Pharmacokinetics of Ceftazidime-Avibactam in Combination with Aztreonam (COMBINE) in a Phase 1, Open-Label Study of Healthy Adults

ABSTRACT

Scant pharmacokinetic (PK) data are available on ceftazidime-avibactam (CZA) and aztreonam (ATM) in combination, and it is unknown if CZA-ATM exacerbates alanine aminotransferase (ALT)/aspartate aminotransferase (AST) elevations relative to ATM alone. This phase 1 study sought to describe the PK of CZA-ATM and assess the associations between ATM exposures and ALT/AST elevations. Subjects (n = 48) were assigned to one of six cohorts (intermittent infusion [II] CZA, continuous infusion [CI] CZA, II ATM, CI ATM [8 g/daily], II CZA with II ATM [6 g/daily], and II CZA with II ATM [8 g/daily]), and study product(s) were administered for 7 days. A total of 19 subjects (40%) had ALT/AST elevations, and most (89%) occurred in the ATM/CZA-ATM cohorts. Two subjects in the CI ATM cohort experienced severe ALT/AST elevations, which halted the study. All subjects with ALT/AST elevations were asymptomatic with no other signs of liver injury, and all ALT/AST elevations resolved without sequalae after cessation of dosing. In the population PK (PopPK) analyses, CZA-ATM administration reduced total ATM clearance by 16%, had a negligible effect on total ceftazidime clearance, and was not a covariate in the avibactam PopPK model. In the exposure-response analyses, coadministration of CZA-ATM was not found to augment ALT/AST elevations. Modest associations were observed between ATM exposure (maximum concentration of drug in serum [C_max] and area under the concentration-time curve [AUC]) and ALT/AST elevations in the analysis of subjects in the II ATM/CZA-ATM cohorts. The findings suggest that administration of CZA-ATM reduces ATM clearance but does not exacerbate AST/ALT elevations relative to ATM alone. The results also indicate that CI ATM should be used with caution.

INTRODUCTION

Metallo-β-lactamases (MBLs) are an emerging resistance determinant in many Gram-negative bacteria and are being detected globally at an alarming speed (1–3). Aztreonam (ATM) is one of the two commercially available β-lactams that are not hydrolyzed by MBLs, but many MBL-producing Gram-negative bacteria express β-lactamases that inactivate ATM. To circumvent this clinical conundrum, clinicians have coadministered ATM with ceftazidime-avibactam (CZA) to treat patients with MBL-producing Gram-negative infections with reported success (4–6). When paired with CZA, avibactam (AVI), a bicyclic diazabicyclooctane β-lactamase inhibitor, protects ATM from hydrolysis by a wide range of β-lactamase enzymes, including extended-spectrum β-lactamases (ESBLs) and Klebsiella pneumoniae carbapenemase (KPC) (7–12). In hollow fiber infection model (HFIM) studies of MBL-producing Escherichia coli and K. pneumoniae, simultaneous administration of ATM 8 g/day given as a continuous or 2-h infusion with CZA resulted in complete bacterial eradication and resistance suppression (13). Furthermore, the CZA-ATM combination regimens in these HFIM studies outperformed the ATM-AVI regimen currently in clinical development (13). While the mechanism of rapid and sustained bacterial killing with CZA-ATM combination regimens has not been established, it is likely due to maximal saturation of the diverse penicillin-binding proteins present in Gram-negative bacteria, flooding of periplasm with β-lactams, and inhibition of available β-lactamases (7, 14).

While the safety and pharmacokinetics (PK) of CZA and ATM administered alone have been well described, it was important to establish the safety and PK of the optimal CZA-ATM combination regimens identified in our HFIM study (13) in a controlled clinical trial prior to advocating for its use in patients with limited or no available treatment options. There are currently scant published PK data when these agents are administered concurrently (15, 16), and it is unknown if use of CZA-ATM in combination will lead to an altered PK profile of each agent due to inhibition of renal or other compensatory clearance mechanisms. Patients who receive ATM are at a greater risk for serum aminotransferase elevations relative to other β-lactams (17), and it is unclear if coadministration of CZA-ATM exacerbates the association between ATM and occurrence of liver enzyme elevations. Given these gaps in the literature, we conducted an open-label, phase 1 study of healthy subjects to evaluate the PK profile of CZA combined with ATM relative to its standalone counterparts and assess for the presence of an ATM exposure-alanine transaminase (ALT)/aspartate transaminase (AST) elevation associations when ATM is administered alone and in combination with CZA.

