Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Mar 19.
Published in final edited form as: J Subst Use Addict Treat. 2025 Jul 3;176:209752. doi: 10.1016/j.josat.2025.209752

Beyond frequency of use: Exploring additional metrics for quantifying cocaine use and their relation to life functioning and treatment outcome

Samantha Ellis a,1, Rachel Sun b,1, Angela Heads c, Jin Ho Yoon c, Heather Webber c, Joy Schmitz c, Margaret Wardle a,*
PMCID: PMC12997293  NIHMSID: NIHMS2154122  PMID: 40617507

Abstract

Background:

Frequency of use – the number of days of use – is the most common metric for quantifying cocaine use and the most common target of treatment. However, frequency may not capture all important aspects of cocaine use. This study examined how other aspects of cocaine use (typical amount spent, variability etc.) relate to life functioning and cocaine use disorder (CUD) treatment outcomes.

Method:

This is a secondary analysis of 3 clinical trials for CUD, N = 207 for the life functioning analysis and N = 173 for the outcomes analysis. Eight metrics of use were calculated from baseline 30-Day Timeline Followback (TLFB) data and entered into multiple regression and hierarchical Poisson regression analyses predicting domains of the Addiction Severity Index and treatment outcomes, respectively.

Results:

Greater typical dollar amount used (i.e., generally greater amounts of cocaine used per occasion) related to better employment functioning, while better treatment outcomes were associated with: 1) less frequent use, 2) greater typical dollar amount used, 3) more variability in dollar amount, 4) less weekday use, 5) greater variability in days between uses, and 6) a downwards trajectory in frequency of use and upwards trajectory in typical dollar amount over the month leading into treatment.

Conclusions:

Beyond days of use, other aspects of use also relate to life functioning and treatment outcomes, although these aspects predicted treatment outcomes better than life functioning. Future work should explore whether other aspects of cocaine use are important targets for harm-reduction-focused treatment.

Keywords: Cocaine use disorder, Life functioning, Treatment outcomes, Frequency of use, Amount spent, Variability of use

1. Introduction

1.1. Background

Cocaine use disorder (CUD) affects over 1.5 million people in the United States and is associated with numerous negative health and social outcomes (SAMHSA, 2018). Traditionally, complete cocaine abstinence was the preferred treatment outcome. However, experts have advocated for reduced cocaine use as a potentially beneficial alternative in the past decade (Amin-Esmaeili et al., 2024; Carroll et al., 2014; Kiluk et al., 2016; Roos, Nich, Mun, Babuscio, et al., 2019a), informed by alcohol use disorder research showing reductions in “heavy drinking days” yields clinically meaningful benefits (Carroll et al., 2014; Falk et al., 2010). In support, the most recent draft of FDA guidance for medications for stimulant use disorders includes “change in pattern of stimulant use” as a preferred outcome (FDA, 2023).

However, this guidance raises the question of how to quantify a “pattern of stimulant use” (e.g., how often, how much, predictable or highly variable, stable or changing, etc.). When attempting to define alternate non-abstinence outcomes for cocaine use, researchers have favored frequency of use – the number of days used over a specific time period (Carroll et al., 2014; Roos, Nich, Mun, Babuscio, et al., 2019a). Frequency is easier to measure through self-report or biological methods compared to amount used, and researchers have clinically validated the relationship of frequency of use to DSM symptoms of CUD (Kellogg et al., 2003; Liu et al., 2020), impairments in employment and relationships, and legal involvement (Reelick & Wierdsma, 2006; Roos, Nich, Mun, Babuscio, et al., 2019a; Roos, Nich, Mun, Mendonca, et al., 2019). These findings resulted in frequency being recommended as a non-abstinence endpoint in draft FDA Guidance (FDA, 2023).

Yet other aspects of drug use beyond frequency may also be clinically significant, as observed in alcohol research. For cocaine, experts have previously dismissed amount of use as a clinically meaningful endpoint due to the lack of standard illicit drug units (Carroll et al., 2014; FDA, 2023; Kiluk et al., 2016). Although researchers can transpose illicit drug use into dollar amounts, critics have questioned the validity of dollar amounts due to inconsistent drug quality and the social or informal nature of illicit markets, which may include shared use or trading sex for drugs (Carroll et al., 2014). However, the question of whether estimates of amount capture meaningful information seems open enough to warrant empirical examination. Regarding varying quality, standard timeline followback methods for alcohol use produce actionable information without taking into account differences in alcohol by volume between, e. g., light beer and India pale ales (Marini et al., 2023). Regarding informal markets, people with experience in a market can still likely estimate the dollar value of a gift or trade. Indeed, other literature suggests that estimates of amount of cocaine capture important information: amount used per occasion is significantly associated with risk of developing CUD (Liu et al., 2020), the dollar value of cocaine used over month is sensitive to treatment effects (McKay et al., 2001), and studies using hypothetical cocaine purchase tasks show that the amount of money someone is willing to spend on cocaine both predicts treatment outcomes and response (Yoon et al., 2020, 2021). These findings support testing amount of use estimated in dollars as a potential measure of use in addition to frequency when considering a “pattern of use.”

Other “patterns of use” may also be important. Researchers have observed poorer treatment outcomes among those who use during the week vs. weekend only for cocaine (Huang et al., 2014), similar to findings for other drugs and alcohol (Bernstein et al., 2015; Lau-Barraco et al., 2016). Similarly, relative changes in use may be important, not just absolute levels - relative change in cocaine use predicts treatment response (past 90-days frequency and amount prior to treatment; Ye et al., 2015a) and life functioning (relatively decreased use independent of baseline use severity; Roos et al., 2019).

