Major studies of opiate related morbidity and mortality
Although the literature is replete with studies of opiate follow-up, many of these suffer from methodological shortcomings such as short follow-up periods or small patient numbers. To circumvent these issues, to confine the enormous subject within reasonable bounds, and to limit consideration to the most authoritative sources, it is proposed to focus on two large, recently published population-based studies. The first study of interest is a five year prospective study supported by the National Cancer Institute, was published in Lancet.74
The second is a 21 year retrospective and historical review of the clinical experiences and mortality experience of Australia’s most populous state, New South Wales. This second study appeared in Drug and Alcohol Dependence.73
The first involved 8,487 opium addicted people and 41,558 controls (a total of 50,045 persons) and amassed 234,928 person-years. The second studied 42,676 opiate addicted patients over 21 years and analyzed 425,998 patient-years. The data of interest in the second study appeared as a Supplementary file in the Web Appendix 6. A subsequent publication which explored the cancer deaths in more detail in the Sydney experience has also appeared.80
The two studies reported on populations who took opiates predominantly by the oral and inhaled routes, and by the oral and parenteral routes, respectively.
The prospective study from Iran was able to utilize advance statistical modelling to show that not only were opiates and death from many causes statistically associated, but the relationship was causal in the forward direction, from opiates to death. Cessation of opiate use required about 10 years for the effect of elevated mortality to regress to baseline.
The cause of death data from both studies is summarized in . The two studies report hazard ratios and standardized mortality ratios respectively. A number of striking findings stand out from this table.
Effect sizes for death by organ system in opiate dependence.
Firstly, opiate use by either route is associated with greatly elevated mortality, even in the group using opiate by the inhaled and oral route. When adjusted for all other factors and when healthy patients are compared with opiate users this effect is 1.90 (95% C.I. 1.55–2.33). The effect appears to be compounded by use of the parenteral route. As noted above, 83% of deaths in the first study were related to chronic disease, as were 37% of those in the Australian study. Interestingly both studies reported higher HR/SMRs in females, by a factor of 1.49 and 1.47 respectively. This was highly statistically-significant (both P < 0.001). This finding remains unexplained. Conceivably, gender differences in immune function may be involved.
Secondly, opiates appear to be associated with cardiovascular death and cancer deaths, findings which, although they have been previously described, have not been demonstrated definitively and are not widely appreciated. Interestingly, similar comments apply to chronic digestive and pulmonary disease. Suicide, as a fatal end point of depression, can also be considered a neuroinflammatory disorder. The subsequent report from Sydney shows that rates of liver, lung and anogenital (including cervical) cancer were elevated and breast cancer rates were reduced.80
Death related to alcoholic disease was noted to be elevated, which relates to earlier comments about disinhibition of the hypothalamic craving mechanism. The rates of overdose found in both studies is as shown. The Iranian study found that 3 of the 4 heroin users died during the five year study period. These deaths can probably be regarded as overdose deaths, and have been listed as such, but the assignment is tentative. The hazard ratio is also highly unstable in view of the small patient numbers.
Previous publications from Iran have found that in males, opium use is the strongest cardiovascular risk factor related to the need to open surgical coronary revascularization. Similarly in a carefully controlled Iranian study, opium use was found to be a more powerful predictive risk factor than tobacco use or diabetes81
and to be associated with coronary artery disease occurring 4 years before that seen in a nonaddicted population.82,83
Thirdly, the degree of agreement of many of the effect sizes calculated is remarkable. Hence there is virtual identity for the effect sizes relating to cardiovascular disease, cancer, lung cancer, and chronic respiratory disease. Such near identity between two large and highly powered studies from two such divergent contexts provide powerful evidence for both the reality of the effect demonstrated1 and its magnitude.
Cardiovascular disease is now understood to have an inflammatory basis, as is much respiratory disease and hepatic cirrhosis. The inflammatory component of depression is widely known and widely studied. Moreover the contribution of chronic inflammation to many cancers is increasingly being appreciated.