RESULTS

A total of 48 healthy adult male and female subjects enrolled in this study received CZA in combination with ATM, CZA alone, or ATM alone and had at least one quantifiable plasma or urine concentration of CAZ, AVI, or ATM measurement (PK population). Baseline demographics of subjects in the PK population are shown by cohort in Table 1. Median (range) age and weight were 32 years (22 to 45 years) and 74.3 kg (53.3 to 104.6 kg), respectively. Half (50%) of the subjects were female, and the median (range) baseline creatinine clearance (CL_CR) and serum creatinine (S_CR) were 118.34 mL/min (75.66 to 209.62 mL/min) and 0.9 mg/dL (0.5 to 1.4 mg/dL), respectively. A total of 19 subjects (40%) experienced an ALT and/or AST elevation of any relatedness (16 subjects had ALT and AST elevations and 3 subjects had an ALT or AST elevation). The maximum severity of ALT elevations was grade 3 in 3 subjects, grade 2 in 7 subjects and grade 1 in 7 subjects. The maximum severity of AST elevations was grade 3 in 2 subjects, grade 2 in 1 subject and grade 1 in 15 subjects. Of the 19 subjects with grade 1 or higher ALT/AST elevations, 17 received ATM alone or in combination with CZA. Two subjects in cohort 4 (ATM continuous infusion [CI] alone) experienced severe ALT/AST elevations, which halted the study. For subjects with ALT/AST elevations, ALT/AST values increased on or after study day 4, and the highest observed ALT/AST values were on study day 7 or 8. All ALT/AST elevations resolved without sequalae after cessation of dosing. Furthermore, all subjects who experienced ALT/AST elevations were asymptomatic, none had elevations in alkaline phosphatase or bilirubin, and there were no clinical findings suggestive of acute liver failure (i.e., liver necrosis, cholestasis, or jaundice).

TABLE 1.

Baseline demographic and clinical characteristics of subjects^a

Characteristic	Data for:b
	Cohort 1	Cohort 2	Cohort 3	Cohort 4	Cohort 5	Cohort 6
Subjects (N)	8	8	8	8	8	8
Female (%)	4 (50)	2 (25)	5 (62.5)	5 (62.5)	4 (50)	4 (50)
Ethnicity [N (%)]
Not Hispanic or Latino	8 (100)	7 (88)	5 (63)	7 (88)	7 (88)	6 (75)
Hispanic or Latino		1 (13)	3 (38)	1 (13)	1 (13)	2 (25)
Not reported
Unknown
Race [N (%)]
American Indian or Alaska Native
Asian			3 (38)		2 (25)	2 (25)
Native Hawaiian or other Pacific Islander
Black or African American	6 (75)	5 (63)	3 (38)	7 (88)	4 (50)	4 (50)
White	2 (25)	2 (25)	2 (25)	1 (13)	1 (13)	2 (25)
Multiracial		1 (13)			1 (13)
Unknown
Median (range) age (yrs)	40 (29–44)	38 (24–45)	34 (27–42)	35 (22–38)	28 (24–35)	32.5 (29–39)
Median (range) wt (kg)	75.6 (58.6–94.1)	85.8 (65.3–99.5)	67.6 (57.2–104.6)	72 (60.6–87.1)	71.6 (53.3–94.1)	78.1 (64.3–100.3)
Median (range) ht (cm)	174.1 (158.9–188)	177.5 (159.3–181.5)	163.5 (155.5–180.9)	168.5 (160.0–187.0)	164.4 163.0–178.4)	167.3 (146.5–186)
Median (range) BMI (kg/m2)	26.25 (21.5–29.9)	27.4 (20.7–31.8)	25.4 (21.4–32)	26 (19.3–30.7)	26.8 (20.1–29.6)	29.20 (22.6–33.7)
Median (range) LBW (kg)	56 (38.7–70.2)	63.9 (43.6–69.1)	45.7 (37.87–71.34)	43.5 (39.4–60.7)	45.8 (36.1–66.7)	51.3 (39.0–71.8)
Median (range) BSA (m2)	1.9 (1.6–2.2)	2.1 (1.8–2.2)	1.8 (1.6–2.2)	1.8 (1.7–2)	1.8 (1.6–2.1)	1.9 (1.6–2.3)
Median (range) baseline SCR (mg/dL)	0.9 (0.6–1.3)	1 (0.6–1.3)	0.8 (0.5–1.1)	0.9 (0.6–1.4)	0.8 (0.6–1.4)	0.8 (0.5–1.4)
Median (range) baseline CLCR (mL/min)	98.4 (92.5–136.3)	119.4 (88.2–153.6)	109.2 (96.5–163.0)	93.8 (87.0–105.9)	100.6 (86.8–128.5)	114.8 (86.7–171.8)
Median (range) baseline ALT (U/L)	17 (11–39)	17 (12–36)	22 (14–27)	18 (10–53)	14 (12–29)	18 (13–43)
Median (range) baseline AST (U/L)	26 (16–37)	24 (17–26)	20 (17–31)	19 (17–34)	20 (17–23)	19 (11–33)
Median (range) highest postbaseline ALT (U/L)	22 (14–146)	30 (17–57)	65 (23–212)	32 (23–620)	44 (22–106)	75 (22–166)
Median (range) highest postbaseline AST (U/L)	32 (19–85)	27 (20–52)	47 (24–151)	32 (23–489)	48 (20–63)	47 (24–101)