In summary, metrics of cocaine use beyond frequency may reveal clinically relevant information. The present study pooled timeline followback data recording both dates and estimated amounts used in dollars in the 30 days prior to entering treatment across three CUD clinical trials to construct multiple cocaine use metrics – frequency, dollar amount, patterning/variability in frequency and dollar amount, and changes in use leading into treatment. Each use metric was compared to the following: 1) life functioning (employment, medical status, etc.) at baseline; 2) total cocaine-negative urine drug screens during the initial 4-week phase of treatment. Although this study is exploratory and prior literature in this area is limited, we hypothesized that other aspects of use would significantly contribute to life functioning and treatment outcomes, even after controlling for frequency of use.

2. Methods

2.1. Study overview

This is a secondary analysis of three clinical trials for CUD (NCT02773212, NCT02896712, NCT04843046; Table 1). The life functioning analyses used standard screening and baseline measures from all trials, while the outcomes analysis used treatment results from the first four weeks of treatment in two studies with similar treatment protocols.

Table 1.

Overview of clinical trials in pooled sample.

Study Behavioral treatment conditions Medication treatment conditions Length of treatment Participants Location
NCT02773212 (Wardle et al., 2023) CM d-amphetamine v placebo 4 weeks 57 adults with at least moderate CUD University of Illinois at Chicago and the University of Texas at Houston
NCT02896712 (Schmitz et al., 2021) CM + ACT v CM + DC Modafinil v placebo 12 weeks 118 adults with at least moderate CUD University of Texas at Houston
NCT04843046 (Schmitz et al., 2024) Inpatient detoxification followed by CBT Pioglitazone v placebo 12 weeks 30 adults with CUD University of Texas at Houston
Open in a new tab

Note: CM = Contingency Management; ACT = Acceptance and Commitment Therapy; DC = Drug Counseling; CBT = Cognitive Behavioral Therapy.

2.2. Participants

Eligible participants were non-hospitalized, treatment-seeking individuals in Houston and Chicago aged 18–60 with moderate to severe CUD and cocaine-positive urine test at screening (a positive baseline urine test provided biological verification of cocaine use prior to treatment). Exclusionary criteria included other substance use disorders (aside from alcohol, nicotine, and cannabis), psychotic disorders, current suicidal or homicidal ideation, and current legal system involvement. Individual trials also had unique medical criteria, which can be found in the ClinicalTrials.gov registrations and published articles. Only participants reporting cocaine use in the month prior to trial enrollment were included, resulting in an N = 207 for the life functioning analysis and n = 173 for the treatment outcomes analysis.

2.3. Measures

2.3.1. Metrics of use

To obtain the raw data used to calculate our metrics, we used the 30-Day Timeline Followback (TLFB) procedure (Robinson et al., 2014; Sobell & Sobell, 1992). The TLFB is a calendar-aided self-report of cocaine use in dollars on each day over the past month, supported with standard memory aids. Per best practices, we used a list of common “eliciting events” to identify key dates for recall (e.g., holidays, birthdays, newsworthy events, doctor’s appointments, incarcerations/hospitalizations, etc.) and wrote them on a shared calendar in collaboration with the participant. Research assistants then attempted to 1) identify any known periods of abstinence, 2) elicit regular patterns of use, 3) work backward from the current date, 4) identify use on key dates, and 5) identify use on dates not yet accounted for. All estimated dollar amounts were written on a shared calendar in collaboration with the participant. Participants generally spontaneously reported their use in dollars and were also generally able to estimate dollar amounts for cocaine acquired through other means (e.g., trade, socially), but research assistants were provided with a typical amount-to-dollar conversion sheet for multiple methods of use in case participants were unable to identify a dollar amount. This TLFB method was standard across the three clinical trials.

From the TLFB, we calculated nine metrics of cocaine use: 1) frequency of use — number of use days, 2) typical dollar amount used — mean dollar amount per use day, 3) variability in dollar amount — variance in dollar amounts, 4) typical days between uses — mean days of abstinence in between days with use, 5) variability in days between uses — variance in days of abstinence in between days with use, 6) weekday use — percentage of use days occurring on a weekday, 7) change in frequency of use — linear slope of a 3-day moving average of use/no use over the month, 8) change in dollar amount used — linear slope of dollar amounts over the month (change metrics were used only in the treatment outcome analyses, because any influence of change on life functioning would be difficult to discern given the fully overlapping reporting periods), and 9) interaction between frequency and typical dollar amount used, a rough estimate of total use in the month that was important in prior research (Liu et al., 2021).

2.3.2. Life functioning

Life functioning measures were composite scores from the medical, social, and employment domains of the Addiction Severity Index (ASI; McLellan et al., 1980). We excluded the drug domain due to overlap with our metrics of use, the legal domain due to near zero variance from our exclusionary criteria, and the psychiatric domain, as diagnoses on the Structured Clinical Interview for the DSM-V (SCID; First et al., 2015) were treated as potential covariates rather than an outcome.

2.3.3. Treatment outcome

Our treatment outcome measure was the total cocaine-negative urines submitted during the initial 4 weeks of treatment with thrice weekly testing (range 0–12), as suggested by (Ling et al., 1997). Urine tests were done using the Readitest 6 Cassette Urine Drug Screen was used with testing done on site at point-of-care, by a trained researcher in accordance with Redwood Toxicology Laboratory protocol.

