Secondary Analysis of Agreement between Negative Timeline Follow Back Report and Negative Urine Toxicology in a Large Trial of Individuals with Opioid Use Disorder

Abstract

Objectives:

Timeline follow back is a self-report measure commonly used as a method of assessing historical drug use in both clinical and research settings. Our study considered rates of agreement between Timeline Follow Back (TLFB) and an objective biological assay of opioid use.

Methods:

We calculated the rates of agreement between negative report of opioid use for the most recent 8 days on timeline follow back and urine toxicology results in a large multisite opioid use disorder treatment trial.

Results:

In total, 3,986 assessments were provided by trial participants with both urine toxicology and TLFB during weeks 1–12, 2,716 during weeks 13–24, and 325 at week 28. Rates of disagreement between negative TLFB and positive opioid urine toxicology were 2.33% of all assessments (21.68% of those with positive urine toxicology) over weeks 1–12, 2.06% of all assessment (25.00% of those with positive urine toxicology) over weeks 13–24, and 9.85% of all assessments (26.02% of those with positive urine toxicology) at week 28.

Conclusions:

Negative timeline follow back appears to be generally associated with negative results on urine toxicology.

Introduction:

Timeline Follow-Back (TLFB) is a self-report measure of substance use that is frequently employed in research trials. It is administered by a trained interviewer who records substance use for each day of a given follow up period, using important events in the respondent’s life as landmarks in helping with recall. This method was developed for gathering estimates of drinking days¹ but has been used to assess all substance use and is considered a valid and psychometrically sound instrument². A 2012 review³ identified two studies considering outcomes in the opioid use disorder population and found a 94% average rate of agreement between TLFB and urine toxicology (UTOX). However, these studies were relatively small (total n=208) and neither included individuals maintained on medication for opioid use disorder (OUD)^4,5. Determining agreement rates between TLFB and biological measures of opioid use is important in interpreting data from any clinical study. We set out to quantify the rate of agreement between negative timeline follow back self-report and the biological measure of UTOX in a large multisite trial of opioid use disorder.

Methods:

We performed a secondary analysis of the X:BOT trial (CTN-0051) which randomized patients seeking treatment for OUD at eight short term inpatient/residential detoxification units to either long-acting injectable naltrexone or buprenorphine/naloxone. Participants were followed on an outpatient basis over a 24-week period. A full description of the study design and rationale has been published previously^6,7. The study was approved by the Institutional Review Boards of New York University Medical Center and each of the participating sites, and all patient-participants provided written informed consent.

TLFB was collected at each site by a research assistant who was trained in the administration of TLFB and instructed to gather the information in a non-judgmental fashion. The study utilized a point of care instant UTOX testing that was confirmed with gas chromatography mass spectroscopy. Samples were not produced under observation but were tested for temperature and adulteration.

Statistical Analysis:

In this analysis we included TLFB assessments and UTOX collected during the 24-week trial and a week 28 follow-up time point. We defined non-agreement for a given visit as no opioid use reported on the day of assessment or on any of the 7 previous days on TLFB but positive UTOX for opioids. The day of assessment was captured on the TLFB collected at the subsequent visit. TLFB In cases where two UTOX assessments occurred within <1 week of each other, TLFB reporting was only included for the days following the prior UTOX assessment. Visits missing TLFB reporting for any of the 7 previous days or the day of assessment were not included in the analysis. We did not consider positive report of opioid use on TLFB and negative UTOX to be non-agreement as the opioids may have been cleared before testing and synthetic opioids may not have been detected by the testing. Opioid use was defined by at least one day of reported use of the following substances: opioid analgesics, methadone, heroin, buprenorphine (if randomized to naltrexone or if not reporting treatment with buprenorphine). For purposes of this analysis positive opioids on UTOX was defined as a positive test for the following: opiates 2000 ng, oxycodone, methadone, or buprenorphine (if randomized to naltrexone or if not reporting treatment with buprenorphine at week 28 follow-up).

