treatment. The simplest type of randomised controlled trial is one in which people with a condition (e.g. opioid dependence) are randomly assigned to receive either the active treatment (e.g. methadone maintenance) or a comparison treatment (e.g. drug-free counselling). The evaluation of treatment effectiveness presupposes a comparison treatment so that one can discover what would have happened if the patient had received a different treatment, including no treatment at all. The aim of randomisation is to ensure that the subjects who are allocated to the treatment and the comparison conditions are equivalent in the long run, that is, over a large number of trials in which the subjects have been randomly assigned to the treatment and comparison conditions. Only when the two groups have been assigned in this way can one be confident that a difference in treatment outcome is more likely to reflect the effects of the treatment than the pre-existing characteristics of the subjects who were assigned to the different treatments.

In order to minimise bias in assessment, both the administration of treatment and the assessment of treatment outcome should be conducted in such a way that neither the person receiving the treatment nor the person assessing its effects are aware of which treatment the subject has received. When this is not possible, at least the assessment of treatment outcome should be conducted by an assessor who is unaware of which treatment the subject has received. The measurement of outcome should also be of demonstrated validity, and a statistical test should be used to decide if any difference in treatment outcome observed between the treatment and control is too large to have arisen by chance.

The demonstration in a single study that a treatment produces a better outcome than a control condition is rarely decisive in evaluating the therapeutic effectiveness. it is the ability to reliably reproduce or replicate such findings that establishes the therapeutic effectiveness. The importance of successful replication of both positive or negative results derives from the possibility of making errors in the statistical comparison of the outcomes of the treatments being compared. We will falsely conclude that treatment is superior to the comparison in 5% of tests, and on occasions we will also incorrectly conclude that treatment is no different from the comparison condition. The percentage of occasions in which the latter happens depends upon the size of the difference we are seeking to detect and the sample size we use in our attempts to detect it. The more reliably a result is reproduced, the greater our confidence in it, because the chance of a run of consistent decision errors decreases dramatically as the number of replications increases. So, for example, if we always demand that the chance of our falsely rejecting the null hypothesis is no greater than 0.05 then the chances of obtaining five false positive results is (0.05)5 = 0.00000003125.

II. Randomised Controlled Trials of Methadone Maintenance

There have been few randomised controlled trials that compared the effectiveness of methadone maintenance with an appropriate control condition. When methadone maintenance was introduced (Dole & Nyswander, 1965), the randomised controlled trial was not part of the culture of treatment evaluation to the same degree it is today. This meant that the opportunity to randomly assign patients to methadone and minimal treatment was only rarely exercised before methadone became a widely available form of treatment. By the time that methadone maintenance had become an important part of the publicly-funded treatment system for opioid dependence in the early 1970s, it was no longer politically acceptable to deny the treatment to people who might have benefited from it. Although it was still ethically acceptable to randomly assign opioid addicts to methadone and other competing forms of treatment, in practice it became difficult to do so because patients who were randomly assigned to treatments of which they did not approve, could obtain their preferred treatment elsewhere (for example, Bale et al., 1980; Bell et al., 1992).

Only three randomised controlled trials have been performed in which comprehensive methadone maintenance has been compared with a control condition over a substantial period of time (Dole et al., 1969; Newman & Whitehill, 1979; Gunne & Grönbladh, 1981). All three studies were undertaken in a context in which methadone maintenance program places were strictly rationed - a fact that made it ethically acceptable to randomly assign patients to either methadone or a control condition. More recently, two randomised controlled trials have compared some form of methadone maintenance with an alternative treatment over short periods of time (45 days or less). These two trials will also be reviewed. Randomised controlled trials have also compared variations of methadone maintenance treatment with one another (e.g. high and low dose methadone maintenance), or methadone with other forms of maintenance using synthetic opioids such as LAAM. However, these studies do not bear as directly on the effectiveness of methadone as do the studies that have been included in this chapter. They are reviewed where relevant in subsequent chapters (for example, on methadone dose).