2.3.4. Potential covariates

We additionally evaluated basic demographics; location (Houston, Chicago); cocaine route of administration from the ASI; presence of either a mood or anxiety disorder, alcohol use disorder, and cannabis use disorder from the SCID; probable diagnosis of PTSD from the PTSD Checklist for DSM-5 (PCL: Blevins et al., 2015); and presence of moderate to severe nicotine dependence from the Fagerström Test for Nicotine Dependence (FTND; Heatherton et al., 1991).

2.4. Procedures

Participants were screened via phone or online questionnaire, with potentially eligible participants then completing an in-person screening including all aforementioned measures except the TLFB and total negative urine samples during treatment. Participants completed the full 30-day TLFB at their enrollment session, prior to starting treatment, and total negative urine samples were collected using standard urine drug screens thrice weekly across a 4-week treatment period for both trials included in the outcomes analyses.

2.5. Data analysis

All variables were mean-centered, and all metrics were standardized for regression analyses. Several metrics and outcomes were non-normal, and transformations (e.g., log) were only partially successful. Thus, the regression analyses retained the transformed variables (transformations are noted in all tables) after verifying all regression diagnostic indices were within appropriate ranges, and Spearman correlations were used for all zero-order analyses. Further, variability in typical dollar amount used, variability in days between uses, and change in dollar amount used all rely on establishing “patterns” of use, which cannot be fully captured with few use occasions. Thus, we excluded n = 32 participants reporting <4 days of use from all zero-order analyses involving these three metrics, and our regression analyses included a dummy variable coded 0: ≥4 days, 1: <4 days, which we interacted with those three metrics to ensure main effects of those variables capture only participants meeting these criteria.

We then tested metrics for multicollinearity. Typical days between uses was removed from all models due to high correlation (>0.80) with frequency (see Table 2). Preliminary multiple regression analyses did not indicate multicollinearity among the remaining metrics (all VIF < 5). For simplicity, we also used preliminary regressions to exclude the total amount interaction from models where it was insignificant. This resulted in 5 metrics — frequency, typical dollar amount used, variability in dollar amount, % weekday use, and variability in days between uses — used in all models. In addition, the ASI employment model included the interaction between frequency and typical dollar amount used, and the treatment outcome model included change in frequency and amount spent. Life functioning was predicted using ordinary least squares regression, and total negative urines was predicted using a hierarchical Poisson regression model with participants nested in their respective treatment conditions. Multiple comparisons were corrected with False Discovery Rate (FDR) adjustment for the number of metrics included in each analysis.

Table 2.

Correlations between metrics of use.

Variable M(SD) 1 2 3 4 5 6 7 8
1. Frequency of use (log) 11.4 (9.04) – −0.15 −0.14 −0.18 −0.92 −0.77 −0.23 0.08
2. Typical dollar amount (log) 53.7 (51.8) – 0.51 0.01 0.11 0.1 0.05 −0.14
3. Variance in dollar amount (log) 1197 (3627) – −0.02 −0.09 0.24 0.06 −0.09
4. % Weekday use (sqrt) 0.71 (22.8) – −0.77 0.06 −0.29 −0.06
5. Typical days between uses (log) 4.69 (4.23) – 0.78 0.11 −0.06
6. Variance in days between uses (log) 13.4 (29.1) – 0.14 −0.05
7. Frequency change 0.09 (1.14) – 0.03
8. Dollar amount change 0.01 (3.79)
Open in a new tab

Per previous recommendations on covariates in clinical trials (Assmann et al., 2000; Pocock et al., 2002), we evaluated all demographic and clinical characteristics as well as study location in Table 3 and included a variable as a covariate if it significantly related to both a metric of cocaine use and the regression outcome. In addition to these variables, the three ASI domains were evaluated as potential covariates for the treatment outcome analyses, as these capture aspects of baseline life functioning (e.g., income) that could reasonably relate to treatment outcomes. These evaluations resulted in covarying for probable PTSD in the ASI medical (p = 0.04) model and both probable PTSD (p = 0.004) and gender (p = 0.02) in the ASI social model. Because some covariates could capture variance that is of theoretical interest, we present all analyses with and without covariates. Additionally, due to theoretical reasons to believe route of administration (e.g. snorting powder cocaine vs. smoking crack cocaine) could be confounded with amount spent, we also reviewed results including only those who used cocaine via smoking crack for both life functioning (n = 181) and treatment outcome (n = 155) analyses. Random effects considered included study in the life functioning analysis and treatment condition in the treatment outcome analysis (this also captures study differences as treatment groups were uniquely coded across studies). Study was insignificant for two of three life functioning models (medical, employment) and introduced convergence issues for the third (social); to avoid unnecessary complexity, we chose not to include study. The random effect for treatment condition was significant and was included in the treatment outcome analysis.

Table 3.

Descriptive statistics: overall sample. (N = 207).