We calculated descriptive statistics on the rates of disagreement between negative timeline follow back and positive UTOX in samples provided in the first 12 and last 12 weeks of the trial as well as the week 28 follow up time point after study completion.

Results:

570 individuals were randomly assigned to receive either XR-NTX (n=283) or BUP-NX (n=287). Three thousand nine hundred eighty-six assessments included both UTOX and timeline follow back data during weeks 1–12 (93 missing), 2,716 assessments were available during weeks 13–24 (69 missing), and 325 were available at week 28 (35 missing). During weeks 1–12, there were 429 (10.76%) occasions when a sample was positive for opioids on UTOX and of these 93 (2.33% of the total sample and 21.68% of those with positive UTOX) of these reported no opioid use on TLFB (Table 1). During weeks 13–24, there were 224 (8.25%) occasions when individuals had positive opioids on UTOX and of these 56 were not in agreement with negative report on TLFB (2.06% of the total sample and 25.00% of those with positive UTOX) (Table 1). At Week 28, there were 123 (37.85%) occasions when individuals had positive opioids on UTOX and of these 32 (9.85% of the total sample 26.02% of those with positive UTOX) reported no opioid use on TLFB (Table 1).

Table 1:

Non-agreement was defined as an encounter with a negative TLFB and positive urine toxicology.

		Timeline Follow Back (TLFB)
Time Period	Urine Toxicology Result	No Use	Use	% Positive Urine Toxicology with TLFB Non-Agreement	% of all assessments with TLFB Non-Agreement
Weeks 1–12	Negative	3384	173
	Positive	93	336	21.68%	2.33%
Weeks 13–24	Negative	2416	76
	Positive	56	168	25.00%	2.06%
Week 28	Negative	185	17
	Positive	32	91	26.02%	9.85%

Discussion:

We calculated rates of non-agreement between timeline follow-back assessment and UTOX on opioid use in a large multi-site trial. At all timepoints, around 20–25% of assessment visits with a positive UTOX had negative self-reports. These numbers are generally consistent with agreement rates across substances in previous studies³. This has been interpreted to indicate that when, as in typical research studies, self-reported substance use is ascertained in a non-judgmental fashion, and positive results are not punished (no policy of rescinding privileges or being discharged from care), self-reports can be relied on as fairly accurate. In clinical settings, if substance use has negative consequences, more patients may be likely to deny use. The 20–25% of encounters with non-agreement may have been due to shame, stigma, misremembering the time of use, or inadvertent use due to contamination in other drugs.

These findings are reassuring that patient report of drug use is generally reliable and may be relevant to larger questions surrounding the utility of UTOX in clinical practice and as a primary research outcome⁸.

This trial had several limitations. The study population, although diverse and from multiple treatment sites, only included those who consented to randomization to either buprenorphine and naltrexone and remained in the trial for follow up visits. Consistent with the effectiveness of these medications, rates of opioid use overall appear to have been low. The numbers of disagreements would likely be higher in a sample with more active drug use (e.g. more positive urines), and proportions of disagreements might differ as well. We only considered non-agreement regarding one week of TLFB due to the limited time UTOX would remain positive after opioid use. This data is therefore not applicable to longer TLFB, although TLFB has been considered accurate up to six months. Finally, individuals were aware they would be tested with more objective toxicology assessment, possibly decreasing their likelihood to falsely report no opioid use. Extrapolation to cases in which no objective assessments are collected is therefore impossible. It would also be important to understand demographic and clinical factors associated with non-agreement although this was beyond the scope of this analysis.

Conclusion:

In agreement with previous reports, these findings suggest that report of abstinence on TLFB is generally consistent with negative UTOX results. For research studies, this suggests that TLFB can be relied upon as a measure of drug use, although testing of urine may be important to discourage denial of use. Clinicians need to maintain some index of suspicion as the data suggest some proportion of patients who are actually using will deny it.