The research team, led by investigators from Guangdong Provincial Hospital of Chinese Medicine and affiliated institutions in Guangzhou and Foshan, China, drew their primary evidence from the MIMIC-IV critical care database, which contains records from Beth Israel Deaconess Medical Center in Boston between 2008 and 2019. From this database they identified 2,381 first-time ICU admissions diagnosed with acute myocardial infarction, covering both ST-elevation and non-ST-elevation heart attacks. In this derivation cohort, 316 patients, or 13.3 percent, died during hospitalization, and 237 patients, or 10.0 percent, died during their ICU stay. The median PWR was 16.20, and patients were divided into four quartiles ranging from below 11.27 in the lowest group to 22.90 or above in the highest.

The pattern in the raw data was striking. Patients in the highest PWR quartile had lower heart rates, lower white blood cell counts, less frequent use of vasoactive drugs such as epinephrine, norepinephrine, and vasopressin, and lower scores on the SOFA, APS III, and SAPS II illness-severity scales compared with those in the lowest quartile. Mortality declined steadily across the quartiles: in-hospital death fell from 17.3 percent in the lowest quartile to 9.1 percent in the highest, while ICU death fell from 14.3 percent to 6.9 percent. Notably, the gradient was strongest for short-term outcomes. Ninety-day mortality showed only a borderline trend, and 180-day and 365-day mortality did not differ significantly across PWR groups, suggesting the ratio functions primarily as an early, short-horizon prognostic signal rather than a long-term one.

To test whether the relationship was linear or followed some more complicated curve, the researchers applied restricted cubic spline analysis, trimming extreme PWR values below the 5th and above the 95th percentiles. The results showed no significant departure from linearity in either cohort, meaning that each incremental rise in PWR carried a roughly proportional reduction in risk across the observed range. Kaplan-Meier survival curves reinforced the message, with patients in higher PWR quartiles showing consistently better cumulative survival over their hospital and ICU stays, a difference that was statistically significant in the derivation cohort and even more pronounced in the external validation data.

The centerpiece of the analysis was a series of multivariable logistic regression models. After full adjustment for creatinine, hemoglobin, calcium, potassium, hypertension, hyperlipidemia, diabetes, atrial fibrillation, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic obstructive pulmonary disease, and the use of statins, aspirin, clopidogrel, and continuous renal replacement therapy, along with SOFA and APS III scores, each one-unit increase in PWR was associated with lower in-hospital mortality, with an odds ratio of 0.976, and lower ICU mortality, with an odds ratio of 0.977. Because a one-unit change in this ratio is small in clinical terms, the researchers also expressed the effect per interquartile range, a span of 11.63 units in the derivation cohort. On that scale, each interquartile increase in PWR corresponded to a 24.6 percent relative reduction in the odds of in-hospital death and a 23.7 percent relative reduction in ICU death. When PWR was treated categorically, patients in the top quartile had roughly 43 percent lower odds of in-hospital mortality and a similar reduction in ICU mortality compared with those in the bottom quartile.

Sensitivity analyses bolstered the finding. Cox proportional hazards regression yielded hazard ratios in the same direction and of similar magnitude, and Fine-Gray subdistribution hazard models, which account for competing events such as alive discharge or transfer, confirmed that the association persisted after this more conservative treatment of the data. The researchers also calculated E-values, a metric estimating how strong an unmeasured confounder would need to be to explain away the observed association entirely. In the derivation cohort, the E-value for in-hospital mortality was 2.92, with a conservative estimate of 1.61, figures the authors argue are substantially larger than the effect sizes of typical confounders, making residual confounding an unlikely full explanation.

Crucially, the association replicated outside the derivation dataset. In the multi-center eICU Collaborative Research Database, which aggregates electronic health records from 208 United States hospitals, 7,054 heart attack patients were analyzed, and the inverse relationship held: each unit increase in PWR was associated with lower in-hospital mortality, with an odds ratio of 0.986, and lower ICU mortality, with an odds ratio of 0.980, after full adjustment. Quartile analyses in this cohort showed the top two PWR groups had significantly lower mortality than the bottom group, with clear dose-response trends. A third, internal validation cohort of 252 patients from Guangdong Provincial Hospital of Traditional Chinese Medicine, admitted between October 2022 and October 2024, showed the same directional pattern, with the highest PWR quartile exhibiting dramatically lower in-hospital mortality than the lowest, an odds ratio of 0.082 after adjustment, though the wide confidence interval in this smaller sample reflects considerable statistical uncertainty.

