Accurate assessment of substance use is crucial for public health interventions, clinical treatment, legal proceedings, and scientific research. High-quality data on substance use is essential for understanding its effects on various aspects of human health, including brain function and cognitive processes. In research contexts, precise measurements of substance use can clarify how different substances impact brain and cognitive functions, ultimately guiding the development of effective treatment and intervention strategies.

Traditional methods of assessing substance use often rely on self-reports, where individuals disclose their substance use behaviors through questionnaires or interviews. These methods are popular due to their simplicity and cost-effectiveness. However, they are subjective and vulnerable to biases such as recall errors, social desirability, and both intentional and unintentional underreporting (1–4). These biases can compromise data accuracy, impacting research outcomes.

Intentional underreporting of drug use may occur because of the sensitive nature of the topic—participants may fear judgment, social stigma, or legal consequences from disclosing their drug use (5). Concerns about personal relationships or professional consequences can also increase the likelihood of underreporting (5). Not all underreporting is deliberate, however. Individuals might forget their drug use, misunderstand survey questions, or be unaware of exposure to contaminated or mixed drugs (5). For example, psychoactive drugs such as heroin and ecstasy (MDMA) are often mixed with other substances, including methamphetamine or new psychoactive substances (NPS) (6–8). A growing concern is the presence of fentanyl and its analogs in drugs like heroin and cocaine, which increases the risk of overdose for users unaware of the contaminants (9, 10).

Advancements in analytical techniques have introduced objective measures of substance use, such as biological sample analysis. These methods offer a more reliable assessment by detecting the presence of substances and their metabolites in bodily fluids or tissues. Common biological testing methods—such as blood, saliva, and urine tests—are well-suited for capturing recent use within a few days, making them effective for evaluating acute and recent drug effects. However, they may not be ideal for identifying long-term drug use or its residual effects (11).

Hair sample analysis provides a unique advantage in substance use research due to its extended detection window—enabling retrospective identification of drug use over several months (12–14). This feature is particularly valuable for studies that aim to assess longer-term patterns of drug use and compare them with self-reported behavior. Unlike urine, saliva, or blood samples—which primarily detect recent consumption over hours to a few days—hair offers a cumulative record of substance exposure over time, making it a more suitable choice for studies examining broader usage trends.

From a practical standpoint, hair collection is non-invasive, does not require specialized medical personnel, and is relatively easy to transport and store. Compared to saliva or urine (if not immediately tested), which can be susceptible to adulteration and often demand immediate processing or refrigeration, hair provides a stable matrix for toxicological analysis (15). Despite these advantages, hair analysis does present challenges: it is more costly than other methods, and participants may be hesitant to provide samples due to cosmetic concerns or discomfort with the amount of hair required. Previous studies have reported hair sample refusal rates ranging from 20% to over 85% in non-clinical samples in research settings (15–18).

Additionally, individual differences in hair growth and potential external contamination are known limitations, although these can be mitigated with standardized washing procedures and validated analytical protocols (19). Moreover, previous research indicates that hair sample analysis is characterized by high specificity but relatively lower sensitivity, which depends on the type of substance and the individual's history of use (20, 21, 45). High specificity ensures a low probability of obtaining false-positive results, making it relatively reliable for verifying self-reported abstinence from substance use. However, the lower sensitivity increases the risk that substances that have actually been consumed remain undetected. Hair analysis demonstrates the highest sensitivity for substances such as opioids (e.g., heroin) and the lowest for cannabis (15). Sensitivity is also influenced by the amount and frequency of substance use; it tends to be lower in cases of low-dose or infrequent consumption (22). Additionally, drug concentration in hair is affected by hair structure and melanin content (23) and may be altered by cosmetic treatments such as perming, straightening, or dyeing, which can result in gender and ethnic differences in measurement accuracy (24). It is important to note that hair toxicology cannot reliably detect very recent substance use because drugs require time to be incorporated into the growing hair shaft. As a result, consumption occurring within approximately the past week is typically not observable in hair samples, despite the method's broad retrospective detection window (22, 25).

Alternative biological matrices were considered but found less appropriate for our research objectives. Saliva testing, though highly acceptable and minimally invasive, detects drug use only within a narrow window (typically up to 48 h post-consumption) (26), limiting its utility for evaluating sustained patterns of use. Blood testing offers precise quantification but is invasive and impractical outside clinical environments. Urine testing has a moderately longer detection period and is more commonly used, yet it remains vulnerable to manipulation and is still limited in capturing chronic or historical use (11).