Demographics M (SD) or % (n)
Gender
Woman 19.2 % (40)
Man 80.8 % (167)
Age (years) 50.2 (7.82)
Ethnicity
Non-Hispanic 88.7 % (184)
Hispanic 11.3 % (23)
Race
African American/Black 78.4 % (162)
Asian 0.9 % (2)
White 11.3 % (23)
American Indian or Alaskan Native 0.5 % (1)
More than one race 4.9 % (9)
Unknown/not reported 4.7 % (10)
Education (years) 12.86 (1.53)
Study location
University of Texas at Houston 87.8 % (182)
University of Illinois at Chicago 12.2 % (25)
Route of Administration
 Nasal 12.2 % (25)
 Smoking 86.4 % (179)
 IV 0.9 % (2)
Comorbid substance use
Alcohol Use Disorder 33 % (71)
Nicotine Use Disorder 65 % (140)
Marijuana Use Disorder 19 % (41)
Comorbid mental health
Internalizing mood disorder 13 % (28)
Post-traumatic stress disorder 30.9 % (66)
Open in a new tab

3. Results

Descriptive statistics for the overall sample are shown in Table 2. Descriptive statistics for the subsample used in the outcome analysis were not significantly different.

3.1. Life functioning analysis results

See Table 4 for life functioning results. At the zero-order level, no metrics significantly predicted life functioning. Regression results showed that a greater typical dollar amount used (p = 0.02) significantly related to better employment functioning; probable PTSD diagnosis was associated with worse medical (p = 0.04) and social (p < 0.001) functioning. All other results were null, and results did not change when including only people who used cocaine via smoking.

Table 4.

Life functioning and cocaine use metrics zero-order correlations and regression results.

Variable Zero-order correlations Multiple regression with covariates Multiple regression without covariates
Employment Medical (sqrt) Social (sqrt) Employment Medical (sqrt) Social (sqrt) Employment Medical (sqrt) Social (sqrt)
Metrics of use
Intercept – – – β = 0.62, p < 0.001*** β = 0.20, p < 0.001*** β = 0.14, p = 0.40 β = 0.62, p < 0.001*** β = 0.23, p <0.001*** β = 0.30, p < 0.001***
Frequency of use (log) ρ = 0.03 p = 0.67 ρ = −0.02, p = 0.80 ρ = 0.01, p = 0.85 β = 0.02, p = 0.78 β = 0.09, p = 0.32 β =−0.05, p = 0.71 β = 0.02, p = 0.78 β = 0.11, p = 0.17 β = −0.02, p = 0.96
Typical dollar amount (log) ρ = −0.12, p = 0.09 ρ = 0.01, p = 0.93 ρ = 0.12, p = 0.10 β = −0.09, p = 0.02* β = −0.01, p = 0.98 β = 0.05, p = 0.46 β = −0.09, p = 0.02* β = −0.01, p = 0.94 β = 0.05, p = 0.67
Variance in dollar amount (log) ρ = 0.10 p = 0.14 ρ = −0.02, p = 0.73 ρ = 0.03, p = 0.72 β = 0.02, p = 0.06 β = −0.01, p = 0.89 β = −0.01, p = 0.71 β = 0.02, p = 0.06 β = −0.00, p = 0.94 β = −0.01, p = 0.96
Variance in days between uses (log) ρ = 0.03, p = 0.66 ρ = 0.09, p = 0.18 ρ = 0.01, p = 0.86 β = 0.02, p = 0.58 β = 0.06, p = 0.12 β = 0.01, p = 0.93 β = 0.02, p = 0.58 β = 0.06, p = 0.11 β = 0.01, p = 0.96
Weekday % (sqrt) ρ = −0.01, p = 0.93 ρ = −0.04, p = 0.58 ρ = 0.17, p = 0.81 β = 0.01, p = 0.94 β = −0.04, p = 0.89 β = −0.01, p = 0.93 β = 0.01, p = 0.94 β = −0.04, p = 0.94 β = −0.02, p = 0.96
Frequency × dollar amount Interaction – – – β = −0.07, p = 0.09 – – β= −0.07, p = 0.09 – –
Dummy ρ = −0.03, p = 0.70 ρ = −0.06, p = 0.36 ρ = −0.10, p = 0.14 β = 0.02, p = 0.83 β = 0.18, p = 0.17 β = −0.17, p = 0.11 β = 0.02, p = 0.83 β = 0.21, p = 0.12 β = −0.14, p = 0.21
Variance in dollar amount × dummy – – – β = 0.02, p = 0.28 β = −0.01, p = 0.49 β = 0.00, p = 0.86 β = 0.02, p = 0.28 β = −0.01, p = 0.53 β = 0.01, p = 0.66
Variance in days between uses × dummy – – – β = 0.02, p = 0.63 β = −0.06, p = 0.22 β = 0.00, p = 0.93 β = 0.02, p = 0.63 β = −0.05, p = 0.21 β = 0.01, p = 0.90
Selected covariates
PTSD – – – – β = 0.11, p = 0.04* β = 0.16, p < 0.001***
Gender – – – – – β = 0.09, p = 0.06
Open in a new tab

Note:

*

p < 0.05.

**

p < 0.01.

***

p < 0.001; all predictor p-values have been FDR-adjusted.

3.2. Outcomes analysis results

At the zero-order level, using less frequently, greater typical dollar amount used, and having more variability of use with respect to both dollar amount used and days between uses were significantly related to better treatment outcomes (Table 5). Results for these metrics were similar in the multiple regression analysis, apart from variability in days between uses, which was associated with worse treatment outcomes in the regression. Additional metrics significantly predicted treatment outcomes in the regression analysis, with less weekday use (p = 0.03), lower frequency of use (p < 0.001), and increases in typical dollar amount used (p = 0.04), resulting in more total negative urines being submitted. Frequency of use (β = −0.59), variability in dollar amount used (β = 0.12), and change in frequency (β = −0.23) were the strongest predictors of treatment outcomes. When including only people who used cocaine via smoking, variability in days between uses, weekday use, and change in amount spent were no longer significant (ps > 0.05), other results were substantively unchanged.

