Clinically-meaningful Individual Differences in Opioid Withdrawal Expression

Abstract

Opioid use Disorder (OUD) is a significant public health concern. An individual with an OUD may experience withdrawal after stopping opioid use. There has been limited exploration of the individual differences in withdrawal expression. This study expands understanding of this issue by examining the presence and frequency at which persons who have ever had opioid withdrawal have experienced different opioid withdrawal symptoms. Using cross-sectional data captured online from Amazon Mechanical Turk, 124 adults with a lifetime experience of opioid withdrawal were included. Respondents were able to indicate ever experiencing 31 individual opioid withdrawal symptoms. If a symptom was ever experienced, respondents would indicate if it was common and whether it bothered them. A cluster analysis was used to explore variability between the withdrawal symptoms. The sample was primarily men (n=76, 61.3%) with an average age of 34.7 (SD=11.6). The typical withdrawal syndrome lasted 6.5 days (SD=4.9) and was most severe at 5.7 (SD=4.9) days. Lifetime endorsement of individual symptoms ranged from a high of 73.4% (anxious) to a low of 43.5% (nausea). The cluster analysis was significant (F(1,122)=215.6, <0.001) with good BIC (0.7). The two clusters are conceptualized here as HIGH (N=73; 59%) and LOW (N=51; 41%) endorsing, with a mean of 21.9 and 8.5 items endorsed. These data add to prior studies by suggesting high variability in the individual expression of opioid withdrawal symptoms. It may be time for the field to develop a consensus regarding opioid withdrawal symptom expression and measurement to enhance clinical care.

Introduction

Opioid use disorder (OUD) is a major public health problem and there are considerable efforts underway to expand access to medications for OUD (MOUD), the gold-standard treatments for OUD. Persons who develop physical dependence on opioids following periods of extended consumption will experience a prominent physical withdrawal syndrome during a period of abstinence. Withdrawal is generally characterized by an acute (3–7 day) syndrome that includes symptoms such as insomnia, gastrointestinal upset, and autonomic hyperactivity (Dunn et al., 2019). Persons often cite withdrawal management as a primary treatment goal, and a growing number of MOUDs (e.g., methadone, buprenorphine, lofexidine) have been developed to specifically mitigate the severity of opioid withdrawal symptoms.

Contemporary understanding of withdrawal is founded on characterizations of persons being withdrawn spontaneously from opioids more than 50 years ago (Andrews, 1944; Himmelsbach, 1942, 1943; Jasinski, 1981) and despite advancements in the understanding of OUD there has been relatively limited exploration of the individual differences in the expression of withdrawal symptomatology over the past several decades. An inherent consequence of the current method of assessing withdrawal severity using both clinician-administered and self-report rating scales is the assumption that opioid withdrawal is a common and unitary syndrome. This could lead to the perception that all persons have equal opportunity to express the symptoms being assessed. However, growing evidence suggests withdrawal is highly variable across individuals such that individuals experience different arrays of symptoms. For instance, studies have now reported preliminary evidence of different opioid withdrawal phenotypes (Dunn et al., 2018), and differences in withdrawal severity based on gender and presence of chronic pain (Ware et al., 2022).

To date there has been no examination of the breadth and scope of individual differences at the level of specific opioid withdrawal symptoms. If experiencing withdrawal is variable, this suggests that not all individuals may have the same potential for experiencing all symptoms, and that each individual may have a unique upper limit on their maximal withdrawal severity (rather than the same unitary opportunity to experience the maximum scale rating). A consequence of this is that individuals who do not routinely express all withdrawal symptoms may be perceived as not experiencing a severe withdrawal syndrome on a rating scale, despite having experienced all symptoms they generally express during the syndrome. Put another way, reliance on rating scales may cause the withdrawal severity experienced by some patients to be under-estimated by a provider, which could have downstream implications for their withdrawal management. These implications may include an increasingly arduous withdrawal experience for patients and clinicians not adequately treating the full range of withdrawal symptoms, or providers.

Data have also demonstrated that differences in the prevalence of individual differences in opioid withdrawal expression and symptomatology may be clinically-meaningful. For instance, one study that administered naloxone to precipitate withdrawal in persons with OUD found that differences in response to naloxone predicted patient withdrawal severity during a subsequent clinical taper (Dunn et al., 2018). Another exploration of individual differences found profound variability in the percent of participants who ever endorsed various self-report and observed symptoms on standard measures of withdrawal during a clinical taper, ranging from vomiting (reported by 37.2% of patients) to increase in heart rate (97.1%) (Dunn et al., 2020). Together, these data provide initial evidence that withdrawal may vary in a clinically-meaningful way across individuals. Given the growing emphasis on developing new MOUDs for managing opioid withdrawal, combined with evidence that individuals may express different symptoms and respond differently to MOUDS, there is significant value in more deeply exploring individual expression in withdrawal to understand its prevalence and related consequences in terms of care. The following study sought to contribute to this understanding by assessing the likelihood that persons who have OUD have ever experienced one or more from an array of potential symptoms of opioid withdrawal. We hypothesized there would be pronounced individual variation in the symptoms reported, supporting the notion that the opioid withdrawal syndrome experienced across individuals is not necessarily uniform.