Given our study's aim—to examine the agreement between self-reported substance use and objective biomarkers over an extended timeframe—hair analysis provided the most appropriate balance of temporal sensitivity, data stability, and feasibility in a non-clinical research context. While we acknowledge its limitations, the capacity to detect multiple substances retrospectively over months outweighed concerns about participant burden or sample variability.

Liquid chromatography-mass spectrometry (LC-MS) is considered the gold standard for detecting drugs, including delta-9-tetrahydrocannabinol (THC) (27). Research suggests that the sensitivity of THC detection in hair depends on usage patterns: for heavy cannabis users, the sensitivity is around 80%, but it drops to 55% when distinguishing any cannabis users from non-users (28). This suggests that hair analysis alone may not accurately detect cannabinoid exposure. As Kroon et al. (29) emphasized, further research is needed to compare hair-derived samples with self-reports to better assess the validity of hair analysis for this purpose.

Despite the increasing use of hair testing in clinical and forensic contexts, relatively few studies have directly compared self-reported substance use to hair toxicology in non-treatment-seeking, research-based populations. Moreover, there remains a lack of clarity regarding which patterns of self-reported cannabis use—such as frequency, quantity, or recency—are most predictive of a positive THC result in hair. This study addresses these gaps by examining the degree of agreement between self-reported substance use and hair toxicology outcomes for four commonly used drugs: cannabis, MDMA, amphetamine, and cocaine. We focus particularly on cannabis use, testing whether THC detection in hair is predicted by frequency of use, grams consumed per week, and time since last use. In addition, we examine the prevalence of underreporting—defined as a mismatch between negative self-report and positive hair test—and assess whether individual characteristics such as sex, age, education, mental health diagnosis, and alcohol use predict this discrepancy.

By exploring these associations in a Central European sample, we aim to contribute to improving the validity of substance use assessments in research settings. This is especially important given the growing reliance on both self-report and biological measures in psychiatric and neuroscientific studies. This research is particularly significant in light of the meta-analyses by Bharat et al. (20), which, while demonstrating high overall agreement between biological and self-reported methods, highlighted important limitations—mainly, a highly heterogeneous evidence base and a sample predominantly (91%) from North America. Given that substance use patterns differ significantly between Europe and the United States (as well as across Europe itself) due to varying cultural, legal, and social influences [e.g., (30–33)], our data from Poland provide an additional perspective. By situating our findings within this broader international context, this study contributes novel insights into the accuracy of self-reporting in a distinct cultural and regulatory setting.

The present study provides important insights into the relationship between self-reported substance use and objective biological measures, such as hair toxicology analysis, particularly for cannabis use. These findings build upon the work of Bharat et al. (20), who conducted a meta-analysis examining the overall agreement between self-reports and biological measures.

Bharat et al. (20) found a generally high level of agreement, though it's crucial to recognize that the evidence base in their study was largely based on North American participants, a region with unique cultural and legal contexts for substance use. In contrast, our study from Poland found descriptivediscrepancies between self-reported substance use and hair toxicology results, with effect sizes suggesting possible patterns for certain substances, such as MDMA and cocaine, where underreporting could be more prevalent due to social stigma or legal concerns.

The finding that approximately 21.3% of participants underreported their drug use, as determined by discrepancies between self-reports and hair toxicology, suggests a substantial level of underreporting and aligns with existing literature on the limitations of self-reported data in substance use research (22, 41). This underreporting is a critical issue in research and can affect data quality, indicating that while not a majority, a notable portion of participants did not fully disclose their substance use. Such discrepancies highlight the potential limitations of relying exclusively on self-report measures.

It is important to note that the small sample sizes across several substance categories limit the study's statistical power to detect significant differences, potentially contributing to the non-significant p-values. To account for this, we reported effect sizes (Φ and Cohen's g) along with 95% confidence intervals for each substance. This approach allows the data to be interpreted in ways that extend beyond simple significance testing and provides descriptive insight into potential patterns of interest. Despite the lack of statistical significance in many χ² comparisons, effect sizes offer valuable insight into the strength of observed discrepancies and are used here in a descriptive rather than inferential manner. For MDMA, we observed a trend toward significance and a large effect size, suggesting differences between self-reports and objective measures. For cocaine, a similarly large effect size was found despite non-significant p-values, which may be consistent with possible underreporting biases, although these findings should be interpreted with caution given the small number of positive and discordant cases. These trends may reflect the complexity of self-reporting, particularly for substances that carry social or legal stigma (42, 43). In the case of amphetamine, the effect size was small, indicating a potential alignment between self-reports and hair samples—though the small sample size should again be noted as a limitation.