Table 5.

Treatment outcomes analysis zero-order correlations and multiple regression results.

Variable Zero-order correlations with cocaine negative urines Multiple regression
Random effects Treatment condition Fixed effects – M = 0.23, SD = 0.49
Intercept – β = 0.81, p < 0.001***
Frequency of use (log) ρ = −0.20, p < 0.001** β = −0.69, p < 0.001***
Amount spent mean (log) ρ = 0.25, p < 0.001** β = 0.27, p < 0.001***
Amount spent variance (log) ρ = 0.24, p < 0.001** β = 0.12, p < 0.001***
Variance in days between uses (log) ρ = 0.18, p = 0.02* β = - 0.13, p = 0.01*
Weekday % (sqrt) ρ = −0.003, p = 0.97 β = - 0.37, p = 0.03*
Frequency change ρ = −0.05, p = 0.62 β = −0.23, p < 0.001***
Amount spent change ρ = 0.01, p = 0.90 β = 0.03, p = 0.04*
Dummy – β = −1.17, p < 0.001***
Amount spent var × dummy – β = −0.19, p < 0.001***
Variance in days between uses × dummy – β = 0.03, p = 0.65
Amount spent change × dummy – β = −0.12, p < 0.001***
Open in a new tab

Note:

*

p < 0.05.

**

p < 0.01.

***

p < 0.001; all predictor p-values are FDR-adjusted.

4. Discussion

This study examined how aspects of cocaine use beyond frequency relate to life functioning and treatment outcomes. We found that metrics aside from frequency of use were meaningfully associated with both life functioning and treatment outcomes and that associations sometimes differed between life functioning and treatment outcomes. For example, frequency of use did not significantly relate to life functioning at baseline but was the strongest predictor of treatment outcomes. Using higher dollar amounts was associated with both better employment functioning at baseline and better treatment outcomes, despite our general expectation that more cocaine use would negatively relate to both life functioning and success in treatment. Such a pattern could represent a confound of higher income, both allowing more use at baseline and contributing to better treatment outcomes. However, the employment functioning metric itself did not significantly relate to treatment outcomes when tested as a potential covariate. This finding makes a single income-based explanation for both relationships less likely, and again suggests potentially different roles of different aspects of use in baseline functioning vs. treatment outcomes.

Examining life functioning in more detail, lower frequency of use (i. e., fewer days of use) was not associated with better life functioning in employment, medical, or social domains, which contrasts with previous work (Reelick & Wierdsma, 2006; Roos, Nich, Mun, Babuscio, et al., 2019a; Roos, Nich, Mun, Mendonca, et al., 2019). Instead, greater typical dollar amount used was associated with better employment functioning. This discrepancy from previous literature may reflect differences in the point at which use was measured — in our study, participants had not yet entered treatment, while other studies measured use and life functioning during or after treatment, when these might reflect treatment response. It is important to note that we cannot infer causality — as noted above, participants with more disposable income may have means to use larger quantities, producing the observed relationship with employment. Longitudinal studies will be necessary to disentangle these relationships. These findings also suggest that studies using dollar amount as an outcome should closely examine income as a potential covariate. It may be that the dollar amount spent relative to income serves as a better measure of impairment.

With respect to treatment effectiveness outcomes, participants who achieved better outcomes (i.e., submitted more negative urine screens) tended to enter treatment with 1) lower frequency of use, 2) greater typical dollar amount used, 3) greater variability in dollar amount used, 4) lower weekday use, 5) more variability in days between uses, 6) a downwards trajectory in frequency of use over the month leading into treatment, and 7) an upward trajectory in the typical dollar amount used. Together, these results suggest more positive treatment outcomes for “binge” users — those who engage in irregular, infrequent, high-amount use primarily on weekends or in social settings. These patterns also suggest that behavior change prior to treatment is important, possibly constituting a proximal indicator of motivation to make changes. The strongest predictor of treatment outcomes was frequency of use, which aligned with expectations (Ahmadi et al., 2009; Palazón-Llecha et al., 2024); however, variance in the dollar amount used and rate of change in frequency were also relatively good predictors. Further, our finding that greater typical dollar amount used and greater variance in dollar amount used predict positive treatment outcomes is novel. Again, we posit this may be reflective of social, binge, or non-habitual use (Green et al., 1994). Supporting this interpretation, and consistent with Huang et al. (2014), we saw weekend (vs. weekday) use was also associated with better treatment outcomes. Our finding that reducing frequency in the month before treatment predicted better outcomes contrasts with Ye and colleagues, who found that prior month reductions in use predicted poor treatment outcomes, which they suggested indicated prematurely attempting abstinence (Ye et al., 2015b). Additional replication in larger samples will be required to address this discrepancy, but our findings are consistent with studies showing reductions in frequency generally have clinical benefits (Amin-Esmaeili et al., 2024; Roos, Nich, Mun, Babuscio, et al., 2019b). Finally, our finding that increases in typical dollar amount used over that same time related to better outcomes are also novel and slightly counter-intuitive but again may reflect a shift to more “binge-like” behavior as a positive sign for treatment.