^aBMI, body mass index; LBM, lean body weight; BSA, body mass index; S_CR, serum creatinine; CL_CR, creatinine clearance; ALT, alanine aminotransferase; AST, aspartate aminotransferase.

^bDrug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5); CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h (cohort 6).

The subjects contributed a total of 3,475 PK samples (2,733 PK plasma samples and 742 PK urine samples) to the analysis for CAZ, AVI, and ATM. Of these, 206 of 3,475 (5.9%) plasma and urine concentration values were below the quantifiable limit (BQL) and excluded, of which 2 of 1,160 (0.2%) were for CAZ, 18 of 1,160 (1.6%) were for AVI, and 186 of 1,155 (16.1%) were for ATM. The reasons for the higher proportion of BQL for ATM were the occurrence of a study halting rule (day 7 plasma PK concentrations for 8 subjects in cohort 4), voluntary study withdrawal (day 7 plasma PK concentrations for 1 subject in cohort 5), and discordant dosing-plasma concentration data (day 7 plasma PK concentrations for 3 subjects in cohort 5). In cohort 4, all subjects on day 6 stopped dosing due to a study halting rule, and their day 7 plasma PK concentrations were excluded, and no PK parameters were estimated on day 7. In cohort 5, 1 subject voluntarily withdrew from the study on day 6. For 3 other subjects in cohort 5, all day 7 CAZ, AVI, and ATM PK plasma data were excluded due to discordance between ATM dosing and observed plasma drug concentration data. For these 3 subjects, all plasma ATM concentrations after dosing were BQL, whereas the electronic case report form indicates the ATM dose was given. Day 7 ATM BQL concentrations for these subjects were excluded due to being classified as outliers. Due to the discordant ATM dosing-plasma concentration data, 45 CAZ and 45 AVI day 7 plasma samples were also treated as outliers and excluded from the noncompartmental (NCA) and population PK (PopPK) analyses (Table 2). Several urine PK data were also excluded from the urine PK analysis due to a missing urine collection interval (Table 3).

TABLE 2.

Summary of key plasma PK parameters in noncompartmental analysis by day, study drug, and cohort^a

