Why Did Weekly Sweat Patches Miss Opiates More Often Than Cocaine? A 63-Patient Urine Comparison

Answer first: In a 2007 outpatient study of 63 people undergoing treatment for opioid dependence, a week-long sweat patch was positive in just 27 of 81 urine-positive opiate comparisons (33.3%), but in 33 of 36 urine-positive cocaine comparisons (91.7%). Collection failure was equally important: only 188 of 536 applied patches (35.1%) were recovered intact at the prescribed seven-day interval and eligible for the principal paired analyses. A patch finding therefore had to be interpreted together with drug class, collection adherence, reference-specimen timing and the substantial missingness—not as a universal replacement for urine testing or a personal impairment test.

The original empirical study is Chawarski, Fiellin, O’Connor, Bernard and Schottenfeld, Utility of sweat patch testing for drug use monitoring in outpatient treatment for opiate dependence, Journal of Substance Abuse Treatment 33 (2007): 411–415. Read the original primary paper (DOI 10.1016/j.jsat.2007.03.004). The counts reported below concern repeated collection events from the same adults. They do not constitute 536 independently recruited study subjects or 536 verified episodes of drug use.

What question did the original investigators test?

The investigators asked whether a nonocclusive adhesive sweat patch, worn for approximately one week, could provide reliable drug-use monitoring for patients in office-based opioid treatment. They compared patch classifications with weekly urine tests and, when available, patients’ weekly reports of illicit opiate or cocaine use. This is primarily a question about collection feasibility and agreement between imperfect reference methods. The team did not randomize participants to take illicit drugs, establish an administered cocaine or heroin dose, or use blood concentrations as a gold standard. Although participants originated from a randomized buprenorphine-treatment trial, this specimen comparison should not be interpreted as a randomized trial of patch-versus-urine analytical accuracy. The valuable question is narrower: among patches that stayed attached and were brought back on time, how often did each patch classification agree with contemporaneous urine and reported behavior?

Who were the 63 independent participants?

The study followed 63 treatment-seeking adults with documented opioid dependence. There were 48 men and 15 women, with mean age 39.1 years and standard deviation 7.7 years. Entry criteria and exclusions were those of the underlying buprenorphine clinical trial; for example, certain severe psychiatric and unstable medical conditions were excluded. This limits straightforward generalization to forensic populations, incarcerated people or everyone receiving outpatient treatment. The 63 individuals contributed unequal numbers of repeated wear periods. The original report gives a mean 8.5 applied patches per participant and a range from one to 26. Those observations cluster within individuals and treatment circumstances. Describing each repeated weekly patch as another independent patient would artificially enlarge the effective sample and overstate certainty in a result that already has substantial collection attrition.

What exactly was placed on skin?

A PharmChek-style nonocclusive collection device was applied to skin and intended to remain attached for seven days. The film was designed to allow certain gases and water vapor to escape while retaining material available for laboratory drug testing. Study personnel applied and removed patches according to the manufacturer’s instructions, inspected returned devices for signs of tampering or loss of adhesion, and asked about deviations. What reached the laboratory was material accumulated in a wearable collection patch, not a known milliliter of freshly secreted sweat collected as a liquid specimen. A patch result is consequently not interchangeable with a sweat concentration in ng/mL, a urine concentration in ng/mL, or a blood concentration. The original publication does not give an independently validated concentration-to-dose conversion for that matrix.

Why is the 536-to-188 collection pathway fundamental?

Researchers applied 536 devices and scheduled each for removal after about one week. Only 291 came back still attached and were removed by research staff, a return-and-attachment fraction of 54.3%. Another 47 were returned detached; 198 were not returned at all. Within the 291 intact returns, 188 had been worn for the intended seven days, while 85 were obtained within two days of the scheduled removal and 18 were removed two to fourteen days after application. Thus the analysis depended on a relatively small selected portion of applied devices. The paper does not prove why each device was lost or mis-timed, nor can one safely treat missing patches as uniformly negative. Collection success is part of method performance, not an administrative nuisance separable from analytical sensitivity.

Original evidence table 1: follow the collection denominator

Does 35.1% describe analytical sensitivity?

No. The 35.1% figure (188/536) describes the yield of eligible devices under the collection protocol, not how frequently a correctly worn patch identifies an opiate molecule. Analytical or clinical sensitivity uses a reference-defined positive group, whereas collection yield includes losses unrelated to chemical detection. It would be misleading to multiply a patch’s detection percentage and return percentage without specifying which population, dependencies and failure mechanisms are modeled. Likewise, the 54.3% returned-attached figure includes patches not removed on the exact prescribed day. For a monitoring program, both operational feasibility and conditional test agreement matter. A method that performs relatively well on returned devices may still leave many weeks with no usable result. Every statistic should carry its proper numerator and denominator.

