Despite considerable research effort, self-injurious behavior (SIB) continues to be a primary reason, together with aggression toward others, that individuals are either retained in institutional (restrictive) environments or are administered psychotropic medication. Today, SIB remains unmanageable, expensive ($150,000–$500,000/year/patient in institutional settings, with national costs well over $3,000,000,000 [1]), and is often life threatening. It is surprisingly prevalent, occurring in ∼30% of individuals with developmental neurological complications, including those with autistic disorders [2,3].

Twenty-five experts invited by the National Institute of Child Health and Human Development (NICHD) to discuss the relations among the genetic, neurobiological, and behavioral causes and treatments for SIB reached two general conclusions. Intentional acts of harm to self, evident in many species, (a) have no known cause and (b) no agreed upon treatment [4]. The apparent absence of visible progress in understanding or treating SIB is not because of a lack of interest or effort. Studies of self-injury have increased exponentially over the past 30 years, rising from just 60 published studies between 1980 and 1984 to over 1,700 studies reported between 2005 and 2010 (Figure 1).

One major obstacle in understanding the mechanisms of SIB and developing coherent treatment plans is the absence of distinctive behavioral phenotypes. Despite the consensus that SIB has variable expression with no known cause, the group of NICHD experts agreed that SIB could be defined, perhaps with greater precision than most complex human behaviors. SIB is a directly observable behavior that can be reliably counted. The NICHD group argued that data collection and analysis had advanced so that complex patterns of SIB should replace or supplement measures of rate and frequency [5,6]. Two distinct patterns of SIB were proposed as possible guides. One pattern consists of bouts that are most likely maintained by environmental contingencies. The second pattern involves protracted periods of SIB that are most likely under the primary influence of biological factors. The vast majority of existing studies, however, have reported frequencies or rates of occurrence of SIB often linked to a single environmental manipulation. It is a significant advantage that many forms of SIB can be counted and time-stamped enabling contemporary studies to use sequential and time series procedures to define their structure and their relations with other behaviors and the environment. We subjected extensive and lengthy observations of maladaptive behavior in its natural environmental context to time series analysis and discovered unique temporal and sequential patterns of these severe maladaptive behaviors [2,3,7-10].

Specifically, we found for a large majority of the individuals studied, SIB was predicted only by its own recent history. In our cohort of severely self-injurious patients, SIB was expressed consistently in successive occurrences revealing a unique pattern of sequential dependence compatible with a “contagious” distribution. Application of time-series methods of analysis that controlled for chance pairings of events indicated that the contagious patterns of SIB were independent from frequency and rate of occurrence. That is, the temporally dependent patterns we observed were not a function of high rates of occurrence. Moreover, and surprisingly, SIB was not associated consistently with other behaviors or with the several staff activities or environmental conditions recorded in these studies. Thus, in a significant majority of these individuals, SIB episodes were self-perpetuating and not related to antecedent or subsequent environmental circumstances, events, or other recorded behaviors.

This novel and surprising finding was consistent with conclusions of the NICHD group that some expressions of SIB may have an underlying biological basis because a solely self-perpetuating behavior is most parsimoniously explained by internal (i.e., biological) motives [2,11-13]. The vast majority of individuals in our cohort exhibited the most primitive level of internally regulated behavioral patterns despite years of behavioral interventions and treatment with various medications [8,9]. Anecdotal and clinical observations of these individuals also strongly suggested a biological basis for their behavior and specifically involvement of the pain and pleasure systems. Typically, SIB is repetitious consisting of hourly, daily, weekly, monthly, or even yearly cycles [14]. Some individuals who repeatedly injure themselves appear immune to the normal experience of pain [12]. They abuse and injure their bodies, hitting or biting themselves, hurling themselves to the ground, and banging their head against solid objects resulting in broken bones, disfigurement, blindness, and even loss of life [2,15,16]. They often work to overcome interventions designed to decrease self-injury in a manner that is consistent with seeking positive reward. For instance, protective devices (such as helmets) may result in individuals exerting greater effort and exhibiting greater rates of behavior to hit and harm themselves [17].

Because many medications that treat pain or induce pleasure are addicting, it is interesting that SIB, in addition to the obvious involvement of pain systems, shares features of addiction such as compulsive and ritualistic (or stereotypic) patterns that either comprise or surround the self-injuring acts. One biological system that has been implicated in SIB and the modulation of pain and pleasure is the hypothalamic-pituitary-adrenal (HPA) stress axis and specifically the proopiomelanocortin (POMC) molecule [3,18,19].

