Summary
A study of 30,533 US adults found that among three routine blood-test inflammatory indices, only the neutrophil-to-lymphocyte ratio independently predicted cardiovascular mortality, with SII and PLR adding little predictive value.
Only NLR independently predicts cardiovascular death among three common inflammatory indices
A large development and external validation study published as a preprint on medRxiv found that among three inflammation-related ratios calculable from a routine complete blood count, only the neutrophil-to-lymphocyte ratio (NLR) independently predicted cardiovascular mortality. The systemic immune-inflammation index (SII) and platelet-to-lymphocyte ratio (PLR) offered little additional predictive value once NLR was accounted for.
The findings, based on 30,533 US adults in the National Health and Nutrition Examination Survey (NHANES), challenge the practice of reporting all three indices as separate cardiovascular risk markers without establishing whether each contributes unique information.
What the study examined
NLR, SII, and PLR are all derived from a standard complete blood count and are widely reported in the medical literature as markers of systemic inflammation and cardiovascular risk. NLR is the ratio of neutrophils to lymphocytes. PLR is the ratio of platelets to lymphocytes. SII is calculated as platelet count multiplied by NLR, meaning it incorporates both platelet and neutrophil-lymphocyte information.
Because SII is mathematically derived in part from NLR, it is unclear whether SII captures information beyond what NLR already provides. The researchers set out to test this directly, and to evaluate whether any of the three indices adds meaningful predictive value for cardiovascular death beyond established risk factors.
How the study worked
The study used two independent cohorts from NHANES, a nationally representative survey conducted by the US Centers for Disease Control and Prevention. The development cohort included 15,141 adults surveyed between 2005 and 2010; the validation cohort included 15,392 adults surveyed between 2011 and 2016. All participants had complete blood count data available and were linked to mortality follow-up records.
The researchers applied survey-weighted Cox proportional hazards models, restricted cubic splines, and multiple imputation to handle missing data. They developed a prediction model in the 2005–2010 cohort and then applied the frozen model to the 2011–2016 cohort for external validation. Model performance was assessed using discrimination (C-index) and calibration.
During a median follow-up of 11.4 years in the development cohort, 795 cardiovascular deaths occurred. The validation cohort recorded 364 cardiovascular deaths over a median of 5.75 years.
NHANES data collection is approved by the National Center for Health Statistics Research Ethics Review Board, and secondary analysis of de-identified public data does not require additional institutional review board approval. The work is currently a preprint posted on medRxiv and has not yet undergone peer review.
What the data showed
In the development cohort, each 1-standard-deviation increase in NLR was associated with a 32% higher hazard of cardiovascular death (hazard ratio [HR] 1.32, 95% confidence interval [CI] 1.21–1.44). The corresponding figures were HR 1.21 (95% CI 1.12–1.32) for SII and HR 1.13 (95% CI 1.04–1.22) for PLR.
When NLR and SII were entered into the same model, only NLR remained independently associated with cardiovascular mortality (HR 1.49, 95% CI 1.23–1.81). SII lost its statistical significance.
The validation cohort showed the same pattern. NLR was the strongest predictor (HR 1.35, 95% CI 1.22–1.49), and neither SII nor PLR remained significant after mutual adjustment.
The frozen model demonstrated good discrimination, with C-index values between 0.8991 and 0.9023, and good calibration (slope 0.96). However, adding NLR to a model that already included established risk factors produced only a small improvement in discrimination, and reclassification results were inconsistent between the two cohorts.
Why it matters
These findings have practical implications for both clinical research and any future clinical use of inflammatory indices. If SII and PLR do not add predictive value beyond NLR, then reporting all three as independent risk markers may be misleading. Researchers and clinicians who consider using SII or PLR should first test whether NLR already captures the relevant information.
The study also highlights a broader methodological point: indices derived from the same underlying blood-count components should not be treated as independent predictors without formal testing. SII is constructed from platelet count and NLR, so it is mathematically nested within NLR; treating them as separate predictors can inflate their apparent importance.
At the same time, the modest incremental value of NLR itself — even though it was the strongest of the three — suggests that these ratios add limited predictive power to a model already containing established cardiovascular risk factors. The C-index remained above 0.89, indicating that the baseline model was already strong, leaving little room for improvement from inflammatory markers alone.
Limitations
The study is limited by its observational design, which means it can demonstrate association but not causation. The analysis focused on cardiovascular mortality as the endpoint; it did not examine non-fatal cardiovascular events or other outcomes. The cohorts are drawn from a single country's survey, and the findings may not generalise to other populations.
The relatively short median follow-up in the validation cohort (5.75 years) compared with the development cohort (11.4 years) is another consideration, though the consistent pattern between cohorts strengthens confidence in the findings.
Additionally, because this is a preprint, the results may change during peer review.