What was actually measured in urine?

Personnel collected urine at least weekly and submitted specimens for commercial laboratory testing. The original methods describe radioimmunoassay screening for opiate and cocaine metabolites. In all, the study reports 575 urine specimens across the observation period, but only 184 urine results were matched with the relevant on-schedule patch specimens for the principal two-by-two comparisons. Urine was therefore a different physical collection and interval from the accumulated patch. The authors did not publish individual quantitative urine concentrations, chromatograms, a contemporary confirmation package for every paired result, or a universal limit of quantification for these urine endpoints. A screening-positive urine result is an observed laboratory classification, not a directly measured level of the corresponding analyte in blood, brain, or sweat.

What did the sweat laboratory classify?

All collected patches were sent for drug testing performed by the product manufacturer or its laboratory. The study describes classifications for opiates and cocaine-related material, with opiate monitoring concerning heroin-related compounds, morphine and codeine, and cocaine monitoring concerning cocaine or metabolites. It refers readers to prior work for procedural details. The five-page report does not furnish analyte-by-analyte instrumental identification criteria, a new validation series, numerical recovery by patch, laboratory blanks or an individual cutoff table sufficient to reconstruct each assay’s analytical performance. It would therefore be improper to label this study itself a GC–MS method-validation experiment or to insert a detection limit borrowed from another laboratory. The research tested how reported patch results behaved in this specific outpatient workflow.

Why were opiate comparisons so discordant?

Of the 81 matched urine samples positive for opiates, 27 had a corresponding opiate-positive patch and 54 had an opiate-negative patch. These 54 discordant events should be described plainly: the selected patch did not flag an opiate at that assessment despite the positive urine. Four additional patches were positive when the matched urine was negative, among 103 urine-negative comparisons. Differences in specimen collection intervals, excretion biology, drug identity, patch adherence and test cutoffs are possible considerations; this experiment did not adjudicate which explanation applied in each of those 58 discordant observations. Most importantly, one cannot attribute a negative patch to absence of illicit use when a contemporaneous urine result independently indicates exposure under its own conditions.

Why did cocaine comparisons look different?

Of 36 cocaine-positive urine comparisons, 33 corresponding patches were classified cocaine-positive and three negative. Conversely, among 148 urine-negative comparisons, 13 sweat patches were positive. The higher proportion of agreement among urine-positive cocaine tests was an observed feature of the valid-wear subset, not proof that sweat patches are intrinsically reliable for every drug, dose, collection site or outpatient setting. These findings are also not independent proof of the exact date of cocaine intake: patch material accumulates during the wear period, whereas a urine sample captures excretion near a collection time, with analyte-specific detection limits. Some positives in the patch but not urine can be consistent with different detection windows; environmental contamination and other explanations require their own evidence.

Original evidence table 2: compare like reference denominators

What does positive agreement with urine actually mean?

For the paired urine comparisons, a urine-positive patch-positive fraction is the proportion of urine-positive observations in which the patch also tested positive. It can be calculated directly from the two-by-two tables, but calling it an intrinsic biological sensitivity assumes the urine comparison is a sufficiently accurate standard for actual use over precisely the same relevant interval. That assumption is problematic because the two specimens have different collection windows, and urine screening may itself miss use. Opiate 27/81 and cocaine 33/36 have conspicuously different descriptive values. Yet the publication does not report a formal paired hypothesis test demonstrating that the difference is statistically significant or attributable to a specific chemical mechanism. No precise comparative p value should be supplied retrospectively as though reported in the source.

Why did the authors use patient self-reports too?

The team also asked participants about their recent drug use. Unlike an instrument reading, a self-report depends on memory, comprehension and willingness to disclose. But the reports help ask whether a device’s laboratory classifications coincide with the behavior participants said occurred over the wear interval. Within matched self-reports, patch positives occurred in 19 of 59 admissions of recent illicit opiate use (32.2%) and 21 of 23 admissions of recent cocaine use (91.3%). Those comparisons used different valid record counts: 109 total opiate-report pairs versus 99 cocaine-report pairs. Neither denominator equals the 184 paired urine results or the 188 valid patches. A coherent interpretation preserves those subsets instead of combining them into one fictitious comprehensive matched cohort.