II. a. Dole, Robinson, Oracca, Towns, Searcy and Caine (1969)

These investigators conducted the first randomised controlled trial of methadone maintenance in New York using imprisoned, recidivist opioid addicts who had at least a four-year history of opioid use. Thirty-four men who became eligible for release over a four month period were invited to participate in the trial, of whom 32 accepted. Half of these (16) were randomly assigned to methadone maintenance (of whom 12 entered treatment), and the other 16 were assigned to a no treatment waiting list. Methadone maintenance was commenced before they left prison and continued after their release.

Both groups were followed up at 12 months post-release, and only one subject in each group was lost to follow-up. There were dramatic differences in favour of methadone maintenance when outcome was assessed by rates of imprisonment and return to daily heroin use. Six of the 12 men who entered methadone maintenance were employed or in school, and three had been gaoled, whereas all 16 of those in the control condition had returned to gaol. Similarly, while all 16 men in the control condition had returned to daily heroin use, none of the men in methadone maintenance had done so, even though 10 out of 12 had used heroin since their release, and three continued to use intermittently.

Even a conservative analysis on the basis of 'intention to treat' favours methadone maintenance -- that is, an analysis in which all 16 men who were originally assigned to methadone maintenance were included (rather than the 12 who entered treatment), and all subjects who were lost to follow-up were counted as treatment failures. The outcomes from an analysis by intention to treat are shown in Table 1 expressed in terms of odds ratios with their accompanying 95% confidence intervals. The odds ratios express the ratio of the odds of the control and methadone maintenance conditions returning to daily heroin use or being imprisoned in the year after treatment. The 95% confidence intervals indicate the smallest and largest values of the odds ratio that are consistent with the sample results.

Table 1 (not shown here) shows that the odds of being imprisoned were 53 times higher, while the odds of returning to daily heroin use were 92 times higher, among those in the control condition than those in methadone maintenance. The number of cases in each group are small which makes the estimates of the odds ratios uncertain, as is reflected in the width of the 95% confidence intervals around each odds ratio. In the case of daily heroin use, for example, the confidence interval ranges between a lower limit of 2.7 and an upper limit of 1048. Contrary to popular prejudice, the fact that such differences are statistically significant with such small samples makes the size of the differences in favour of methadone all the more impressive.

II. b. Newman and Whitehill (1979)

These investigators conducted a randomised controlled trial of methadone versus placebo maintenance among heroin addicts in Hong Kong. The trial was made possible by the late introduction of methadone maintenance treatment to the colony, which meant that people who were randomly assigned to the control condition were unable to obtain it elsewhere. The patients included in the trial were the first 100 male addicts who met the same criteria that had been used in the Dole et al. (1969) study, namely, they had at least a four-year history of opioid addiction, at least one failed attempt at rehabilitation by other means, and evidence on urinalysis of daily opioid use.

All those who consented to participate in the trial were first admitted to the treatment unit for two weeks and stabilised on 60 mg of methadone. They were then randomly assigned to be placed on methadone maintenance or placebo maintenance after discharge. Both groups were offered extensive follow-up counselling and treatment. The methadone group received a high dose of methadone determined by the patient (average 97 mg per day) while those in the placebo condition were withdrawn from methadone under double blind conditions. In the methadone condition, patients who continued to use heroin as monitored by urinalysis (more than six positive urines), and those who failed to comply with the requirement for daily dosing (by missing six consecutive doses) were discharged from the program.

Both groups were followed for three years and outcome was assessed in terms of the numbers retained in treatment. The differences in treatment retention were dramatic (Table 2--not shown here). By the end of 32 weeks five of the 50 placebo controls and 38 of the 50 methadone treated group were still in treatment. By the end of three years the numbers still in treatment were one and 28 respectively (OR = 62.4, 95% CI: 8.0, 487.9). The reasons for discontinuing treatment also favoured the methadone group: 31 of the 49 patients from the placebo group were discharged for continued heroin use compared with only eight of the 22 patients in the methadone group. There were three deaths in the study, all in the methadone group. In only one case was there any suspicion of an overdose. The other two deaths were from causes not related to continued heroin use (although both deaths were probably attributable in part to the adverse health effects of prior opioid use).