Drug and cohortb	Statistic summary	Day 1			Day 7
		Cmax mg/L	Cmin (Css) (mg/L)	AUC0–tau (mg*h/L)	Cmax (mg/L)	AUC0–tau (mg*h/L)	AUCinf (mg*h/L)	t1/2 (h)	Kel (h-1)	Vss (L)	CL (L/h)	RAUC0–tau
Ceftazidime
Cohort 1	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	66.4	5.4	221.0	62.5	216.0	234.0	2.6	0.27	19.6	8.5	0.98
	Geometric CV (%)	24.5	23.0	16.4	19.2	12.20	11.1	4.9	4.9	19.5	11.1	13.1
Cohort 2	N	8	8	NR	8	8	8	8	8	8	8	NR
	Geometric mean	70.3	23.3	NR	31.2	223.0	302.0	2.6	0.27	12.2	6.7	NR
	Geometric CV (%)	17.3	90.4	NR	22.1	20.8	21.4	21.2	21.2	24.9	22.3	NR
Cohort 5c	N	8	8	8	4	4	4	4	4	4	4	4
	Geometric mean	81.0	5.9	253.0	87.5	260.0	276.0	2.6	0.27	14.2	7.2	0.95
	Geometric CV (%)	21.7	35.2	13.8	13.6	10.8	9.7	4.3	4.3	22.2	9.7	4.8
Cohort 6	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	69.2	6.1	239.0	69.4	228.0	247.0	2.6	0.27	18.8	8.1	0.95
	Geometric CV (%)	19.8	27.0	15.0	12.1	12.4	11.5	9.3	9.3	21.0	11.5	5.2
Avibactam
Cohort 1	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	12.2	0.48	35.4	10.5	30.6	31.9	2.5	0.27	25.5	15.7	0.87
	Geometric CV (%)	21.6	41.6	16.2	22.9	12.1	11.7	5.8	5.8	20.6	11.7	14.5
Cohort 2d	N	8	8	NR	8	8	7	7	7	7	7	NR
	Geometric mean	12.9	3.5	NR	4.5	32.8	41.8	2.2	0.31	16.9	12.0	NR
	Geometric CV (%)	17.5	113.0	NR	17.7	17.8	17.7	12.5	12.5	26.3	18.6	NR
Cohort 5c	N	8	8	8	4	4	4	4	4	4	4	4
	Geometric mean	15.5	0.68	43.4	14.1	37.5	38.9	2.5	0.28	18.1	12.9	0.8
	Geometric CV (%)	19.2	44.2	17.2	12.1	12.0	12.0	6.1	6.1	26.5	12.0	3.8
Cohort 6	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	14.0	0.67	41.8	12.7	34.0	35.6	2.4	0.28	23.6	14.0	0.8
	Geometric CV (%)	22.1	40.0	18.7	16.8	18.7	18.6	8.6	8.6	26.1	18.6	7.5
Aztreonam
Cohort 3	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	88.5	13.7	264.0	77.9	236.0	261.0	1.6	0.44	14.2	7.7	0.89
	Geometric CV (%)	12.6	20.9	14.0	12.0	13.2	14.4	12.1	12.1	12.4	14.4	8.2
Cohort 4	N	8	8	NR	NR	NR	NR	NR	NR	NR	NR	NR
	Geometric mean	93.7	58.5	NR	NR	NR	NR	NR	NR	NR	NR	NR
	Geometric CV (%)	5.5	11.0	NR	NR	NR	NR	NR	NR	NR	NR	NR
Cohort 5c	N	8	8	8	4	4	4	4	4	4	4	4
	Geometric mean	70.2	13.1	219.0	74.0	221.0	244.0	1.6	0.42	11.4	6.2	0.91
	Geometric CV (%)	17.4	16.3	14.7	11.7	11.1	9.42	10.7	10.7	19.9	9.4	3.9
Cohort 6	N	8	8	8	8	8	8	8	8	8	8	8
	Geometric mean	91.9	18.5	292.0	89.0	275.0	313.0	1.8	0.39	13.5	6.4	0.94
	Geometric CV (%)	18.2	21.0	15.7	18.6	17.0	16.8	17.5	17.5	21.3	16.8	4.9

^aC_max, maximum concentration; C_min, trough concentration at steady state (present for cohorts 1, 3, 5, and 6); C_ss, plasma concentration at steady state (presented for cohorts 2 and 4); AUC_0–tau, AUC from time zero to tau (tau, 8 hours for ceftazidime and avibactam and 6 hours for aztreonam) after the dose; AUC_inf, AUC from time of dosing extrapolated to infinity on day 7; CL, total body clearance; V_ss, volume of distribution; t_1/2, terminal phase disposition half-life; K_el, terminal elimination rate constant; RAUC: accumulation ratio for day 1 AUC_0-tau/day 7 AUC_0-tau; NR, not reported.