What was the documented-use index, and what is its weakness?

The investigators created an additional reference by restricting analysis to instances where a urine test and self-report agreed: both positive for use or both negative. Against that constructed index, estimated patch sensitivity was 36.5% for opiates (19/52 reference-positive events) and 95.0% for cocaine (19/20). Estimated specificity was 95.0% for opiates (38/40 reference-negative events) and 92.6% for cocaine (63/68). These numbers follow from selected 92-event and 88-event subsets, respectively. This is not an independent gold standard; agreement between an imperfect screening specimen and self-report can exclude difficult cases and change apparent accuracy. The two test systems may also have correlated errors or mismatched windows. Report the index’s construction whenever quoting its attractive-looking cocaine figure.

Do positive patch results identify the exact drug and route?

Only to the extent supported by the specific laboratory’s identification procedure and analyte scope. A class-level opiate screen cannot always distinguish heroin intake from every alternative without compound-specific and metabolite-specific confirmation, timing and matrix context. Morphine, codeine and 6-monoacetylmorphine are not interchangeable names or interpretations. Similarly, cocaine and benzoylecgonine are distinct analytes, and detecting one in a patch is not the same as determining administered dose or whether the drug was smoked, injected, inhaled or otherwise used. The present study was designed to compare qualitative monitoring results, not to validate an individual’s route of administration. No participant was experimentally administered an illicit opiate or cocaine dose as part of the patch comparison.

Could a patch be negative despite documented exposure?

Yes, and that is the practical reason the 54 opiate urine-positive/patch-negative observations matter. They occurred among patches retained for the principal comparison, not merely among lost devices. A patch may fail to reach a laboratory’s detection threshold because of analyte-specific skin transport, variable perspiration and sebum, wear conditions or other factors. Some of those mechanisms are biologically plausible, but the article did not quantify each participant’s sweat production or independently resolve the cause of individual discordances. No dose-dependent false-negative probability curve or validated correction for patch position was established. A single negative patch, especially from a method with limited opiate agreement in this setting, should not be described as proof that an observed person abstained throughout the wear interval.

Could a patch be positive when matched urine is negative?

Also yes. In these valid matched observations the pattern occurred four times for opiates and thirteen times for cocaine. A longer cumulative patch interval may include exposure no longer detectable by urine on the collection day, but that interpretation is not guaranteed. Drug transfer from skin or surroundings, patch handling, laboratory procedure and differences in targeting parent compounds versus metabolites can contribute uncertainty. The original study expressly discusses environmental contamination as a concern in the field; it did not reconstruct contamination or ingestion for every positive result. A patch-urine mismatch is a reason to inspect raw collection times, placement, integrity, analytes and laboratory documentation, rather than to declare one specimen fraudulent or use a single positive as a chronometer.

Why does uneven follow-up create selection bias?

Some participants contributed just one device while others contributed as many as 26, and the missing 245 devices included both detached and never-returned patches. If weeks involving greater drug use were more likely to produce lost patches, the analyzed set could underestimate exposure; if missingness followed another pattern, the direction might differ. The authors suggested a plausible bias but did not experimentally establish the behavior of missing observations. Such informative missingness can also distort apparent sensitivity because only a selected 188 valid patches entered the main comparison. Simply excluding missing results and then claiming complete cohort-level monitoring performance would obscure the 65% of applications that failed to provide an on-schedule acceptable patch for analysis.

What do the findings not establish statistically?

The study reported observed contingency counts, percentages and constructed-index estimates, not a definitive multivariable proof of why detection differed by drug class or a universally transferable confidence interval for each forensic setting. The descriptive fractions are genuine observations, but the paper supplies no formally tested null result proving equal detection performance across all specimen collection methods. Nor was there a confirmed pharmacokinetic analysis that could map sweat positives to contemporaneous blood drug concentrations. Absence of a reported p value should remain absence of a reported p value; one should not substitute invented significance language. The authors’ cautious conclusion was that sweat patches should not yet replace urine monitoring in clinical and research practice under the tested conditions.

Which biological compartments were not examined?

There was no quantitative parent cocaine, heroin, morphine or metabolite measurement in a legally collected human venous whole-blood specimen. There was likewise no separately quantified serum or plasma concentration, no contemporaneous oral-fluid or hair drug quantitation, no vitreous or tissue toxicology, and no measured exhaled drug aerosol or breath-alcohol concentration. The urine observations concerned immunoassay classifications; the patches concerned accumulated sweat-device material. No validated transfer function connects an opiate patch result to blood mg/L, a specific organ effect or a driver’s performance. Every specimen has its own kinetics, chemical targets, collection period and detection limit; one matrix’s result cannot be relabeled as another’s concentration.