Other studies show elevated rates of cardiovascular risk factors in opiate dependence, including tobacco use, weight gain, and unhealthy eating.84–88
Various endocrinopathies have been reported in COA, including hyperglycaemia and diabetes, growth hormone stimulation, hypothalamic hypogonadotrophic hypogonadism, insulin resistance, hyperprolactinaemia, blunting of the hypothalamic-pituitary-adrenocortical responsiveness, syndrome of inappropriate antidiuretic hormone secretion (SIADH), and osteopaenia-osteoporosis. 44,89–92
It should also be noted that many of these changes are similar to those seen in biological human ageing. As so carefully demonstrated by the cardiovascular surgeons in Tehran, these changes occur several years earlier in opiate dependence.82,83
Indeed, in these studies opium use in men was a more powerful predictive factor than diabetes.81
Similar observations have been made for other age related indices such as hair greying, chronic periodontitis, and stem cell depression.72,93–95
Cancers of the esophagus, larynx, and bladder have also been reported at elevated rates in large Iranian studies.96–98
Since naltrexone has been shown to reverse the proinflammatory effects of TLR ligation, it is expected that any of these pathologies will reverse under long-term antagonist treatment.
Oral use of opioids
An important question relates to whether these considerations are different for opiates used for clinical purposes such as long-term pain relief. The management of long-term pain is a difficult and very real clinical problem, and one upon which much research has been focused in recent years. It should however be noted that the above pathophysiological discussion is not specific to opioids administered by any particular route. Hence large studies have found that mortality amongst patients given opiates for pain is also greatly elevated, and although these publications did not provide detailed cause of death data, it may reasonably be surmised that pathophysiological mechanisms such as those described above contributed to the elevated rates of mortality observed.99,100
Moreover the large Iranian study discussed above considered patients in whom opiate use was via the oral route.74
The point was also made in that study that the level of use of these patients was relatively low. The large Australian study described in detail above also considered the standardized mortality rate (SMR) of patients in treatment and patients out of treatment, which may be considered as patients using mainly oral narcotics and patients mainly using parenteral routes of administration.73
The SMRs from all causes were greatly elevated in both groups, though more so in the group using parenterally. Hence it is likely that pain patients exposed to long-term clinical use of opiates also experience similar risks, although somewhat less so than those exposed by parenteral routes. This explains some of the present research drive to develop newer non-addictive forms of analgesics which are not immune-active.
Randomized clinical studies of extended release naltrexone
Although there are several anecdotal case reports and clinical series relating to extended release naltrexone (EX-NTX), for the sake of conciseness and quality of evidence, the author’s intention is to confine the present discussion to the randomized clinical trials of EX-NTX.
shows a summary of six recently published trials of EX-NTX.2,17,19–21,101
Together this represents 384 patients treated in the active implant arms and 321 in the control groups. The studies are not all the same and their differences in design together provide a useful overview of a diverse clinical experience with EX-NTX. Some of the statistics listed in the table have been quoted from the publication concerned and some have been recalculated using EpiInfo software (Version 184.108.40.206, CDC, Georgia) or “R” (Version 2.13.1, CRAN Archive), based on the published data. Preliminary interim analysis and results from a large Russian study have long been available.12
However, these have recently been superseded by the definitive report from the completed study,101
and it is these latter results which are included in the present review.
Summary of randomized clinical trials.
It is particularly noteworthy that whilst most of these studies are not numerically large, they uniformly report highly statistically-significant results. That so uniform a pattern of results has been reported from so many locations is strongly indicative of a powerful and reliable treatment effect. Indeed the Krupitsky 2011 paper included patients from 13 sites in Russia. The question of EX-NTX is particularly relevant to the Russian Federation as agonist treatment is not lawful in that country, as indeed it is not in 122 of the 192 (63%) member nations of the United Nations.2,12,19
Most studies have compared naltrexone to placebo. The Comer study compared 18 placebo-treated patients to 22 patients treated with two Depotrex injections and 20 patients treated with one. The serum levels achieved in the single injection group were quite low, with mean peaks of around 2 ng/mL, so that this groups represents a worst case scenario. Hulse compared depot naltrexone with oral naltrexone. Tiihonen studied a group of dually addicted heroin and amphetamine-dependent patients to more closely approximate the real world experience in contemporary Russia. Krupitsky notes in many reports that most of the cohorts upon which he reports are young, with a mean age of 22, and only 2.5 years of drug use.12
Most live at home with their parents. Otherwise the duration of heroin dependence seen in the trial patients listed is typical of that seen in many series in the literature. Both the Krupitsky 2010 and 2012 studies utilized 3 two month Prodetoxone implants.12
Similarly the Krupitsky 2011 study utilized 6 once-monthly injections of Vivitrol.19
This is an important design nuance as much of the discussion and controversy relating to NI overlooks the fact that, as in agonist treatments, repeated administrations of the treatment can be performed. NI is not necessarily a “one-off” treatment as it is frequently describe. Indeed, it is important to note that in many of the centers where it is used most successfully, repeated administration, along with psychosocial support, are commonplace.102
Finally, in the most recently published study, Krupitsky compared three groups, with 102 participants in each of the implant naltrexone, oral naltrexone, and placebo (implant and oral) groups.