^bDrug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5); CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h (cohort 6).

^cN = 4 due to 4 subjects missing aztreonam doses on day 7 in cohort 5; all day 7 plasma PK data for ATM were excluded from the PK analysis. All subjects in cohort 4 on day 7 for ATM were excluded from the PK analysis due to stopped dosing.

^dN = 7 due to 1 subject’s K_el being not in acceptance criteria (R² adjusted < 0.8).

TABLE 3.

Summary of key urine PK parameters in noncompartmental analysis by day, study drug and cohort^a

Drug and cohortb	Statistic summary	Day 1					Day 7
		Ae0–t (mg)	Ae0–24 (mg)	Fe0–t	Fe0–24	CLR (L/h)	Ae0–t (mg)	Ae0–24 (mg)	Fe0–t	Fe0–24
Ceftazidime
Cohort 1	N	8	8	8	8	8	7	6	7	6
	Geometric mean	1,230	3,560	0.62	0.59	5.58	1,290	3,920	0.64	0.65
	Geometric CV (%)	18.20	13.90	18.30	13.90	22.60	31.60	27.50	31.60	27.50
Cohort 2c	N	7	7	7	7	7	8	8	8	8
	Geometric mean	2,110	4,690	0.53	0.59	6.77	1370	4,250	0.69	0.71
	Geometric CV (%)	14.60	15.10	14.30	15.00	22.00	31.20	18.00	31.20	18.30
Cohort 5	N	8	8	8	8	8	6	6	6	6
	Geometric mean	1,780	5,540	0.89	0.92	7.03	2,370	6,010	1.18	1.00
	Geometric CV (%)	7.77	4.79	7.77	4.79	14.60	32.30	17.60	32.30	17.60
Cohort 6c	N	7	7	7	7	7	8	8	8	8
	Geometric mean	2,120	6,070	1.06	1.01	9.07	1,770	5,820	0.89	0.97
	Geometric CV (%)	27.00	14.60	27.00	14.60	29.20	27.90	10.20	27.90	10.20
Avibactam
Cohort 1c	N	8	8	8	8	8	7	6	7	6
	Geometric mean	383	1,110	0.78	0.74	10.8	363	1,160	0.73	0.78
	Geometric CV (%)	15.6	12.5	15.3	12.5	25.2	43.30	33.2	43.3	33.2
Cohort 2c	N	7	7	7	7	7	8	8	8	8
	Geometric mean	599	1,300	0.60	0.65	11.4	349	1,130	0.70	0.76
	Geometric CV (%)	17.3	18.3	17.3	18.4	20.7	30.70	18.2	30.7	18.6
Cohort 5c	N	8	8	8	8	8	6	6	6	6
	Geometric mean	488	1,500	0.98	1.00	11.2	615	1620	1.23	1.08
	Geometric CV (%)	8.7	5.1	8.7	5.1	18.90	35.8	17.7	35.8	17.7
Cohort 6c	N	7	7	7	7	7	8	8	8	8
	Geometric mean	582	1,630	1.16	1.09	14.0	444	1,530	0.89	1.02
	Geometric CV (%)	25.1	13.3	25.1	13.3	33.2	31	8.8	31.0	8.8
Aztreonam
Cohort 3	N	8	7	8	7	8	8	7	8	7
	Geometric mean	924	4,630	0.46	0.58	4.2	1,100	4,540	0.55	0.57
	Geometric CV (%)	21.0	22.0	21.0	22.0	24.1	15.4	13.7	15.4	13.7
Cohort 4	N	8	8	8	8	8	5	4	5	4
	Geometric mean	1,060	4,570	0.39	0.46	3.8	1.040	3,200	0.78	0.85
	Geometric CV (%)	28.6	27.2	28.1	27.2	26.6	36.0	8.1	35.9	10.7
Cohort 5	N	8	8	8	8	8	7	6	7	6
	Geometric mean	790	4,530	0.53	0.75	4.4	1,250	3,340	0.83	0.56
	Geometric CV (%)	34.8	11.4	34.8	11.4	37.4	43.0	60.9	43.0	60.9
Cohort 6	N	7	7	7	7	7	8	8	8	8
	Geometric mean	1,050	6,210	0.53	0.78	4.4	1,110	6,250	0.55	0.78
	Geometric CV (%)	13.4	12.2	13.4	12.2	23.7	51.3	10.2	51.3	10.2