II. c. Gunne and Grönbladh (1981)

These authors conducted a randomised controlled trial of the Swedish methadone maintenance program that was closely modelled on the original Dole and Nyswander (1965) approach. As with the Newman and Whitehill (1979) study, it was possible to undertake such a study because methadone maintenance was only introduced into Sweden in the early 1970s, and the number of places in the program was strictly rationed because of political opposition to methadone maintenance as a form of treatment (Grönbladh & Gunne, 1989).

The criteria used to select persons who were eligible for inclusion in the study were substantially the same as those of Dole et al. (1969), and Newman and Whitehill (1979), namely, at least a four-year history of opioid addiction, a previous failed attempt at rehabilitation, and evidence from urinalysis of daily opioid use. Those who were under the age of 20 were excluded, as were those who used other drugs, or who were facing criminal charges. The methadone maintenance program in this case involved substantial vocational rehabilitation during an inpatient admission of up to six months. All subjects who were assigned to the control condition refused drug-free treatment. The two conditions under comparison, then, were methadone maintenance in a setting of intensive vocational rehabilitation and no treatment.

This study differed from the other two in that it used a sequential design. Instead of assigning a predetermined number of subjects to methadone or a control condition, the trial continued until a statistically significant difference emerged in favour of either condition. This occurred after 36 subjects had been recruited, 17 of whom were assigned to methadone, and 19 to the comparison treatment (with two subsequently being excluded because they enrolled in methadone elsewhere).

The outcomes were initially assessed at the end of two years-- the point at which those initially assigned to the control condition became eligible for entry to methadone. Twelve of the 17 in the treatment condition were no longer regularly using opiates or other drugs, and were either employed (10) or undertaking further education (2). The remaining five treatment subjects continued to abuse opioids or hypnotics, and had been discharged from the program. Only one of the 17 subjects in the control condition had ceased drug abuse; 12 continued to abuse opioid drugs; two had died and two were in prison (OR = 38.4, 95% CI: 4.0, 373.1). That is, at the end of the two-year follow-up, among subjects who entered methadone treatment the odds of discontinuing regular illicit drug use were 38 times the odds of doing so among the subjects who were initially offered drug-free treatment. The results reported by Gunne and Grönbladh among the treatment patients in their study were the same as those obtained among another 174 patients who were admitted to the methadone maintenance program over a 20-year period (Grönbladh & Gunne, 1989).

II. d. Vanichseni, Wongsuwan, Staff of BMA Narcotics Clinic No. 6,
Choopanya and Wongpanich (1991)

Vanichseni et al. conducted a randomised controlled trial comparing 45-day methadone detoxification with 45 days of methadone maintenance. The subjects of the trial were 240 heroin injectors in Bangkok, Thailand who applied for detoxification and who had at least six prior detoxifications. They were randomly assigned to methadone assisted withdrawal over 45 days (the standard detoxification regime in Thailand), or to methadone maintenance for the equivalent period (average dose 74 mg per day). Outcome was assessed by continued illicit heroin use, as indicated by morphine positive urines during twice-weekly urinalysis, and retention in treatment. There was no further description of the treatment program.

Predictably there were major differences between the withdrawal and maintenance groups on both outcome measures. The drop-out rates by the end of the 45-day period were 66% and 24% respectively (OR = 6.05, 95% CI: 3.44, 10.62), with the withdrawal group showing dropping-out of treatment earlier than the maintenance group. That is, the odds of dropping-out of treatment were six times higher among those on the withdrawal program than those on the maintenance regimes. The percentages of morphine positive urines were 53% and 28% in the withdrawal and maintenance groups respectively (OR = 10.33, 95% CI: 3.40, 31.35). That is, the odds of providing a morphine positive urine were more than 10 times higher for those on the withdrawal program than among those on the maintenance regime.