Thompson and Caruso [6] recognized that some forms of SIB were “neurochemically driven and independent of environmental events.” Early studies either of basal or of resting levels of a variety of peptides, proteins, transmitters, and amines in plasma or cerebrospinal fluid from patients exhibiting SIB, however, generated inconclusive results. This was not surprising because there was little consistency among studies regarding the rigor of diagnosis, the molecule measured, the tissue assayed, or the conditions assessed [20-26].

In our initial study to examine the relation between circulating endogenous opioids and response to naltrexone, we collected blood samples from ten patients within 2–5 min of a self-injuring act and during a control period [18]. At least 1 month later patients were administered three different doses of naltrexone in a double-blind, placebo-controlled crossover study over a 10-week period. All patients were videotaped during the study and behavior was coded with a computer-assisted program. Patients with the highest change in plasma levels of BE after SIB had the most and statistically significant positive response to naltrexone. These results were consistent with several other reports. First, Ernst et al. [21] reported that baseline levels of BE were positively related to changes in behavior (clinical global impressions, CGI) after treatment with naltrexone in five young autistic children. Second, Bouvard et al. [38] found that C-terminal BE decreased after naltrexone only in good responders. Third, Scifo et al. [63] found that increases in SIB and response to naltrexone in some patients, were related to high levels of endogenous opiates (i.e., good responses to naltrexone were observed in patients with high levels of BE). Fourth, Cazzullo et al. [64] reported that patients responding with decreased BE levels after treatment with naltrexone had better and more pervasive behavioral improvement than patients who did not have physiological changes after naltrexone.

In a follow-up study of nine additional patients (total of nineteen), we [3] found that plasma BE was uncoupled from the usually coreleased ACTH [65-71] after an episode of SIB. This unusual pattern was not a function of time of day that blood was sampled, and it confirmed our earlier observations [7,18] and provided additional support for this specific biological marker among a diverse group of subjects who share a behavioral aberration. In addition, stronger support was generated for the effectiveness of naltrexone in reducing SIB.

Positive responses to low doses of naltrexone were observed in subjects who did not exhibit increased BE after SIB. That is, low doses of naltrexone were effective in reducing SIB only in subjects either who did not exhibit a surge in BE after SIB or whose baseline level exceeded the level after SIB. The relation between BE and response to the lowest dose of naltrexone was consistent with our earlier results, and statistically significant with the addition of nine subjects. Previously we [3] suggested that SIB had functional significance because it increased endogenous opiates and thereby delivered positive consequences (i.e., pleasure/euphoria/pain modulation). We argued that the highest dose of naltrexone most effectively blocked this mechanism in subjects with the highest levels of BE after SIB. The results from the follow-up study suggested an alternative possibility related to baseline levels of, or baseline relations between, POMC peptides. Because the lowest dose of naltrexone was most effective in subjects with the highest (morning) baseline (relative to post-SIB) levels, we speculated that the association between baseline ACTH and BE could influence the response to naltrexone (based on evidence that supported reciprocal functions of BE and ACTH [72]]. If our speculations were accurate, subjects with the greatest difference between morning BE and ACTH levels would be the most responsive to low doses of naltrexone because there would be less attenuation of the opioid influence. The test of this possibility confirmed our speculations because we found that subjects with high levels of morning (chronic) BE and low levels of ACTH were associated with positive responses to the low dose of naltrexone. This possibility may be compatible with findings that chronic exposure to opioids resulted in supersensitivity to the effects of low doses of opiate antagonists [73,74].

These findings have since been extended using a fundamentally different method for quantifying changes in the temporal patterning of SIB following treatment with naltrexone. The THEME method developed by Magnusson [75,76] has been used to detect highly significant, nonrandom, hierarchical patterns in the temporal organization of SIB with respect to other observed behaviors. Kemp et al. [10] reported that these temporal patterns (“T-patterns”) of SIB were significantly correlated with basal levels of BE. In the subgroup of subjects receiving naltrexone (discussed above), the percent change in these T-patterns of SIB (between weeks the subjects were receiving placebo and weeks they were receiving naltrexone) was found to be significantly correlated with post-SIB (i.e., samples were collected immediately after SIB) levels of the N-terminal fragment of beta-endorphin (βEN). No such correlations were found for changes from placebo on any “control” T-patterns (those containing non-injurious, stereotyped behaviors or staff interactions) nor for any samples collected during other periods. These results are shown in Table 1.