This rate of underreporting is consistent with previous studies that have identified both intentional and unintentional biases in self-reporting, driven by various psychological, social, and contextual factors (44). Several factors likely contributed to the observed underreporting in our study. One likely contributing factor is social desirability bias, where participants may downplay their substance use to present themselves in a more favorable light. In research settings—especially those involving sensitive topics—participants may fear judgment or stigmatization, leading to underreporting (5). Another plausible factor is the fear of legal or social consequences. Even in confidential studies, participants may hesitate to disclose illegal behaviors. This concern may be particularly relevant in countries like Poland, where drug use is both legally restricted and socially stigmatized.

Unintentional underreporting may additionally arise from recall difficulties, especially when substance use is sporadic, irregular, or occurred in social contexts involving intoxication (e.g., parties). Participants may struggle to remember frequency, quantity, or timing accurately. Additionally, involuntary exposure to adulterated drugs may explain some cases of underreporting. For instance, MDMA and other party drugs are often mixed with methamphetamine or other psychoactive compounds (6, 7), meaning users might test positive for substances they did not knowingly consume.

While the effect sizes suggest possible differences between self-reports and objective measures, our statistical analyses of potential predictors of underreporting did not reveal any significant associations with sociodemographic variables or alcohol consumption. Neither sex, age, education level, mental health status, nor drinking behavior significantly predicted underreporting, but rather influenced by personal, social, or contextual factors. It may also point to recall issues, particularly in cases of infrequent or incidental drug use.

In the case of THC, underreporting was observed in 9.3% of participants. Interestingly, a higher percentage of participants self-reported cannabis use (53.3%) than those who tested positive via hair analysis (42.7%), which may point to limitations in the sensitivity of hair toxicology in detecting cannabis use. Previous studies have shown that THC detection in hair varies considerably depending on use intensity—while sensitivity can reach around 80% for heavy users, it drops to approximately 55% for occasional users (28). The presented findings are consistent with previous research indicating that hair toxicology demonstrates high specificity but significantly lower sensitivity, particularly among occasional users (20–22). Therefore, negative hair test results may reflect methodological limitations of hair analysis, particularly among infrequent users. At the same time, the presence of underreporting in a subset of participants underscores the need to incorporate objective measures alongside self-reports. These findings suggest that neither method may be sufficient on its own—while hair testing may miss some true cases of use, self-reports are susceptible to bias. Taken together, this pattern highlights the importance of a multi-method approach to improve the accuracy and completeness of cannabis use assessment.

The study also examined self-reported cannabis use patterns as predictors of THC detection in hair. Frequency of use and quantity consumed (grams per week) were significant predictors, whereas duration of use and time since last use were not. These results are consistent with prior research indicating that recent and cumulative exposure are more relevant for hair detection than lifetime use history.

It should be noted that 12 participants (13.8%) from the larger EEG project (the original study from which the data were drawn) declined to provide hair samples. Despite assurances of confidentiality and the non-invasive nature of the procedure, the primary reason for refusal was cosmetic, particularly a reluctance to cut visible hair. While this attrition reduced the final sample size for hair-based measures, the refusal rate in our study is relatively low when compared to prior research. Previous studies have reported hair sample refusal rates ranging from 20% to over 85% in community-based or non-clinical populations in research settings (15–18). Nevertheless, future research should consider incorporating alternative biological sampling methods (e.g., saliva or urine), which may enhance participation rates without compromising data quality.

In light of these results, it is essential for future research to address the challenges of underreporting and identify the factors contributing to discrepancies between self-reports and biological measures. For example, future studies could investigate the impact of survey format (e.g., anonymous online questionnaires vs. face-to-face interviews) or perceived confidentiality on response accuracy. Expanding the sample to include participants from more diverse cultural and geographic contexts would also help to clarify whether certain populations are more prone to underreporting.

These findings underscore the importance of integrating objective measures such as hair toxicology to supplement self-reported data—especially for substances that carry a higher risk of underreporting. At the same time, self-reports offer valuable contextual information, such as usage motives or settings, that biological measures cannot capture. Therefore, a combined approach may be the most informative. It is important to acknowledge, however, that biological testing methods—particularly hair analysis—require specialized equipment and can be costly, which may limit their use in large-scale research.

In conclusion, this study highlights the complexities of substance use assessment and the importance of triangulating methods to improve accuracy. Despite limitations such as small sample size and the exclusive use of hair testing, the results contribute to ongoing efforts to improve the reliability of substance use data—especially in underrepresented populations and culturally specific contexts.