Method

Recruitment and screening

This study was conducted by using the crowdsourcing platform Amazon Mechanical Turk (MTurk) to recruit survey respondents between 10/2021 and 11/2021. While the goal of the study was to learn more about individual variation in opioid (and other substance) withdrawal, the survey was advertised as a “survey on health behaviors” to mask the purpose of the survey prior to eligibility determination. Respondents who lived in the United States were eligible to complete a brief screening form to determine final study eligibility. To be eligible for these analyses, respondents had to be >18 years old, endorse >1 lifetime use of opioids, and report experiencing >1 lifetime instance of opioid withdrawal. Of the 468 respondents who completed the eligibility screener, 295 (63.0%) were eligible, 287 (61.3%) agreed to participate and 247 (52.8%) answered a series of multiple choice and qualitative embedded quality check questions correctly, and 124 endorsed opioid use (26.5%) as their primary substance of use and were included in the final evaluable sample. This study was acknowledged by the Johns Hopkins University IRB. Respondents earned $0.15 for completing the eligibility screener and eligible respondents earned $3.00 for completing the survey.

Measures

Demographic and Drug-Use Characteristics

Self-reported age, gender, race, ethnicity, education level, living area (urban, suburban, rural), marital status, and income level were collected. Respondents also indicated on a checklist (yes/no) whether they’d used any substances from a list of opioid and non-opioid substances in the past 30 days.

Opioid Withdrawal

Respondents were asked general questions about their withdrawal experience, including whether they had experienced the full course of withdrawal (defined as the “beginning to end of withdrawal”) in the past 6 months and 60 days (yes/no), for how many days their (1) “typical” and (2) “last” complete withdrawal syndrome had lasted and (3) on which day their worst symptoms generally present (answered on a continuous measure from <1 to >40 days). They were then presented with 31 potential opioid withdrawal symptoms that were sourced from two commonly-administered scales (e.g., Clinical Opiate Withdrawal Scale [Wesson & Ling, 2003]; Subjective Opiate Withdrawal Scale [Handelsman et al., 1987]), which were cross-referenced to include unique symptoms reported by other subjective scales (Opiate Withdrawal Scale [Bradley et al., 1987]; Subjective Opiate Withdrawal Questionnaire [Loimer & Grunberger, 1991]) and early reports about withdrawal expression (Himmelsbach, 1942; Wang1974). (see Figure 1). Respondents were provided a list of all queried symptoms and asked to endorse “which symptoms you experience when stopping using opioids” (yes/no). For all endorsed symptoms respondents were asked whether the symptoms were ones they (1) “usually experience” and (2) “bother you” (yes/no) (as metrics of frequency and clinical significance). Several simple multiple choice (“what best describes you [human, machine]”) and open-text (“please type the following statement: Have a nice day”) quality control and attention checks were embedded throughout the survey. Any response that did not achieve 100% accuracy on these checks was excluded from analyses.

Figure 1. Individual opioid withdrawal symptom lifetime experiences.

Bars present percent of respondents (Y-axis) who endorsed having no lifetime experience (black bars) of a symptom in response to opioid abstinence, relative to percent of respondents who endorsed having some limited lifetime experience of a symptom (dark gray bars) and those for whom a symptom was commonly experienced during a period of opioid abstinence (light gray bars), as a function of the queried opioid withdrawal symptoms (X-axis).

Statistical analyses

This study hypothesized that respondents who experienced opioid withdrawal would have variability in their expression of symptoms and that some respondents would report never having experienced some of the queried symptoms. Study analyses focused on descriptive assessments of symptom expression as well as determining whether symptom patterns were clustered meaningfully within patient subgroups, as an initial step toward identifying meaningful differences in withdrawal expression. General demographics and questions related to the overall withdrawal experience are presented descriptively. Ever experiencing a symptom was considered the primary outcome as it represents the most conservative method for identifying the expression of a symptom (e.g. independent of whether it was common and/or bothersome). To explore this relationship further, the total number of symptoms ever experienced were summed (range 0–31) and analyzed using a cluster analysis. A two-step cluster analysis applying log-likelihood distance was applied to inform Bayesian Information Criteria (BIC), which yielded a 2 or 3 cluster solution. Hierarchical clustering, applying between-groups linkage with Euclidean distance, further supported a 2 or 3 cluster solution. Examination of the cluster assignment revealed appropriate distancing in the 2-cluster solution, which is termed “HIGH” and “LOW” here for ease of interpretation. Cluster assignments were generated using K-means clustering and used as the basis for between-group comparisons of demographic and drug use characteristics, compared using chi-squared analyses for dichotomous outcomes and independent groups t-tests for continuous outcomes. All analyses were conducted in SPSS version 28 with alpha set at 0.05. The dataset used for this study will not be made publicly available. A preliminary working draft of the results was added as an online preprint for approximately 20 days, from early September 2022 to September 27, 2022. The document was removed by our request on September 27, 2022.