Subgroup analyses examined whether the association differed by sex, age above or below 65, and the presence of hypertension, diabetes, heart failure, atrial fibrillation, COPD, end-stage renal disease, acute kidney failure, or sepsis. After applying the Benjamini-Hochberg procedure to control the false discovery rate across multiple interaction tests, no subgroup variable significantly modified the relationship. Diabetes interactions appeared nominally significant before correction but dissolved after adjustment, underscoring the importance of controlling for multiple comparisons. Decision curve analysis, which quantifies the net clinical benefit of a predictive model across a range of risk thresholds, showed that adding PWR to a base model of clinical covariates improved net benefit at several clinically relevant thresholds, for example around the 0.35 threshold probability for in-hospital mortality, where the difference in normalized net benefit reached statistical significance.

Why should a simple ratio of two blood cell types carry prognostic weight? The authors ground the finding in the biology of thrombo-inflammation. Acute myocardial infarction is fundamentally a event of plaque rupture and coronary thrombosis, but it also triggers a robust systemic inflammatory response. Leukocytes flood the injured vascular territory, releasing pro-inflammatory cytokines, matrix metalloproteinases, and oxidative stress products that worsen myocardial injury and impair recovery. Platelets, meanwhile, do far more than form clots: they bind to leukocytes through P-selectin and its ligand PSGL-1, form leukocyte-platelet aggregates, and release mediators such as CD40 ligand, platelet factor 4, and RANTES that amplify inflammation further. A low PWR, reflecting both persistently low platelets and persistently elevated white cells, may therefore signal unchecked inflammation, platelet consumption, and a deranged balance between the two arms of this vicious cycle. Consistent with this framing, mediation analysis suggested that disease severity, captured by the SAPS II score, mediated 25.9 percent of the protective association of PWR with ICU mortality, while vasoactive drug use, a proxy for circulatory failure, mediated 18.6 percent. The authors caution that these mediation results are exploratory, since the overlapping timing of measurements and the absence of molecular markers in the databases prevent direct mechanistic validation.

The study has limitations that temper enthusiasm. It is retrospective and observational, subject to selection and information biases. PWR was measured only once, within the first 24 hours, so dynamic changes during the acute and recovery phases could not be captured. The internal validation cohort allowed assessment of in-hospital mortality only, and the public databases lacked detailed medication histories, symptom-to-reperfusion times, and procedural specifics of coronary angiography and stenting. The authors themselves emphasize that larger, multi-center prospective studies are needed before PWR can be formally incorporated into risk stratification. Still, the appeal of the marker is hard to overstate: it requires no additional testing, no specialized equipment, and no added cost, and it can be computed the moment a complete blood count returns. If prospective validation confirms these findings, a single division of two numbers already on every admission lab panel could give clinicians an early, inexpensive warning sign about which heart attack patients need the most aggressive monitoring and treatment in the critical first days.

Subject of Research: Association between the platelet-to-white blood cell ratio and short-term in-hospital and ICU mortality in acute myocardial infarction patients

Article Title: Association between platelet-to-white blood cell ratio and short-term prognosis in patients with acute myocardial infarction: a retrospective cohort study

Article References: Kang, Z., Liao, Y., Zheng, C., Li, Z., Yao, G., & Zou, X. (2026). Association between platelet-to-white blood cell ratio and short-term prognosis in patients with acute myocardial infarction: a retrospective cohort study. Frontiers in Cardiovascular Medicine, 13, 1935579. https://doi.org/10.3389/fcvm.2026.1935579

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1935579

Keywords: platelet-to-white blood cell ratio, acute myocardial infarction, in-hospital mortality, ICU mortality, risk stratification, MIMIC-IV, eICU database, retrospective cohort study, thrombo-inflammation, biomarker, complete blood count, cardiovascular prognosis