The practical significance of the findings for the effectiveness of methadone maintenance is uncertain. It is reassuring, if unsurprising, that Thai heroin addicts who enter methadone maintenance are more likely to remain in treatment, and less likely to continue to inject heroin while in treatment, than are those who are placed on the standard withdrawal regime. But this is hardly a rigorous test of the effectiveness of methadone maintenance in retaining patients in treatment and minimising their injection of heroin in the longer term, which is measured in years rather than days.

II. e. Yancovitz, Des Jarlais, Peyser, Drew,
Friedmann, Rigg and Robinson (1991)

These authors reported a randomised controlled trial of 'interim' methadone versus limited contact while on a waiting list to enter a comprehensive methadone maintenance treatment program. 'Interim' methadone involved the 'provision of limited services to patients awaiting treatment positions in comprehensive methadone programs'. It consisted of an initial medical examination, education about AIDS, and the daily dispensation of oral methadone medication 'to prevent narcotic withdrawal symptoms and to block the euphoric effects of heroin' (p. 1185). No vocational or other social rehabilitation or counselling was provided.

The subjects for the study were 301 heroin addicts recruited from the waiting lists of 23 methadone maintenance treatment programs in New York. Initially, patients were randomly assigned to one of three conditions: interim methadone; waiting list with frequent contact and urinalysis; and waiting list without contact. Once a treatment place became available they entered comprehensive treatment and left the study. Initially, recruitment into the trial was good but this soon slowed dramatically when potential participants perceived the one in three chance of receiving methadone as too low. The protocol was subsequently simplified to a two-group design (interim methadone versus frequent contact) and the duration of both conditions was limited to one month, after which all entered comprehensive treatment.

By comparison with the results of urinalysis, the self-reports under-reported drug use; therefore, only the urinalysis results were reported. These showed that the proportion of patients in interim methadone who had used heroin declined from 63% to 29% while the proportion remained stationary in the frequent contact control group (62% and 60%) (OR = 3.55, 95% CI: 1.86, 6.77). That is, the odds of using heroin during the one month trial were over three times higher among the frequent contact control than among the interim methadone maintenance group. There was no change in cocaine use in either group.

The difference in favour of interim methadone persisted when the partial results of the patients with incomplete data were examined (by using the urinalysis result closest to the conclusion of the trial). In this case the proportions of patients with urines positive for morphine were 36% in the interim methadone group and 60% in the frequent contact control group. Sixteen months after the trial the proportion of patients who had subsequently enrolled in comprehensive methadone maintenance treatment were 72% in the interim methadone condition and 56% in the frequent contact condition. Thus, the odds of subsequently enrolling in comprehensive methadone maintenance were two times higher in the interim methadone group than in the frequent contact control group (OR = 2.01, 95% CI: 1.24, 3.24).

This study illustrates the difficulties of conducting randomised controlled trials of methadone maintenance treatment. The need to entice participants into the trial, and to keep them in treatment, limited the duration of the trial, and hence limited the inferences that could be drawn about the effectiveness of interim methadone. As was the case with the trial of Vanchseni et al., the reduction in heroin use was reassuring but hardly compelling proof of the effectiveness of interim methadone. The increase in the number of persons who subsequently entered comprehensive treatment was perhaps of greater clinical significance.

III. Summary

The three controlled trials of comprehensive methadone maintenance produced similar results: all showed that methadone maintenance was more effective than either placebo or no treatment in retaining people in treatment, in reducing opioid use, and in reducing the rate of incarceration. This is an impressive result for studies that have included small sample sizes, and have been conducted in three different countries over a period of about 15 years.

The two more recent controlled studies of time-limited methadone maintenance programs with a minimum of support services provide evidence of the short-term effectiveness of methadone maintenance in retaining patients in treatment and reducing their heroin use while they remain in treatment. Although the results of the randomised controlled trials are strongly supportive of the effectiveness of methadone, there are arguably too few replications to enable definitive conclusions to be drawn about the effectiveness of methadone. Our confidence in the results of the randomised controlled trials will be enhanced to the degree that similar results have been reported in larger observational studies of effectiveness.