Results

Respondent Demographics

Table 1 provides the demographic and substance use characteristics of the sample. The sample was primarily men (n=76, 61.3%) who lived in an urban area (n=76, 61.3%) with an average age of 34.7 (SD=11.6).

Table 1.

Sample Characteristics

	Mean (SD) or N (%)
Age, Mean Years (SD)	34.7 (11.6)
Past 30-day heroin use, Mean Days (SD)	7.8 (5.9)
Past 30-day fentanyl/other synthetic opioid use, Mean Days (SD)	9.0 (7.9)
Past 30-day prescription opioid use, Mean Days (SD)	11.0 (7.4)
Male (n, %)	76 (61.3%)
Living Environment (n, %)
Urban/City	76 (61.3%)
Suburban/Suburbs	29 (23.4%)
Rural/Country	19 (15.3%)
Working full time (n, %)	117 (94.4%)
Ever overdosed on opioids (n, %)	93 (75.0%)
Ever injected a drug (n, %)	87 (70.2%)
Experience “usual” withdrawal symptoms (n, %)a	111 (89.5%)
Experienced opioid withdrawal in past 6 months (n, %)	106 (85.5%)
Experienced opioid withdrawal in past 60 days (n, %)	105 (84.7%)

N=124; SD= Standard Deviation

^a- “usual” withdrawal operationalized as a syndrome typical for that study respondent

Withdrawal Outcomes

Most respondents had experienced opioid withdrawal recently (within the past 6 months; [n=106, 85.5%] and past 60 days [n=105, 84.7%]). Their typical opioid withdrawal syndrome lasted for an average of 6.5 days (SD=4.9, range 1–30) and was most severe at 5.7 (SD=4.9, range 0–29) days, though the most recent mean withdrawal experience lasted an average of 8.8 days (SD=5.9, range 0–30).

There was substantial variability in the percentage of respondents who never, rarely, and commonly experienced each symptom (Figure 1) and who found the symptom bothersome (Supplemental Table 1). Lifetime endorsement of individual symptoms ranged from a high of 73.4% (anxious) to a low of 43.5% (nausea) and no individual symptom was endorsed by all 100% of the respondents. Anxious (73.4%), headaches (68.5%), and depression (62.1%) were rated the most frequently endorsed lifetime symptoms. Anxiety (63.7%), headaches (52.4%), and depression (49.2%) were also the most commonly-experienced symptoms, and anxious (47.6%), headaches (40.3%), and cravings (40.3%) were the most bothersome symptoms.

Cluster analyses supported a 2 and 3 cluster solution, and examination of cluster assignment revealed the 2-cluster solution had better between-cluster differentiation than the 3-cluster solution. The cluster analysis was significant (F(1,122)=215.6, <0.001) with good BIC (0.7). The two clusters are conceptualized here as HIGH (N=73; 59%) and LOW (N=51; 41%) endorsing, with a mean of 21.9 and 8.5 items endorsed, respectively. Figure 2 displays the percent of respondents in each cluster who reported having ever experienced a symptom. As can be seen in the figure, fewer than 50% of persons in the LOW endorsing group reported experiencing any symptoms with the exception of anxiety and headaches. In contrast, every symptom was endorsed by >50% of the HIGH endorsing sample. Between-group comparisons of these groups revealed no significant differences in gender χ2 (1)=.22, p =0.64, ever injecting a drug χ2 (1)=1.23, p =0.27, and ever overdosing on opioids χ2 (1)=1.88, p = 0.17. Similarly, no significant age differences were found between the HIGH (M=34.2, SD=12.3) and LOW (M=34.2, SD=10.7) clusters t(122)=−.43, p=0.67.

Figure 2. Endorsement of >1 lifetime experience of individual symptoms.

Data represent LOW (open circles) and HIGH (closed circles) withdrawal severity clusters based upon hierarchical clustering analyses. Y-axis presents percent of respondents from each cluster endorsing lifetime experience of queried opioid withdrawal symptoms (X-axis).