^aAe_0–t, Cumulative amount excreted into the urine from time of dosing up to time t (t, 8 hours for ceftazidime and avibactam and 4 hours for aztreonam); Fe_0–t, the fraction of dose excreted in urine from the time of dosing to time t; CL_R, ceftazidime and avibactam, calculated as CL_R = Ae_0–8/AUC_0–8 using the same time interval for urine amount and AUC in plasma; aztreonam, calculated as CL_R = Ae_0–4/AUC_0–4 using the same time interval for urine amount and AUC in plasma.

^bDrug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5); CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2-h every 6 h (cohort 6).

^cN < 8 due to missing the urine collection interval, and the subjects were excluded from the urine PK analysis. For AVI and CAZ, the subjects missing urine collection interval 0 to 4 h and 4 to 8 h were excluded from the urine PK analysis, which included one subject in cohort 1 missing the day 6 Ae_0–4, one subject in cohort 2 missing the day 1 Ae_0–4, one subject in cohort 5 missing the day 6 Ae_4–8, another subject in cohort 5 missing the day 6 urine collection data due to study withdrawal, and one subject in cohort 6 missing the day 1 Ae_0–4. One subject in cohort 1 was excluded only for the Ae_0–24 calculation on day 6 due to missing the 8 to 12 h collection interval. For ATM, the subjects missing the urine collection interval 0 to 4 h included 2 subjects in cohort 4 missing the day 6 Ae_0–4, one subject in cohort 4 with no day 6 urine collection due to discordant dosing on day 6, and one subject in cohort 6 with a missing day 1 Ae_0–4h. Additionally, one subject in cohort 3 had missing day 1 and day 6 Ae_4–8, one subject in cohort 4 had a missing Ae_8–12 on day 6, and one subject in cohort 5 had a missing Ae_4–8 on day 6, and all were excluded only for the Ae_0–24 calculation.

Noncompartmental PK analyses.

The plasma mean concentration versus time plots of CAZ, AVI, and ATM on study day 1 and day 7 stratified by cohort are shown in Fig. 1. Following the first single-infusion doses of CZA in combination with ATM, CZA alone, or ATM alone for cohorts 1, 3, 5, and 6, plasma concentrations of CAZ, AVI, and ATM increased rapidly and declined in a multiphasic manner with plasma concentrations measurable, on average, up to 24 h postdose on day 7. Following first loading infusion doses and then CI dosing of CZA alone, or ATM alone for cohorts 2 and 4, plasma concentrations of CAZ, AVI, and ATM increased rapidly on day 1, maintained the steady state plasma concentration for 7 days, and then declined in a multiphasic manner with plasma concentrations measurable, on average, up to 24 h post dose on day 7 after the last dose.

FIG 1.

Mean (standard deviation) plasma concentration-time profile after first dose on day 1 and after last dose on day 7 for ceftazidime, avibactam, and aztreonam. (A) Ceftazidime day 1 plasma profile; (B) ceftazidime day 7 plasma profile; (C) avibactam day 1 plasma profile; (D) avibactam day 7 plasma profile; (E) aztreonam day 1 plasma profile; (F) aztreonam day 7 plasma profile. Drug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5); and CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h (cohort 6).