Our study had several limitations. First, our data was observational, and we cannot determine causality. We were also unable to replicate the post-treatment life functioning results used to validate frequency, as our measure of life functioning (ASI) was not re-administered after treatment. Third, because frequency of use was also our treatment outcome, the significance of frequency-related factors is somewhat tautological. We also operationalized quantity as dollar amount, an imperfect measure of amount used (Carroll et al., 2014), as crack and powder cocaine have different costs and purities, and individuals had to estimate the value of cocaine acquired through other means (e.g., exchanging sex for drugs, shared use). Our results, particularly for amount of use, were substantively similar when we excluded people using cocaine intranasally and intravenously. However, as only a small number of participants reported a means of use other than smoking, we could not conduct more sophisticated analyses of any effects of route of administration/type of cocaine used. Our sample size was also smaller with somewhat restricted demographics (primarily Black and male). Although this is typical for clinical trials in CUD, it is not representative of the general population of people who use cocaine (Cano et al., 2020). Finally, this analysis was exploratory, and all results should be considered preliminary. Future investigations involving larger and more diverse samples are warranted to confirm our findings.

In summary, frequency of use has historically been the most common metric used to describe “pattern of cocaine use,” but in this study, it did not show a particularly strong relationship with life functioning in the social, employment, and medical domains compared to other aspects of use. Further, although frequency was the strongest predictor of treatment outcomes, other aspects of use were also meaningfully predictive, suggesting that the field may benefit from exploring whether other metrics of cocaine use (e.g. amount used, weekday vs. weekend us, variability/predictability) may be valid targets of harm-reduction focused treatment (Bagot & Kaminer, 2018; Marlatt & Witkiewitz, 2002; Roos, Nich, Mun, Babuscio, et al., 2019b). Notably, all aspects of cocaine use were better at predicting future cocaine use than explaining life functioning, which aligns with calls to consider life functioning separately from drug use outcomes in research and clinical practice (Hagman et al., 2022; Laudet, 2011; Witkiewitz et al., 2019). Altogether, these findings suggest that a more nuanced characterization of cocaine use, using various metrics, may help us develop cocaine use profiles of greater meaning. Although this was a preliminary study, we hope it will encourage further collection and examination of these other aspects of use in clinical trials, allowing for more empirical insight into what constitutes a problematic “pattern of cocaine use.”

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

This work is funded by grants from the National Institutes of Health for Heather Webber [K01DA058765], Joy Schmitz [R01DA048026, R01DA039125], and Margaret Wardle [K08DA040006].

Footnotes

CRediT authorship contribution statement

Samantha Ellis: Writing – original draft, Conceptualization, Formal analysis. Rachel Sun: Conceptualization, Writing – original draft, Formal analysis. Angela Heads: Writing – review & editing, Writing – original draft. Jin Ho Yoon: Writing – original draft, Writing – review & editing. Heather Webber: Writing – review & editing, Writing – original draft. Joy Schmitz: Writing – original draft, Writing – review & editing, Supervision. Margaret Wardle: Writing – review & editing, Supervision.