As the indices reported in the different studies varied, it is not necessarily a simple exercise to compare their results. All of the studies found a highly statistically-significant increase in the achievement of opiate-free urine results with odds ratios ranging from 1.58 to 4.7 and above. As listed, the P
-values are highly significant. The Comer study was important as it showed a dose-dependent response with the two injections performing better than one on most measured indices. The Comer report also calculated a number needed to treat (NNT) index. This figure was 2.8 on an intention-to-treat basis and 2.36 on a treatment-completion basis.17
These figures are very low indeed and indicate a much higher treatment effect than many other commonly used treatments such as anti-hypertensives, antibiotics in primary care, and lipid lowering agents.
Conversely the rate of return to dependent drug use was shown to be statistically significantly reduced, with odds ratios from 0.05 to 0.22.
Five of the studies reported marked reductions in opiate craving scores. Interestingly the Comer study, which was numerically small, also reported highly significant reductions in craving for cocaine, benzodiazepines, amphetamine, and methadone (all P < 0.05). The reduction in cannabis craving in that study was P = 0.08, which did not quite reach significance. However, cannabis was smoked by a minority of study participants, which may not be representative of other opiate-dependent populations.
Interestingly most studies reported significantly reduced rates of depression or enhanced scores for general well-being.
The largest study in this area which has just been published is that by Krupitsky in 2012.101
It is noteworthy for several reasons. Whilst agonist medications are not considered legal in the Russian Federation, the nation seems particularly well covered by a network of 138 addiction dispensaries, including 115 inpatient detoxification units and 12 addiction hospitals with over 25,000 beds and 5,600 psychiatrists and narcologists working mainly in addiction. The authors compared the relapse free outcomes in three groups (n = 102 each) of implant naltrexone, oral naltrexone, and placebo groups with blinding by placebos for both implants and tablets. All patients required accountability to a suitable caretaker such as a parent or partner—as is typical of Russian studies—and the unavailability of a caretaker was sufficient reason for exclusion from the study (13 patients). The mean age was 28.2, rather older than many of the other studies. The mean duration of heroin use was 8.0 years. The protocol involved the use of three implants at bimonthly intervals. However, patients who relapsed back to heroin use were excluded from further participation in the study. Since in the real world patients would simply be re-implanted, the results of this study should be considered as the lower bound of the efficacy of NI. These authors reported that 54 of 102 NI patients (52.9%) were relapse-free at six months, which was significantly better than the outcome in the double placebo group (log rank statistic = 68.4, df = 1, P
= 6.17 × 10−17
). Patients also received follow-up nine and twelve months after trial registration, but results at these time points were poor. The authors concluded that extended periods in treatment were required for most patients. Oral naltrexone performed better than the double placebo group on both clinical relapse and opiate positive urine criteria (P
= 0.014 and 3.4 × 10−6
). No deaths were reported.
Meta-analysis of results
Patients in these studies can largely be considered similar. Perhaps the largest differences related to study duration, which were of 2, 6, 6, 6, 2.5 and 6 months. Other individual features of the studies are shown in and . Patient exclusion criteria in all studies were similar. Differences in design have been noted above. For the reasons stated above, these experimental protocols likely represent the lower limit of the effectiveness of EX-NTX.
Meta-analysis of 6 month EX-NTX studies.