The PK parameters for CAZ, AVI, and ATM in plasma on day 1 and day 7 are shown in Table 2. For CAZ, the maximal concentration (C_max) and area under the curve from 0 to 8 h (AUC_0–8) on day 7 for cohorts 1, 5, and 6 were similar to the day 1 C_max and AUC_0–8, and the geometric mean accumulation ratio for the AUC over one dosing interval (RAUC_0–Tau) ranged from 0.95 to 0.98 across cohorts 1, 5, and 6. The geometric mean half-life (t_1/2) was 2.6 h across all cohorts. The geometric means (coefficient of variation [CV]) of clearance (CL) on day 7 were 8.5 (11.1%) L/h and 6.7 (22.3%) L/h for cohorts 1 and 2 (CZA alone), respectively, and 7.2 (9.7%) L/h, and 8.1 (11.5%) L/h for cohorts 5 and 6 (CZA-ATM), respectively. For AVI, the C_max and AUC_0–8 on day 7 for cohorts 1, 5, and 6 were similar to the C_max and AUC_0–8 on day 1, and the geometric mean RAUC_0–8 ranged from 0.80 to 0.87 across cohorts 1, 5, and 6. The geometric mean t_1/2 was 2.2 to 2.5 h across the cohorts. The geometric means (CV) of CL on day 7 were 15.7 (11.7%) L/h and 12.0 (18.6%) L/h for cohorts 1 and 2 (CZA alone), respectively, and 12.9 (12.0%) L/h and 14.0 (18.6%) L/h for cohorts 5 and 6 (CZA-ATM), respectively. For ATM, the C_max and area under the curve from 0 to 6 h (AUC_0–6h) on day 7 for cohort 3, 5, and 6 were similar to the C_max and AUC_0–6 on day 1 and the geometric mean RAUC_0–6 ranged from 0.89 to 0.94 across cohorts 3, 5, and 6. The geometric mean t_1/2 was 1.6 to 1.8 h across cohorts 3, 5, and 6. The geometric means (CV) of CL on day 7 were 7.7 (14.4%) L/h for cohort 3 (ATM alone) and 6.2 (9.4%) L/h and 6.4 (16.8) L/h for cohorts 5 and 6 (CZA-ATM), respectively.

The PK parameters in urine on day 1 and day 6 for CAZ, AVI, and ATM are summarized in Table 3. For CAZ and AVI, the fraction of drug excreted was higher for the CZA-ATM cohorts (cohorts 5 and 6) relative to the CAZ-AVI-alone cohorts on day 1 and day 6. However, some subjects in cohorts 5 and 6 displayed a fraction of dose excreted in urine (Fe) above 100%, potentially due to errors in the dosing-concentration record or urine collection. For ATM, the extent of urinary elimination of ATM was also higher in the combination cohorts than in the ATM-alone cohorts on day 1 and day 6. There were modest increases in the mean renal clearance (CL_R) of ATM observed when combined with CZA compared to ATM alone. However, the results should be interpreted cautiously since CL_R was calculated only after a 4-h interval, at which point the full amount of unchanged drug is not recoverable from the urine.

Population PK analysis.

(i) Ceftazidime. Based on the comparison of objective function values (OFV) for nested models or Akaike’s information criterion (AIC) for nonnested models, a 2-compartment model with a combined additive and proportional error model in plasma and an additive error in urine was selected as the structural model for CAZ (Appendix SA in the supplemental material). The precision of PK parameters for the base model was overall good (relative standard error [RSE], 3.5 to 18.1%), and the residual proportional error was 16.2% for plasma (Table S1). The deviations from the population-typical value PK parameters (ETA) versus covariate plots suggested that ATM coadministration was a significant covariate on CL_R and nonrenal clearance (CL_NR), lean body weight (LBW) was a significant covariate on CL_R, apparent volume of the central compartment (V₁), intercompartmental clearance (Q), and apparent volume of peripheral compartment (V₂); S_CR was a significant covariate on CL_R; and CI was a significant covariate on Q. After the covariate step (appendix SA), the final PopPK model for CAZ included LBW as a statistically significant covariate for CL_R, V₁, Q, and V₂; S_CR as a statistically significant covariate for CL_R; CZA-ATM coadministration as a statistically significant covariate for CL_R and CL_NR; and CI administration as a statistically significant covariate for Q (Table 4).

TABLE 4.