How should a forensic toxicologist evaluate a real monitoring dispute?

Begin with the actual device and documented chain of custody: initial application date, anatomical placement, adhesion throughout wear, date and circumstances of removal, observed tampering, storage, shipment and handling. Obtain the laboratory’s actual assay method, target compounds, identification standards, controls, instrument data and reporting rules for the relevant date. Then match the urine collection to the same time interval without treating a spot specimen as an identical temporal sample. Document all missing or rejected patches, not just results that survived. If a dispute involves a person’s use or impairment, that question requires independent case-specific evidence; the published 63-person observational comparison is not a substitute for examining what was actually collected and measured in that matter.

Where do the separately published site guides fit?

For a broader treatment of differences between blood, urine, sweat and other materials, see the independently published biological-matrices guide. The separately published specimen-collection guide addresses site, timing, integrity and documentation of evidence. The confirmatory-testing guide explains why a laboratory-specific analytical procedure must support identification. These links offer general context, not additional data or a purported independent replication of the Chawarski comparison. The primary five-page research report remains the controlling source for its own sample size, collection flow and comparison counts.

What is the narrow bottom line for drug-monitoring policy?

In this outpatient cohort the process placed many patches but recovered relatively few at precisely seven days. Even within the eligible subset, opiate-positive urine and sweat classifications coincided only about one third of the time, whereas the cocaine urine-positive comparisons coincided much more frequently. That difference is worth further analytical and clinical investigation, but it neither validates all cocaine patches nor negates all opiate patches. A practical monitoring program must assess specimen acceptability, adherence, compound scope, laboratory methods, participant circumstances and the cost of missing data before replacing a conventional collection method. The study informs method selection; it cannot establish any litigant’s exact consumption time, individual impairment, true intent or event causation.

Frequently asked forensic questions

1. Why was the study called a 63-person study if it used 536 patches?

There were 63 independently recruited adults. The 536 patches were repeated weekly applications to those same adults. Multiple visits, repeated laboratory specimens and different drug-class comparisons do not create additional independent biological participants.

2. Does 33.3% mean the sweat instrument itself had only 33.3% analytical sensitivity?

No. It is the fraction of urine-positive opiate events accompanied by a patch-positive result among 184 matched urine-patch events. Analytical limits, reference-window agreement and missed applications are distinct questions. The full assay validation needed to call it an intrinsic instrument sensitivity was not supplied.

3. Why are 188 and 184 both used?

There were 188 eligible on-schedule intact patches. Only 184 had a matching urine laboratory result in the reported comparison. Different missing-data requirements also produce 109 opiate self-report pairs and 99 cocaine self-report pairs; these denominators should never be pooled casually.

4. Did the investigators statistically prove cocaine detection was better than opiate detection?

They reported conspicuously different descriptive paired fractions: 33/36 for cocaine versus 27/81 for opiates. The source does not report a formal inferential test for that drug-class difference, and there was no chemical causal experiment to explain it.

5. Can a seven-day patch prove a drug was taken on a particular day?

No. The device collects material over a wear interval, which is not a validated time-stamped ingestion record. Exposure timing and the contribution of skin deposition, individual excretion and handling remain uncertain without independent information.

6. Did the study measure morphine or cocaine in legal whole blood or determine DUI impairment?

No. It reported qualitative classifications for weekly urine and sweat collection devices plus self-report. It did not quantify venous whole-blood, separated plasma or serum drug concentrations, administer a driving test, or validate an individual intoxication threshold.

7. What is the single most important practical limitation?

Only 188/536 applied patches met the prescribed intact seven-day return criterion. That operational shortfall matters alongside the 54 urine-positive opiate comparisons with negative patches. A supposedly convenient monitor is not effective when much of the collection history produces no valid result.

Read the original primary evidence

Chawarski MC, Fiellin DA, O’Connor PG, Bernard M, Schottenfeld RS. Utility of sweat patch testing for drug use monitoring in outpatient treatment for opiate dependence. Journal of Substance Abuse Treatment. 2007;33:411–415. Original research DOI 10.1016/j.jsat.2007.03.004. This independent scientific analysis preserves distinctions between 63 participants, repeated patch applications, paired tests and the unmeasured case-specific endpoints. It does not reproduce the publisher’s figures or original formatted tables.