A Four-Dollar Blood Index Predicted Poor Sleep in Addiction Better Than Anything Else in the Model

Sleep complaints are close to universal in addiction treatment and almost impossible to triage. Everyone reports bad sleep on intake; nobody has a cheap way to tell whose insomnia is a withdrawal symptom, whose is a co-occurring depression, and whose is going to be the reason they relapse in six weeks.

A paper published October 20 in Frontiers in Pharmacology argues that a number already sitting in the complete blood count can help sort them. The systemic inflammation response index — monocytes multiplied by neutrophils, divided by lymphocytes — ranked first among all variables a machine-learning model used to classify sleep status in people with substance use disorder, ahead of age, body mass index and race.

The analysis ran on 2,533 adults from three NHANES cycles (2013 through 2018) and then on a separate, deliberately cleaner cohort of 44 people with opioid use disorder recruited in Changsha, China, compared against 44 matched controls. Among the NHANES participants with substance use disorder, 436 reported sleep trouble and 816 did not. Comparing the highest SIRI quartile to the lowest, the fully adjusted odds ratio for sleep disturbance was 2.00. The gradient was monotonic: 1.15 in the second quartile, 1.29 in the third.

What the Mediation Analysis Adds

The more interesting number is not the odds ratio. It is 0.239 — the share of the total association between substance use disorder and sleep disturbance that the authors estimate runs through SIRI.

That means roughly 24 percent of the link operates through a pathway a routine blood draw can see, and 76 percent does not. In the Changsha cohort, the opioid group had higher SIRI than controls (0.93 against 0.74, p=0.045) and Pittsburgh Sleep Quality Index scores of 11 against 4.5, with a partial correlation of 0.327 between the two after adjusting for age, sex, BMI, nicotine dependence, alcohol screening scores and marital status.

The authors are explicit that this is not a diagnostic threshold. They propose a stratification framework: patients with opioid use disorder whose SIRI stays elevated might be a pro-inflammatory subgroup worth watching more closely for sleep deterioration, treatment resistance and relapse. SIRI would sit beside the PSQI and clinical judgment, not replace them.

Why It Would Be Easy to Overread

The design is cross-sectional, so the arrow could run in either direction — or both. Poor sleep raises inflammatory markers in experimental studies; inflammation degrades sleep architecture. The NHANES sleep measure is a single yes-or-no question about whether a clinician ever said the participant had trouble sleeping, which the authors note would bias estimates toward the null; the Chinese cohort used the full PSQI, so the two datasets are not measuring the same thing.

The clinical sample is small and unusually pure — no major psychiatric comorbidity, no antidepressants, no anti-inflammatory drugs, almost all with two decades of heroin use by hot smoking. That is a clean experiment and a poor mirror of an American treatment caseload, where polysubstance use and dual diagnosis care are the norm. The authors also concede their training-set AUCs (CatBoost 0.999) indicate overfitting.

What remains is a cheap hypothesis worth testing in a cohort that follows people forward. If a four-dollar index from a CBC flags which patients will still have insomnia at ninety days, that changes intake practice. Until then, it is a lead.