The results from the four studies reporting six month results may be pooled directly. Together these four studies document the course of 292 EX-NTX patients, 136 oral naltrexone patients, and 253 placebo-treated patients. The numbers reported to be drug-free (either as clinical abstinence or opiate-free urines) were 132, 25, and 44 respectively, representing abstinence rates of 45.2%, 18.4% and 17.4% (). When the four chisquared tables implies in are considered serially they are significant with an adjusted odds ratio of 6.00, (95% C.I. 3.87–9.50, P < 10−10).
Adverse events in trials
The study by Kunoe was the only one to document a death in the program. In that study one patient died in the placebo group. One patient assigned to the NI group left hospital on day three before he could commence NI. Therefore, he was allocated to the NI group in the intention-to-treat analysis (ITT) and to the placebo group on a treatment completion (TC) basis. Also pertinent to this question is the report of a follow-up study conducted after a large Russian trial of 306 patients. Whilst no deaths were reported in the NI group, there was one in the oral naltrexone group and four in the placebo group.103
Interestingly it has been reported that of 45,000 patients treated with NI in the USA, only 19 deaths in such patients have been reported to the FDA. Only one of these has been ascribed to possibly being associated with the NI.103
Adverse effects were generally reported as mild to moderate. Most involved the injection site and were readily manageable, although on occasion the implant was removed due to infection.
Three study reports2,17,19,101
mentioned that liver enzyme elevations were not clinically problematic at the usual doses employed. This contrasts with the FDA black box warning, relating to a potential elevation of serum liver enzymes, which appears on the product information for oral naltrexone, and which was derived from studies of the use of the drug at high dose. Four reports specifically stated that there was no cases of implant removal by patients.
Two other theoretical risks were also discussed by some authors but not observed in practice. These were the risk of opiate overdose by overriding the competitive antagonism of the antagonist-induced blockade and the risk of late overdose after the implant had worn off.
Naltrexone’s side effects are well established. Indigestion, insomnia, and depression are commonly reported in opiate-dependent patients, but these generally relate to residual detoxification symptoms.104
They are readily managed in the clinic. In patients who are truly drug-free, naltrexone can be administered almost without apparent effect. Indeed, it is so benign that it has been referred to as a “non-drug” by leading researchers from NIDA.9
Of more concern has been the hypothetical overdose risk, which in the case of naltrexone may relate to the dual scenarios of patients either attempting to override the competitive blockade effect whilst the implant is still active, or alternatively a return to their use after the period in which the implant is active. The first scenario has not been reported clinically. As noted below, the second case is only seen very rarely.
Tissue reactivity has been a major concern with implantable NTX or EX-NTX. Recent studies of the Australian implant, both in rat and in human trials, have not found severe local reactions. In 58 biopsies taken from clinical samples from 1 to 38 months, a moderate level of fat necrosis was seen in the first 6 months, falling to a mild level over the next 12 months. Some of this was ascribed to the local response to surgery. No foreign body reaction, fibrosis, or chronic inflammation was detected. In most cases no implant material was detectable at 25 months or later.105
Similar largely benign results were obtained in rats.106
A similar ultrasound study performed in 71 patients with the Perth NI in situ on days 2 to 1808 after implantation showed that the while implant beads were originally dispersed throughout the subcutaneous tissue, they gradually coalesced into a single lump as biodegradation ensued. No local naltrexone material was detectable by day 1201. These changes correlated with clinical findings on palpation.107
The suggestion that NI may be associated with inordinate death rates, as has come from certain uncontrolled case series108,109
has generated significant controversy and has formally been refuted by many authorities.103,110–117
As noted above, elevated death rates have not been a feature of any of the randomized trials of NI, nor has it been problematic either in 45,000 US patients treated, nor in the follow-up period after completion of some of the large Russian clinical trials.103
