The primary exposure was the first arterial PaO2 value obtained after placement of a qualifying upper‑body invasive venous catheter, subject to four timing constraints: the PaO2 had to be recorded at or after the catheter insertion index time, within 24 hours of insertion, during the same ICU stay, and while at least one qualifying catheter remained in situ. For each eligible ICU stay catheter episodes were reconstructed from documented insertion and removal times; the interval from insertion to the selected PaO2 was calculated in minutes. PaO2 was analyzed both continuously (odds ratios per 10 mmHg increase) and categorically (≤80, 81–100, >100 mmHg, with ≤80 mmHg as the reference). A secondary sensitivity analysis assessed PaO2/FiO2 as an alternative oxygenation exposure.
Data were sourced from the Medical Information Mart for Intensive Care IV database, version 2.2 (MIMIC‑IV v2.2), covering encounters from 2008–2019 at a single academic center. ICU stays were eligible if patients were ≥18 years, had an ICU length of stay ≥24 hours, at least one qualifying central venous catheter episode overlapping the ICU stay, and an arterial PaO2 measurement meeting the timing criteria. Catheters documented in upper‑body locations were retained; lower‑body catheters were excluded. From an initial extraction of 30,003 catheter records, application of eligibility rules yielded a final analytic cohort of 11,277 hospital admissions from 10,789 unique patients.
The primary outcome was a hospitalization‑level documented acute upper‑body deep or central venous thrombosis identified using a prespecified list of ICD‑9 and ICD‑10 diagnosis codes in the MIMIC‑IV diagnoses table. The primary definition included acute thrombosis of upper‑extremity deep veins, axillary, subclavian, or internal jugular veins. Superficial thrombosis and anatomically unspecified codes were excluded from the primary outcome. Two additional outcome constructs were tested in sensitivity analyses: a broader definition that included superficial or less specific upper‑body venous thrombosis codes, and a strict laterality‑compatible definition matching catheter laterality with laterality‑specified diagnoses. No imaging‑adjudicated reference standard was available, coding sensitivity and specificity could not be estimated, and diagnosis timestamps were unreliable; therefore, the temporal relationship between the selected PaO2 and thrombus onset could not be confirmed in every admission.
Potential confounders captured for adjustment included demographic factors (age, sex, race/ethnicity used for description and sensitivity analysis), comorbidity burden (Charlson Comorbidity Index), laboratory values (most recent hemoglobin and platelet count within 24 hours before the PaO2, and lactate preferentially from the same arterial blood gas), catheter characteristics (type, laterality, number of active catheters), timing variables (interval from insertion to PaO2 and from ICU admission to PaO2), respiratory variables (FiO2 at the PaO2 measurement and respiratory‑support status), and illness severity indicated by the non‑respiratory component of the first‑day SOFA score. Race/ethnicity was excluded from the primary fully adjusted model but included in descriptive and sensitivity analyses.
Logistic regression models examined the association between PaO2 and the hospitalization‑level thrombotic outcome. Models were sequentially adjusted: starting unadjusted, then adding patient characteristics and labs, and finally including catheter characteristics, measurement timing, FiO2, respiratory support, and non‑respiratory SOFA. Multiple imputation was used to address missing covariate data. Patient‑clustered robust standard errors accounted for repeated admissions by the same individual. The study analyzed admission‑level associations and was not designed as a time‑to‑event analysis because reliable onset timestamps for thrombosis were not available.
Among 11,277 hospital admissions, 195 admissions (1.73%) met the primary documented thrombotic outcome. In unadjusted models, higher PaO2 was associated with lower odds of the outcome (OR per 10 mmHg increase 0.970; 95% CI 0.952–0.987). After sequential adjustment for measured confounders, the association attenuated. In the fully adjusted model the authors report no independent association between the early post‑insertion PaO2 and documented upper‑body venous thrombosis (OR per 10 mmHg 1.008; 95% CI 0.989–1.027; P = 0.430).
Most secondary and sensitivity analyses supported the primary null finding. A 60‑minute sensitivity analysis produced estimates that depended on the adjustment set, indicating some sensitivity to measurement timing and model specification. The PaO2/FiO2 ratio was examined as an alternative exposure in sensitivity testing, though the main inference reported focused on PaO2 values as specified.
In this single‑center, retrospective MIMIC‑IV cohort, a single arterial PaO2 measurement obtained within 24 hours after qualifying catheter insertion was not independently associated with documented acute upper‑body deep or central venous thrombosis after multivariable adjustment. The findings argue against using a solitary early post‑insertion PaO2 for thrombosis risk stratification or as a rationale to change oxygen therapy solely to prevent catheter‑related upper‑body venous thrombosis in similar settings.
Strengths include a large, well‑characterized EHR cohort with detailed timing of catheter episodes and routine arterial blood gas data, predefined exposure and outcome algorithms, and multivariable adjustment with multiple imputation and clustered standard errors. Key limitations are reliance on diagnosis codes without imaging adjudication, lack of reliable thrombosis onset timestamps preventing confirmation that the exposure preceded the outcome, potential residual confounding from factors that influence both oxygenation and catheter selection or thrombotic risk, and the use of a single early PaO2 measurement rather than repeated oxygenation assessments.
The study concludes that a single early post‑insertion arterial PaO2 measurement was not independently associated with documented upper‑body venous thrombosis in this critically ill cohort. The authors recommend future research using repeated oxygenation measurements and accurately timed, imaging‑confirmed thrombotic outcomes to better evaluate any causal relationship between oxygenation and catheter‑associated thrombosis.