How do methamphetamine use and BMI jointly relate to interleukin-6 and TNF-alpha?

Two risk factors that reverse when combined

Methamphetamine use and obesity are each linked to inflammation, so the intuitive expectation is that the two together would produce the highest cytokine levels. This study found the opposite pattern for the combination. Across 560 specimens from 140 men, frequent methamphetamine use and higher body mass index (BMI) were each positively associated with interleukin-6 (IL-6), a cytokine released during immune activation. But the interaction term between frequent use and BMI was negatively associated with IL-6, meaning the expected elevation shrank as BMI rose among frequent users.

The same reversal appeared for tumor necrosis factor-alpha (TNF-alpha), another pro-inflammatory cytokine, though only once the interaction was included. The authors describe this as a paradoxical relationship and interpret it cautiously: the data show association, not causation, and the mechanism is proposed rather than demonstrated.

A three-level use variable built from two sources

Self-reported drug use can be distorted by recall and by the wish to give a socially desirable answer, so the authors did not rely on the questionnaire alone. They combined a urine drug screen, which captures recent exposure, with a computer-assisted self-interview asking how often participants used methamphetamine in the past six months. That produced three levels: no use (no self-report and negative urine screens), intermittent use (monthly, less than monthly, or once in the past six months, or a positive urine screen despite reporting no use), and frequent use (weekly or daily use in the past six months).

Across visits, 61.6% of specimens fell in the no-use group, 19.1% in intermittent use and 19.3% in frequent use. This composite measure is the study's key design choice: it lets the analysis distinguish occasional from heavy exposure rather than treating any positive test as equivalent.

A selected cohort of men in Los Angeles

The analysis is a secondary use of data and stored specimens from the mSTUDY, a longitudinal cohort of men who have sex with men in Los Angeles recruited in part through an LGBTQIA+ community organization and a university research clinic. By design, about half the cohort lives with HIV and half does not, and half has active substance use. The 140 participants contributed four biannual visits each, from September 2014 through March 2020, giving 560 specimens; 70 participants with recent methamphetamine exposure were compared with 70 who did not use substances.

Median age was 35 years and median BMI was 26.5. Because the sample is small, selected for HIV and substance use, and entirely male, the results describe this group rather than the general population. The authors note that sex differences in how these cytokines relate to BMI are plausible and untested here.

Frequent users had the lowest BMI

Before turning to cytokines, the descriptive results matter for interpretation. Median BMI was lowest among specimens with frequent methamphetamine use (25.7), higher with intermittent use (26.3), and highest with no use (27.0), a difference the authors report as significant. Sexually transmitted infection positivity also rose with use frequency, reaching 23.2% of specimens in the frequent-use group against 14.5% in the no-use group.

This matters because BMI and methamphetamine use are not independent in these data: the heaviest users tended to be leaner. Any model that treats the two as separate predictors is therefore describing a population in which the two exposures are already entangled, which is exactly why the authors added an interaction term.

Why the models work on a log scale

Cytokine concentrations are heavily right-skewed, with a few very high values stretching the distribution, and each participant contributes repeated measurements. The authors therefore used mixed-effects generalized linear models with a gamma distribution and a log link, which accommodates that skew and accounts for the repeated visits, and they report exponentiated coefficients. An exponentiated coefficient above 1 indicates a proportional increase in the cytokine level; below 1 indicates a proportional decrease.

Adjusted models controlled for age, HIV status and STI diagnosis. Time was initially included but showed no significant main effect or time-by-group interaction, so the final models compare levels between groups across the study period rather than tracking change within individuals over time. That distinction limits how the results should be read: this is a between-group comparison built on longitudinal data, not a demonstration that inflammation rose or fell within a given person.

IL-6 rises with each factor but not with both

In the main-effects model, BMI was positively associated with IL-6 (adjusted exponentiated coefficient 1.04, p = 0.004), and specimens from visits with frequent use had higher IL-6 than those with no use (1.74, p < 0.001). Intermittent use trended higher but was not statistically associated in either unadjusted or adjusted models.

The interaction model changed the picture. Frequent use alone carried a much larger coefficient (18.03, p = 0.001) and BMI remained positive (1.06, p < 0.001), but the frequent-use-by-BMI interaction was 0.92 (p = 0.002). Because the interaction multiplies the two main effects, the net association with IL-6 depends on BMI: the large positive frequent-use term applies at the lowest BMI, and each additional unit of BMI scales it down. The negative interaction is the paper's central quantitative claim, and it is a statement about the shape of the joint relationship, not about either factor alone.

TNF-alpha needs the interaction to appear

TNF-alpha behaved differently, which is one reason the two cytokines should not be treated as interchangeable readouts of the same process. In unadjusted models, frequent use was associated with higher TNF-alpha (1.16, p = 0.013), but after adjusting for age, STI diagnosis, HIV status and BMI, the frequent-use difference was no longer significant (1.10, p = 0.10). BMI itself was not significantly associated with TNF-alpha in the main-effects model (1.00, p = 0.31).

Only in the interaction model did the pattern emerge: frequent use (3.18, p < 0.001) and intermittent use (2.32, p = 0.005) were both positively associated with TNF-alpha, while the frequent-use-by-BMI interaction (0.96, p < 0.001) and the intermittent-use-by-BMI interaction (0.97, p = 0.005) were both negative. The authors suggest that much TNF-alpha activity in adipose tissue acts locally rather than systemically, and that its short half-life may explain why BMI alone did not show up in the main-effects model.

A proposed energy-balance explanation

The authors offer a physiological reading rather than a proven pathway. Peripherally, IL-6 promotes energy expenditure, thermogenesis and the breakdown of protein and fat; centrally, both IL-6 and TNF-alpha act on the hypothalamus to suppress appetite. Methamphetamine independently suppresses appetite through dopaminergic signaling in the hypothalamus. On this account, frequent users with higher BMI may have a blunted inflammatory response, reduced energy expenditure and increased appetite, which would help sustain the higher body weight.

The paper supports the plausibility of this chain with prior work cited in its discussion, including observations that mice lacking TNF-alpha or IL-6 carry more body fat, and that patients treated with cytokine inhibitors such as tocilizumab or infliximab gain BMI. None of this is tested in the present cohort, so the mechanism remains an interpretation of the association, not a finding of the study.

What the interaction does and does not establish

The practical weight of the paper sits in one comparison: among men with higher BMI, more frequent methamphetamine use was associated with lower IL-6 and TNF-alpha, while each factor on its own pointed toward more inflammation. That makes BMI a modifier worth accounting for when inflammation is assessed in people who use methamphetamine, and it cautions against reading a single cytokine value as a straightforward index of stimulant exposure or of adiposity.

The boundaries are equally specific. This is an observational secondary analysis, so causal inference is not possible and confounding cannot be excluded. Cannabis use was too common to isolate methamphetamine entirely, though sensitivity analyses indicated it did not change the findings. BMI cannot separate fat from lean mass or visceral from subcutaneous fat, which the authors flag as relevant to inflammatory risk. And the sample is 140 men who have sex with men with high HIV prevalence, so the interaction should be examined in other populations before it informs anyone's reading of an individual patient's cytokine panel.

Sources used in this answer