Characteristics of the study cohort
Of the 236 patients, the median age was 55 years. The majority of them were male (78.8%). 35.2% had cardiovascular disease, 86.0% were on mechanical ventilation, 70.3% were taking vasoactive drugs, and 52.5% were suffering from sepsis shock. The median score for APACHE II and SOFA at baseline were 20 and 8, respectively. The incidence of in-hospital mortality and 30-day mortality were 46.6% and 40.7%, respectively. The total bacterial clearance rate was 22.9% (54/236) compared to 36.9% (87/236) for CRKP clearance. Additionally, 106 (44.9%) received TG (TG group), 102 (43.2%) received PB (PB group), and 28 (11.9%) received TG combined with PB (TG+PB group). Meropenem, imipenem, and sulbactam cephalosporin were mainly in combination. More men, middle-aged with pulmonary infections, multiple sites of infection, sepsis, unstable hemodynamics, critical illness, and high mortality characterized our cohort, which makes it different from those in other studies. The clinical characteristics of the patients are listed in Table 1.
In the TG+PB group with 28 cases, 57.1% of the patients had multiple infection sites, 42.9% had blood infections. The median APACHE II score and SOFA score were 20.50 and 8.00, respectively (Table 2). The median serum creatinine level before and after treatment were 201.00 µmol/L and 165.5 µmol/L, respectively. Moreover, 35.7% cleared CRKP, 10.7% cleared all bacteria, 64.3% died within 30 days, and the median 30-day survival was 16.50 days. The hospitalization cost was 56.09 thousand dollars. More critically ill may have masked some of the efficacy of TG+PB-based therapy. Because of the limit of the small sample size, comparsion analysis were not employed among TG+PB group and other groups.
TG group vs. PB group
The TG group had more patients with simple abdominal infection than those in the PB group (19.8% vs. 8.8%, P = 0.040). At baseline, patients in the TG group had a lower median APACHE II score (18.00 vs. 20.00, P = 0.001), SOFA score (6.50 vs. 8.00, P < 0.001) and more of them were taking vasoactive agents (63.2% vs. 79.4%, P = 0.015). Regarding the secondary outcome, patients in the TG group took vasoactive agents for fewer days (2.50 vs. 5.00, P= 0.015), underwent mechanical ventilation for fewer days (6.00 vs. 13.00, P = 0.003), and incurred lower hospitalization costs (40.28 vs. 50.80 thousand dollars, P = 0.003) than those in the PB group. Concerning the primary outcome, fewer patients in the TG group were able to clear CRKP than those in the PB group (31.1% vs. 43.1%, P = 0.073). The 30-day mortality in the TG group was lower than that in the PB group (33.0% vs. 42.2%, P = 0.223), as was in-hospital mortality (37.7% vs. 51.0%, P = 0.075) (Table 2).
Due to the differences of the baseline between TG and PB groups, propensity-score matching was performed to control the confounders. 58 pairs of cases were matched. After PSM, the baseline characteristics of the two groups were balanced, including the age, infection factors, therapy factors, underlying dieases, and other potential prognostic factors. The propensity-score were TG: 0.49± 0.13 vs. PB: 0.49± 0.13 (P=0.975)(Figur 1 ). PB group showed higher CRKP clearance (46.6% vs. 22.4%, P = 0.011) and complete bacterial clearance (32.8% vs. 15.5%, P = 0.051) than TG group. PB group had lower 30-day mortality than TG group without statistical significance (34.5% vs. 37.9%, P = 0.847). Furthermore, PB group showed more expensive hospital cost than TG group (Table 3). Why were bacterial clearance statistically significant, but 30-day mortality not ? We taken binary logistic and COX regressions to verified further.
Binary logistic regression of 30-day mortality
Univariate analyses about 30-day mortality and bacterial clearance and subsequent logistic regression analysis were applied. The results indicated that age, septic shock, mechanical ventilation, vasoactive agents, and TG/PB treatment may be associated with different incidences of 30-day mortality (Supplementary Table 1). Binary logistic regression was performed to demonstrate these factors.
For 30-day mortality, CRKP clearance was an independent protective factor (OR 0.312, 95%CI 0.159-0.612, P = 0.001). A shorter duration of PB was an independent risk factor compared to longer duration (4-8 days vs. > 8 days, OR 2.974, 95%CI 1.297-6.820, P = 0.010). In addition, the duration of vasoactive agents was an independent risk factor (1-4 days vs. 0 days, OR 2.90, 95%CI 1.146-7.352, P = 0.025; >4 days vs. 0 days, OR 8.843, 95%CI 3.845-22.338, P = 0.000). No significant risk difference was found between TG-based and PB-based therapies (PB vs. TG, OR 0.554, 95%CI 0.096-3.210, P = 0.510). (Table 4).
Cox regression analysis for 30-day mortality
Finally, Cox regression survival analyses were carried out to compare the incidence of 30-day mortality and survival. CRKP clearance (P = 0.000), longer duration of PB (P = 0.000) and less use of vasoactive agents (P < 0.050) were independent protective factors for 30-day cumulative survival. While the PB group had a better survival outcome than the TG group, the difference was not significant ( PB vs. TG, OR, 0.858, 95%CI 0.306-2.402, P = 0.770). (Table 5 and Figure 2).