References

  1. Ahmadi J, Kampman KM, Oslin DM, Pettinati HM, Dackis C, & Sparkman T (2009). Predictors of treatment outcome in outpatient cocaine and alcohol dependence treatment. American Journal on Addictions, 18(1), 81–86. 10.1080/10550490802545174 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Amin-Esmaeili M, Farokhnia M, Susukida R, Leggio L, Johnson RM, Crum RM, & Mojtabai R (2024). Reduced drug use as an alternative valid outcome in individuals with stimulant use disorders: Findings from 13 multisite randomized clinical trials. Addiction (Abingdon, England), 119(5), 833–843. 10.1111/add.16409 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Assmann SF, Pocock SJ, Enos LE, & Kasten LE (2000). Subgroup analysis and other (mis)uses of baseline data in clinical trials. The Lancet, 355(9209), 1064–1069. 10.1016/S0140-6736(00)02039-0 [DOI] [PubMed] [Google Scholar]
  4. Bagot KS, & Kaminer Y (2018). Harm reduction for youth in treatment for substance use disorders: One size does not fit all. Current Addiction Reports, 5(3), 379–385. 10.1007/s40429-018-0217-8 [DOI] [Google Scholar]
  5. Bernstein J, Cheng DM, Wang N, Trilla C, Samet J, & Saitz R (2015). Recreational drug use among primary care patients: Implications of a positive self-report. The Annals of Family Medicine, 13(3), 257–260. 10.1370/afm.1750 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Blevins CA, Weathers FW, Davis MT, Witte TK, & Domino JL (2015). The posttraumatic stress disorder checklist for DSM-5 (PCL-5): Development and initial psychometric evaluation. Journal of Traumatic Stress, 28(6), 489–498. 10.1002/jts.22059 [DOI] [PubMed] [Google Scholar]
  7. Cano M, Oh S, Salas-Wright CP, & Vaughn MG (2020). Cocaine use and overdose mortality in the United States: Evidence from two national data sources, 2002–2018. Drug and Alcohol Dependence, 214, Article 108148. 10.1016/j.drugalcdep.2020.108148 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Carroll KM, Kiluk BD, Nich C, DeVito EE, Decker S, LaPaglia D, … Ball SA (2014). Toward empirical identification of a clinically meaningful indicator of treatment outcome: Features of candidate indicators and evaluation of sensitivity to treatment effects and relationship to one year follow up cocaine use outcomes. Drug and Alcohol Dependence, 137, 3–19. 10.1016/j.drugalcdep.2014.01.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Falk D, Wang XQ, Liu L, Fertig J, Mattson M, Ryan M, … Litten RZ (2010). Percentage of subjects with no heavy drinking days: Evaluation as an efficacy endpoint for AlcoholClinical trials. Alcoholism: Clinical and Experimental Research, 34 (12), 2022–2034. 10.1111/j.1530-0277.2010.01290.x [DOI] [PubMed] [Google Scholar]
  10. FDA, C. for D. E. (2023, October 5). Stimulant use disorders: Developing drugs for treatment. FDA. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/stimulant-use-disorders-developing-drugs-treatment. [Google Scholar]
  11. First MB, Williams JBW, Karg RS, & Spitzer RL (2015). Structured clinical interview for DSM-5, research version. American Psychiatric Association. [Google Scholar]
  12. Green A, Pickering H, Foster R, Power R, & Stimson GV (1994). Who uses cocaine? Social profiles of cocaine users. Addiction Research, 2(2), 141–154. 10.3109/16066359409109139 [DOI] [Google Scholar]
  13. Hagman BT, Falk D, Litten R, & Koob GF (2022). Defining recovery from alcohol use disorder: Development of an NIAAA research definition. American Journal of Psychiatry, 179(11), 807–813. 10.1176/appi.ajp.21090963 [DOI] [PubMed] [Google Scholar]
  14. Heatherton TF, Kozlowski LT, Frecker RC, & Fagerström K-O (1991). The Fagerström test for nicotine dependence: A revision of the Fagerström Tolerance Questionnaire. British Journal of Addiction, 86(9), 1119–1127. 10.1111/j.1360-0443.1991.tb01879.x [DOI] [PubMed] [Google Scholar]
  15. Huang H, Li Y, & Guan Y (2014). Joint modeling and clustering paired generalized longitudinal trajectories with application to cocaine abuse treatment data. Journal of the American Statistical Association, 109(508), 1412–1424. 10.1080/01621459.2014.957286 [DOI] [Google Scholar]
  16. Kellogg SH, McHugh PF, Bell K, Schluger JH, Schluger RP, LaForge KS, … Kreek MJ (2003). The Kreek-McHugh-Schluger-Kellogg scale: A new, rapid method for quantifying substance abuse and its possible applications. Drug and Alcohol Dependence, 69(2), 137–150. 10.1016/s0376-8716(02)00308-3 [DOI] [PubMed] [Google Scholar]
  17. Kiluk BD, Carroll KM, Duhig A, Falk DE, Kampman K, Lai S, … Strain EC (2016). Measures of outcome for stimulant trials: ACTTION recommendations and research agenda. Drug and Alcohol Dependence, 158, 1–7. 10.1016/j.drugalcdep.2015.11.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Lau-Barraco C, Braitman AL, Linden-Carmichael AN, & Stamates AL (2016). Differences in weekday versus weekend drinking among nonstudent emerging adults. Experimental and Clinical Psychopharmacology, 24(2), 100–109. 10.1037/pha0000068 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Laudet AB (2011). The case for considering quality of life in addiction research and clinical practice. Addiction Science & Clinical Practice, 6(1), 44–55. [PMC free article] [PubMed] [Google Scholar]
  20. Ling W, Shoptaw S, Wesson D, Rawson RA, Compton M, & Klett CJ (1997). Treatment effectiveness score as an outcome measure in clinical trials. NIDA Research Monograph, 175, 208–220. [PubMed] [Google Scholar]
  21. Liu Y, Cheong J, Vaddiparti K, & Cottler LB (2020). The association between quantity, frequency and duration of cocaine use during the heaviest use period and DSM-5 cocaine use disorder. Drug and Alcohol Dependence, 213, Article 108114. 10.1016/j.drugalcdep.2020.108114 [DOI] [PubMed] [Google Scholar]
  22. Liu Y, Vaddiparti K, Cheong J, & Cottler LB (2021). Identification of typologies of cocaine use based on quantity, frequency, and duration of use: A latent profile analysis. Journal of Addiction Medicine, 15(3), 211. 10.1097/ADM.0000000000000749 [DOI] [PubMed] [Google Scholar]