Discussion

These data add to the limited evidence suggesting there is profound individual variation in how opioid withdrawal is expressed across persons with OUD. Variability was present both in the lifetime expression of withdrawal experience as well as whether the symptoms were common and perceived as clinically-meaningful by respondents. A formal cluster analysis further suggested that differences in the endorsement of symptoms clustered into distinct subgroups across respondents. There are a few clinical implications from these data. First, they support prior demonstrations of individual variability in the individual expression of withdrawal. Second, they provide evidence that persons with OUD likely express withdrawal differently. This suggests that assuming all individuals can achieve the same level of severity on a common rating scale could be systematically underestimating withdrawal severity of persons who were identified here as being in the LOW withdrawal cluster group. This underestimation is problematic as it may lead to the opioid withdrawal syndrome being inadequately treated for some individuals and could stifle discovery for innovative withdrawal remediation strategies. Collectively these data support additional prospective research into the mechanisms and clinical impact of these differences.

Although withdrawal management is a core goal for OUD treatment, the manner by which withdrawal is measured is highly decentralized. A recent scoping review identified eighteen different opioid withdrawal scales that assess 10 – 550 symptoms (Nuamah et al., 2019). The variability in items across scales suggests there is no general consensus in the field with regard to what constitutes an opioid withdrawal syndrome and it is likely that some important symptoms may be routinely overlooked. In this study, the three symptoms that were rated as being the most commonly experienced and/or most bothersome (anxious, headaches, depression) are not uniformly collected across all withdrawal measures. These collective findings are troubling because they suggest that using questionnaire-based thresholds of mild, moderate, or severe withdrawal for clinical decision-making could mask the personalized nature of withdrawal. The data collected here also suggest that some of the physiological symptoms that are weighted more heavily in withdrawal scales (and thus elevate the score more quickly) such as vomiting, diarrhea, and tremors, were only commonly experienced by a third of respondents. In a clinical setting, individuals who do not endorse or display these symptoms may not be identified as being in an advanced state of withdrawal despite being elevated on all the symptoms they expect to experience. It is possible that low endorsement of symptoms reflects a less severe physical dependence, and that all individuals have the same future potential to experience symptoms. Symptoms may also differ by the primary opioid used, the duration of use, or comorbid polysubstance use (none of which can be assessed here). Yet it is also possible that some individuals will never express some symptoms of withdrawal. These are important empirical and clinical questions that warrant more investigation. Future investigations may include mixed methods studies directly asking individuals open-ended questions about their opioid withdrawal experiences.

This study is limited by its retrospective review of lifetime experience of opioid withdrawal symptoms, which is subject to recall bias. In the interest of brevity and the focus on the etiology of various symptoms, additional information (such as the duration of opioid use prior to withdrawal, previous experience with MOUD and/or other opioid withdrawal treatments, and concurrent use of polysubstances) that would help establish a potential mechanism for the differences observed was not collected or analyzed. It is also not possible to verify opioid use status and other features of use that may help distinguish the HIGH versus LOW endorsers. It is crucial for additional work in this area to more thoroughly assess differences in respondents’ demographics and/or drug use characteristics to elucidate whether differences in withdrawal expression are state (e.g., related to acute dependence) or trait (e.g., related to some underlying physiological mechanism) in nature to help advance a more personalized opioid withdrawal management for persons with OUD.

Conclusion

The contemporary management of opioid withdrawal is premised on the notion that patients have a common syndrome and the same upper limit for withdrawal expression. These data expand prior studies and suggest there may be high variability in the individual expression of opioid withdrawal symptoms. Assuming withdrawal is a linear experience might inhibit progress with regard to understanding the mechanistic basis for withdrawal symptoms and the development of medication or interventions. Notably, one early withdrawal rating scale advocated that endorsement of >1 symptom from a symptom cluster (e.g., gastrointestional, autonomonic hyperactivity) was sufficient evidence of withdrawal (Judson et al., 1980). This categorical definition of withdrawal as being present (yes, no) deviates from our current common practice of rating withdrawal along a severity continuum. In the context of the current opioid crisis and efforts to identify mechanistically-informed treatments, it may be time for the field to develop a consensus regarding opioid withdrawal symptom expression and measurement. Further, by exploring the value of categorical and continuous thresholds that are tailored to individuals will be an important step towards efforts to advance care for persons with OUD.

Public Significance:

We identified two classes: HIGH withdrawal and LOW withdrawal. Findings from our study suggest high variability in the individual expression of opioid withdrawal symptoms. Considering our contemporary understanding of withdrawal is founded on persons being withdrawn from opioids more than 50 years ago, developing a contemporary consensus regarding opioid withdrawal may improve measurement and clinical interventions.