The investigators analyzed 2,533 eligible participants from three National Health and Nutrition Examination Survey cycles: 2013–2014, 2015–2016 and 2017–2018. The survey analysis included 1,252 participants classified as having substance use disorder and 1,281 healthy controls. Substance use disorder was diagnosed according to the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, using self-reported information on lifetime use of cannabis with or without other illicit drugs, and cocaine, heroin and/or methamphetamine.

Sleep disturbance was identified with a single yes-or-no question asking whether a doctor or other health professional had ever told participants that they had trouble sleeping. Because the data were cross-sectional, the findings show associations but cannot establish whether substance use, inflammation, impaired sleep or another factor preceded the others. SIRI was calculated by multiplying the monocyte count by the neutrophil count and dividing the result by the lymphocyte count.

The formula combines circulating measures associated with innate immune activity and adaptive immune function, allowing the researchers to assess systemic inflammatory balance using routine hematology measurements. Among participants with substance use disorder, those reporting sleep disturbance had a median SIRI of 1.217, compared with 1.043 among those without sleep disturbance. The difference was statistically significant, with a reported p-value below 0.001.

The clinical cohort also showed a difference in the same direction: mean SIRI was 0.93 in the opioid use disorder group and 0.74 in healthy controls. Adjusted survey analysis showed a graded relationship between SIRI and the odds of reporting sleep disturbance. The model accounted for age, sex, education, race, marital status, annual family income and body mass index.

The odds ratio was 1.29, with a 95% confidence interval of 1.19–1.39, for SIRI from 1.125 to 1.622 and 2.00, with a 95% confidence interval of 1.85–2.17, for SIRI above 1.622. These are cross-sectional odds ratios rather than estimates of future risk, and the categories were defined within the analyzed sample. The researchers also performed a mediation analysis across the full survey sample, treating substance use disorder status as the exposure, SIRI as the mediator and sleep disturbance as the outcome.

The estimated total effect was 0.168, with a 95% confidence interval of 0.120–0.216; the average direct effect was 0.128, with a 95% confidence interval of 0.080–0.173; and the average indirect effect through SIRI was 0.040, with a 95% confidence interval of 0.026–0.056. The estimated proportion of the association statistically mediated through SIRI was 23.9%, with a 95% confidence interval of 15.3%–35.1%. The remainder was described as direct or potentially mediated through unmeasured pathways.

The analysis supports a partial statistical relationship, but its cross-sectional design does not demonstrate causal mediation. The clinical dataset provided a focused test of the same pattern in opioid use disorder. Between January and December 2024, researchers recruited 44 people with opioid use disorder from Changsha, Hunan Province, and 44 healthy controls matched on age, education, sex, body mass index, smoking and drinking.

The opioid group had used opioids for a median of 18 years, and all reported hot smoking as their primary route of administration. Urine screening was positive for heroin, and participants met Diagnostic and Statistical Manual of Mental Disorders criteria for opioid use disorder. The groups did not differ significantly in most demographic characteristics, although marital status differed significantly.

Sleep quality in the clinical cohort was measured with the Chinese version of the Pittsburgh Sleep Quality Index, an 18-item questionnaire organized into seven domains and scored from 0 to 21, with higher scores indicating poorer sleep quality. The median score was 11 in the opioid use disorder group and 4.5 in healthy controls, a significant difference with a reported p-value below 0.001. Within the opioid group, SIRI was positively correlated with PSQI, with a partial correlation coefficient of 0.327 and a p-value of 0.037 after adjustment for age, sex, body mass index, education, Fagerstrom Test for Nicotine Dependence score, Alcohol Use Disorders Identification Test score and marital status.

The correlation was modest, and the sample was small and geographically narrow. Machine-learning analysis offered a second way of examining whether SIRI contributed information about sleep status within the survey population. Eleven algorithms were evaluated, including LASSO, CatBoost, gradient boosting, support vector machines, decision trees, LightGBM, neural networks, naive Bayes, XGBoost, random forest and logistic regression.

The data were divided into stratified training and test sets at a 70%–30% ratio, with ten-fold cross-validation used for training and hyperparameters selected according to mean cross-validated area under the receiver operating characteristic curve. Missing continuous variables were imputed with medians, categorical variables with modes, and variables with at least 10% missingness were excluded before modeling. CatBoost had the highest test-set area under the ROC curve at 0.880, followed by random forest at 0.846.

The other algorithms had test-set areas ranging from 0.500 to 0.688. In the training set, CatBoost and random forest achieved areas of 0.999 and 0.996, respectively, indicating a degree of overfitting. Test-set calibration and decision-curve analyses identified CatBoost as demonstrating optimal performance and the greatest net benefit.

The investigators used SHAP analysis to interpret model predictions. SIRI ranked first in both models. In CatBoost, the next leading features were race, body mass index, age and sex; in random forest, they were neutrophil count, body mass index, monocyte count and white blood cell count.

These results identify SIRI as an influential feature in the fitted models, not as a validated diagnostic test. Biologically, the findings were consistent with previous observations discussed by the investigators that substance use and withdrawal can be accompanied by systemic immune responses, while inflammatory signaling can influence the central nervous system through humoral mediators, neural pathways and transport across the blood-brain barrier. Neutrophils and monocytes participate in innate immune regulation and can generate pro-inflammatory factors including tumor necrosis factor-alpha and interleukin-6.

The paper also discusses possible immune effects of opioids, including signaling through toll-like receptor 4 on myeloid cells. The present study did not directly measure these proposed pathways, test anti-inflammatory treatments or demonstrate that inflammation caused sleep disruption. Its evidence is observational and cross-sectional.

Important limitations constrain the conclusions. Sleep was self-reported in both datasets, creating the possibility of recall and classification bias. The NHANES measure was a single question, whereas the clinical cohort used the multidimensional PSQI, so sleep estimates from the two populations were not directly equivalent.

Detailed route-of-use information was unavailable in the survey data, limiting comparisons between intravenous and other administration methods. The clinical cohort contained only 44 participants with opioid use disorder, excluded people with major psychiatric comorbidities, antidepressant use or anti-inflammatory treatment, and represented one Han Chinese community in Changsha. External validation, longitudinal follow-up and experimental work are therefore needed before the observed associations can establish temporal order, causation or clinical utility.

The evidence supports SIRI as a promising marker for further study alongside established sleep assessments, rather than as a standalone diagnostic or treatment-selection tool.

Subject of Research: Association of the systemic inflammatory response index with sleep disturbance in substance use disorders

Article Title: Systemic inflammatory response index: a potential mediator between substance use disorder and sleep disturbance

Article References: Yang, L., Tang, Y., Wang, X., Chen, X., Zou, Q., Zeng, Y., Chen, Q., & Liu, J. (2026). Systemic inflammatory response index: a potential mediator between substance use disorder and sleep disturbance. Frontiers in Pharmacology, 17, Article 1964417. https://doi.org/10.3389/fphar.2026.1964417

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1964417

Keywords: systemic inflammatory response index, substance use disorder, sleep disturbance, opioid use disorder, peripheral inflammation, NHANES, machine learning, cross-sectional study, Pittsburgh Sleep Quality Index, blood biomarkers, Systemic, inflammatory