  23. Marini C, Northover NS, Gold ND, Rogers UK, O’Donnell KC, Tofighi B, … Bogenschutz MP (2023). A systematic approach to standardizing drinking outcomes from timeline followback data. Substance Abuse: Research and Treatment, 17. 10.1177/11782218231157558, 11782218231157558. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Marlatt GA, & Witkiewitz K (2002). Harm reduction approaches to alcohol use: Health promotion, prevention, and treatment. Addictive Behaviors, 27(6), 867–886. 10.1016/S0306-4603(02)00294-0 [DOI] [PubMed] [Google Scholar]
  25. McKay JR, Alterman AI, Koppenhaver JM, Mulvaney FD, Bovasso GB, & Ward K (2001). Continuous, categorical, and time to event cocaine use outcome variables: Degree of intercorrelation and sensitivity to treatment group differences. Drug and Alcohol Dependence, 62(1), 19–30. 10.1016/S0376-8716(00)00156-3 [DOI] [PubMed] [Google Scholar]
  26. McLellan AT, Luborsky L, Woody GE, & O’Brien CP (1980). An improved diagnostic evaluation instrument for substance abuse patients. The addiction severity index. The Journal of Nervous and Mental Disease, 168(1), 26–33. 10.1097/00005053-198001000-00006 [DOI] [PubMed] [Google Scholar]
  27. Palazón-Llecha A, Caparrós B, Trujols J, Duran-Sindreu S, Batlle F, Madre M, & Mallorquí-Bagúe N (2024). Predictors of cocaine use disorder treatment outcomes: A systematic review. Systematic Reviews, 13(1), 124. 10.1186/s13643-024-02550-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Pocock SJ, Assmann SE, Enos LE, & Kasten LE (2002). Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: Current practice and problems. Statistics in Medicine, 21(19), 2917–2930. [DOI] [PubMed] [Google Scholar]
  29. Reelick NF, & Wierdsma AI (2006). The addiction severity index as a predictor of the use of mental health care. Psychology of Addictive Behaviors, 20(2), 214–218. 10.1037/0893-164X.20.2.214 [DOI] [PubMed] [Google Scholar]
  30. Robinson SM, Sobell LC, Sobell MB, & Leo GI (2014). Reliability of the timeline Followback for cocaine, cannabis, and cigarette use. Psychology of Addictive Behaviors, 28(1), 154–162. 10.1037/a0030992 [DOI] [PubMed] [Google Scholar]
  31. Roos CR, Nich C, Mun CJ, Mendonca J, Babuscio TA, Witkiewitz K, … Kiluk BD (2019). Patterns of cocaine use during treatment: Associations with baseline characteristics and follow-up functioning. Journal of Studies on Alcohol and Drugs, 80(4), 431–440. 10.15288/jsad.2019.80.431 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Roos CR, Nich C, Mun CJ, Babuscio TA, Mendonca J, Miguel AQC, … Kiluk BD (2019a). Clinical validation of reduction in cocaine frequency level as an endpoint in clinical trials for cocaine use disorder. Drug and Alcohol Dependence, 205, Article 107648. 10.1016/j.drugalcdep.2019.107648 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Roos CR, Nich C, Mun CJ, Babuscio TA, Mendonca J, Miguel AQC, … Kiluk BD (2019b). Clinical validation of reduction in cocaine frequency level as an endpoint in clinical trials for cocaine use disorder. Drug and Alcohol Dependence, 205, Article 107648. 10.1016/j.drugalcdep.2019.107648 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. SAMHSA. (2018). Results from the 2017 National Survey on Drug Use and Health (NSDUH): Key substance use and mental health indicators in the United States | SAMHSA Publications and Digital Products. https://store.samhsa.gov/product/results-2017-national-survey-drug-use-and-health-nsduh-key-substance-use-and-mental-health.
  35. Schmitz JM, Lane S, Weaver M, Narayana PA, Hasan KM, Russell DD, … Green C (2021). Targeting white matter neuroprotection as a relapse prevention strategy for treatment of cocaine use disorder: Design of a mechanism-focused randomized clinical trial. Contemporary Clinical Trials, 111, Article 106603. 10.1016/j.cct.2021.106603 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Schmitz JM, Stotts AL, Vujanovic AA, Yoon JH, Webber HE, Lane SD, … Green CE (2024). Contingency management plus acceptance and commitment therapy for initial cocaine abstinence: Results of a Sequential Multiple Assignment Randomized Trial (SMART). Drug and Alcohol Dependence, 111078. 10.1016/j.drugalcdep.2023.111078 [DOI] [PubMed] [Google Scholar]
  37. Sobell LC, & Sobell MB (1992). Timeline follow-back. In Litten RZ, & Allen JP (Eds.), Measuring alcohol consumption: Psychosocial and biochemical methods (pp. 41–72). Humana Press. 10.1007/978-1-4612-0357-5_3. [DOI] [Google Scholar]
  38. Wardle MC, Hoots JK, Miloslavich K, Nunez C, de Dios C, Holden C, … Schmitz JM (2023). Deficits in consummatory reward relate to severity of cocaine use. Drug and Alcohol Dependence, 109950. 10.1016/j.drugalcdep.2023.109950 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Witkiewitz K, Wilson AD, Pearson MR, Montes KS, Kirouac M, Roos CR, … Maisto SA (2019). Profiles of recovery from alcohol use disorder at three years following treatment: Can the definition of recovery be extended to include high functioning heavy drinkers? Addiction, 114(1), 69–80. 10.1111/add.14403 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ye J, Li Y, & Guan Y (2015a). Joint modeling of longitudinal drug using pattern and time to first relapse in cocaine dependence treatment data. The Annals of Applied Statistics, 9(3). 10.1214/15-AOAS852 [DOI] [Google Scholar]
  41. Ye J, Li Y, & Guan Y (2015b). Joint modeling of longitudinal drug using pattern and time to first relapse in cocaine dependence treatment data. The Annals of Applied Statistics, 9(3). 10.1214/15-AOAS852 [DOI] [Google Scholar]
  42. Yoon JH, Suchting R, McKay SA, San Miguel GG, Vujanovic AA, Stotts AL, … Schmitz JM (2020). Baseline cocaine demand predicts contingency management treatment outcomes for cocaine-use disorder. Psychology of Addictive Behaviors, 34, 164–174. 10.1037/adb0000475 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Yoon JH, Suchting R, de Dios C, Vincent JN, McKay SA, Lane SD, & Schmitz JM (2021). Decreased cocaine demand following contingency management treatment. Drug and Alcohol Dependence, 226(June), Article 108883. 10.1016/j.drugalcdep.2021.108883 [DOI] [PubMed] [Google Scholar]

RESOURCES