Moreover the issue has been carefully addressed in at least two clinical reviews of large case series. One study from Perth compared the survival experience of 341 NI treated patients with 553 methadone registrants treated between 2001 and 2002 and observed for 1,594 and 2,572 patient-years. They found 6 deaths in the NI group, a crude mortality of 1.8%, and an age standardized mortality compared to methadone of 0.645 (0.123–1.17). In four cases where the cause of death was known, death occurred by motor vehicle trauma in two cases—one of which was suicide— while the other two deaths were caused by a fit and an accidental polydrug overdose. This last death occurred 1026 days after the last NI was administered. One further patient was treated with both methadone and NI, dying 277 days after methadone and 899 days after NI, from a combined polydrug overdose. Deaths in the methadone group largely reflected overdose mortality, as is generally seen.118
This group has also done careful follow-up work with a cluster of especially chaotic patients whose clinical course was marked by repeated overdoses, and who, in the normal course of events, might have been expected to experience an inordinate elevation of their mortality.119
They found that prior to NI antagonist treatment, 7 patients had 82 events requiring hospital admission over 34.4 patient years of observation. However, after NI treatment 8 patients experienced 21 events in 98.5 patient-years of observation, whereas 234.4 would have been expected at the prior rate. This represents a standardized mortality rate of 0.0009 compared to their pre-treatment mortality. The two time periods measured were 1,794 and 4,494 patient days respectively. The two event rates were 0.0457 per patient-day (99% C.I. 0.0345–0.0603) and 0.0047 (99% C.I. 0.0015–0.0111) which are very clearly non-overlapping (Fisher test: O.R. = 0.1023; 95% C.I. 0.0599–0.1674, P
). This work was subsequently expanded to show a reduction from 21 overdose episodes in 20 clients in the six months prior, to zero overdoses in the six months after NI treatment.120
A second study also from Australia followed 255 NI and 2,518 buprenorphine treated patients for 1,332 and 8,030 patient-years.121
The crude mortality observed in each group was 3.00 and 5.35 per 1000 patient-years, and the age standardized mortality rate of NI compared to buprenorphine was 0.676 (95% C.I. 0.014–1.338). Most age and sex stratified subgroup comparisons within this study favored the NI group (P
= 0.002). Death in the NI group occurred 324 to 2,209 days after implant insertion. Identified causes of death in these NI patients included suicide, murder, polydrug overdose (at 324 days), and cerebral abscesses and status epilepticus from intravenous injection of buprenorphine. In particular the in-treatment mortality for NI patients, where treatment was defined as the period for which the implant was working, was 0.00 in 258 patient-years.
Together these two studies present reassuring registry-controlled data from 596 patients and 2,927 patient-years of observation.
Combined with other remarks, such as the comment that only one report of death in 45,000 patient treatment episodes reported to the FDA was believed to possibly be implant-related and the fact that there appear so far to be no mortality in the 305 treated Russian patients post-trial,103
the overall picture of mortality emerging from the literature after NI is reassuring.
Whilst it is widely said that naltrexone may have poor clinical uptake because it does not have the euphoric effects of agonist treatments, in practice there does seem to be keen demand in many countries for such treatments. The congruence between the often stated wish of drug-dependent patients for the drug-free lifestyle and the non-addictive nature of antagonist based treatments has been noted above.6
It is difficult to argue that patients should not at least be given the option to choose which treatment pathway they wish to pursue. In this context, the right to choose multiple treatment episodes and different treatment modalities at different phases of their addictive history appears to be difficult to refute.
Place in therapy
The array of treatments provided to opiate-dependent patients at the present time appears to be related to the confluent factors of the historical availability, development of agonist treatments, and the early failure of implantable forms of antagonists apparently related to difficulties of local tissue reactivity. However, the present range of treatment options was the result neither of rational clinical trial design nor of an adequate understanding of the fundamental nature of the deranged pathophysiology of opiate addiction. At the time of writing, opiate addiction is still not completely understood either in its neurobiological or its immunostimulatory-immunosuppressive dimensions, and the definitive series of clinical trials comparing current agonist based and NI treatments is yet to be conducted.
Data presented in this review shows that issues of efficacy, safety, and patient uptake have been addressed beyond reasonable dispute. Naltrexone implants and depot injectables are available from several countries. As the treatment becomes more widespread, and as regulatory authorities permit their increasing use because of this gathering data, it seems that a more rational, evidence driven approach to treatment selection will become possible.
Whether psychological testing or pharmacogenomic approaches will allow the prediction of optimum treatment routes based upon personal genomic or epigenomic profiling is a subject for further research.