We have made similar observations in our long-term studies of naltrexone and SIB [19]. POMC fragments were measured in twelve self-injurious patients before and after long-term (3-month) treatment with naltrexone. POMC fragments were sampled from blood collected at the end of the baseline and placebo-controlled treatment phases of the study. Two patterns emerged. One group (responders) displayed persisting improvement in SIB and lower relative levels of BE after initial exposure to naltrexone. Chronic administration of naltrexone to this group was associated with increased SIB and elevated relative levels of BE. Return to placebo improved their behavior (reduced SIB) and their levels of BE returned to basal levels. The second group (nonresponders) was characterized by absence of persisting improvement after acute treatment with naltrexone and by elevated basal BE levels. Chronic treatment with naltrexone improved their behavior but did not alter their BE levels. Long-term positive responders to acute doses of naltrexone were associated with less disregulation of ACTH and BE.

When it was established that the body had its own opiate system [77], the endogenous opiates became prime suspects responsible for maintaining SIB. Perhaps individuals who self-injure have elevated thresholds for pain or derive pleasure from painful stimulation. Exposure to, or levels of, endogenous opiates could explain these possibilities. Reduction in SIB following treatment with opiate blockers would provide evidence for the opiate hypothesis of self-injury. Results from studies to test these possibilities, however, are complex. The complexity is related primarily to the fact that patients exhibiting SIB and evaluated after treatment comprise a mixture of etiologies, pathologies, and motivations. Despite the tremendous amount of error introduced with a heterogeneous population, there is substantial evidence that opiate blockers are efficacious in reducing SIB.

The observations that opiate blockers reduce SIB are important for at least two reasons. First, naltrexone produces a clinically significant reduction in a serious and life-threatening behavior for some individuals typically who have not been responsive to any other type of treatment. We have observed startling improvements in individuals who have failed all rational treatments. Some adults in our studies have had protective headgear discontinued for the first time since early childhood. Others have developed adaptive skills and have acquired the ability (or the privilege) to leave institutions for the first time in their lives after treatment with naltrexone. Second, the results with naltrexone are important because they suggest that a specific biological system may be disregulated in a subgroup of patients. Because the opiate blockers have few effects in the absence of opiates [78], effective treatment with these drugs must engage the endogenous opioid system. Reports that resting levels of endogenous opiates or levels of endorphin after an SIB episode predict positive responses to opiate blockers provide support for this assumption and the foundation for rational treatment strategies based on biological criteria.

From the current review we can draw several conclusions. There is consensus in the literature that doses between 1.0 and 2.0 mg/kg or a fixed dose of 100 mg are the most effective for reducing SIB [16,53,54]. At these doses, at least half of the chronically self-injuring patients exhibit at least a 25% reduction in their behavior. It is important to acknowledge that the studies that have reported reductions of 25% or greater, typically conducted direct observations of the patients and did not rely on global clinical ratings. As reviewed by Symons et al. [53], male patients respond more favorably to naltrexone treatment than do female patients. There is consensus that naltrexone is a safe drug without major or contraindicating side effects. Most individuals entered into naltrexone trials exhibited the most severe SIB for which all other forms of treatment had been ineffective. Often their behavior presented life-threatening consequences and always their SIB prevented them from enjoying a less restrictive environment. Against this background, the possible benefits of reduced SIB by treatment with naltrexone exceed the risk of side-effects. We are aware of clinical decisions to treat SIB with naltrexone even when patients presented with risk factors such as chronic hepatitis [79]. To our knowledge, there have not been serious side effects solely due to administration of naltrexone among MR/NDD patients. There is consensus among the studies that naltrexone is an effective treatment because the endogenous opioid system is engaged by SIB [3,16,18,19,57]. SIB has characteristics that resemble addictive behavior (compulsive, ritualistic, destructive) and altered pain threshold. Both of these characteristics implicate the opioid system and support the logic of opiate blockers as reasonable treatments. There is consensus that long-term treatment of SIB with naltrexone apparently is not harmful and may be effective. There is consensus that generally, treatment with naltrexone appears to be effective in about half of the adult patients examined. Following effective acute treatment with naltrexone some patients have shown a rebound to pre-treatment levels [61]. There is consensus that naltrexone should be avoided during periods when patients are known to be in pain requiring narcotic analgesics, such as surgeries or with bone fractures. Alternately, nonnarcotic analgesics should be administered to patients receiving naltrexone [80]. Finally, it should be noted that there are reports of paradoxical increases of SIB following treatment with naltrexone [81]. In our own research [61], we have observed increased rates of SIB in individuals who showed sustained improvements throughout a 12-month hiatus following a brief period of acute treatment with naltrexone, and then were given longterm treatment (3 month) with naltrexone.