Final population PK models for ceftazidime, avibactam, and aztreonam^a

Final ceftazidime PopPK model			Final avibactam PopPK model			Final aztreonam PopPK model
Parameter	Estimate	RSE (%)	Parameter	Estimate	RSE (%)	Parameter	Estimate	RSE (%)
CLR (L/h)	5.69	3.8	CLR (L/h)	12.20	2.7	CLR (L/h)	4.79	3.3
V1 (L)	10.50	4.0	V1 (L)	12.40	4.5	V1 (L)	3.77	37.9
Q (L/h)	6.93	10.3	Q (L/h)	7.81	9.3	Q (L/h)	28.7	24.3
V2 (L)	8.40	4.0	V2 (L)	9.06	4.9	V2 (L)	9.87	13.7
CLNR (L/h)	2.89	8.6	CLNR (L/h)	1.39	25.5%	CLNR (L/h)	2.66	5.8%
θ(CLR-LBW)	0.0146	17.5	θ(V1-LBW)	0.0194	22.9%	θ(CLR-CLcr)	0.622	20.4%
θ(V1-LBW)	0.0166	22.8	θ(Q-LBW)	0.021	42.2%	θ(V1-BSA)	4.11	34.5%
θ(Q-LBW)	1.42	32.5	θ(V2-LBW)	0.0214	23.0%	θ(CLNR-CZA/ATM)	–0.391	14.0%
θ(V2-LBW)	0.0254	1.01	θ(CLR-CLcr)	0.00453	21.0%
θ(CLNR-CZA/ATM)	–0.905	6.3	θ(Q-CI)	–0.407	20.3%
θ(CLR-CZA/ATM)	1.39	4.9
θ(CLR-SCR)	–0.494	25.3
θ(Q-CI)	–0.396	21.8

^aV₁, the volume of the central compartment; Q, intercompartmental clearance; V₂, volume of the peripheral compartment; CL_R, renal clearance; CL_NR, nonrenal clearance; RSE, relative standard error; IIV/BSV, interindividual variability/between-subject variability; CV, coefficient of variation (%); BMI, body mass index; LBM, lean body weight; BSA, body mass index; SCR, serum creatinine; CLCR, creatinine clearance; CZA/ATM, coadministration of ceftazidime/avibactam with aztreonam; CI, continuous infusion.

^bIIV/BSV V2 was fixed to 0 in the final model; the parameter estimate was near its boundary with high shrinkage of 99%.

^cIIV/BSV V1 was fixed to 0 in the final model, as the parameter estimate was near its boundary (shrinkage, 99%).

The final CAZ PopPK model was as follows:

$${\text{CL}}_{\text{R}}\left(\text{L/h}\right)=5.69\cdot e^{0.0146\cdot \left(\text{LBW}-52\right)}\cdot {\left(\frac{\mathit{SCR}}{0.9}\right)}^{-0.494}\cdot {1.39}^{\text{concomitant CZA}/\text{ATM}}$$

$$V_1(\text{L})=10.5\cdot e^{0.0166\cdot (\text{LBW}-52)}$$

$$Q\left(\text{L/h}\right)=6.93\cdot {\left(\frac{\text{LBW}}{52}\right)}^{1.42}\cdot (1\hspace{0.17em}+\hspace{0.17em}\mathit{CI}\cdot (-0.396))$$

$$V_2\hspace{0.17em}(\text{L})=8.4\hspace{0.17em}\cdot \hspace{0.17em}{e}^{0.0254\cdot (\text{LBW}-52)}$$

$${\text{CL}}_{\text{NR}}\left(\text{L/h}\right)=2.89\cdot (1\hspace{0.17em}+\hspace{0.17em}\text{concomitant CZA}/\text{ATM}\cdot (-0.905))$$

(ii) Avibactam. A 2-compartment model with a combined additive and proportional error model in plasma and a combined additive and proportional error in urine was selected as the structural model for AVI based on the comparison of OFV for nested models or AIC for nonnested models (Appendix SA). The precision of PK parameters for the base model was overall good (RSE, 3.3 to 24.8%), and the residual proportional error was 19.7% for plasma and 21.3% for the urine (Table S2). The ETAs versus covariate plots suggested that CL_CR was a significant covariate on CL_R; LBW was a significant covariate on V₁, Q, and V₂; and CI was a significant covariate on Q. After the covariate step (Appendix SA), the final AVI PopPK model included the LBW as a statistically significant covariate for V₁, V₂, and Q; CL_CR for as a statistically significant covariate CL_R; and CI administration as a statistically significant covariate for Q (Table 4).

The final AVI PopPK model was as follows:

$${\text{CL}}_{\text{R}}\left(\text{L/h}\right)=12.2\cdot e^{0.0045\cdot ({\text{CL}}_{\text{CR}}-118.69)}$$

$$V_1(L)=12.4\cdot e^{0.0194\cdot (\text{LBW}-52)}$$

$$Q\left(\text{L/h}\right)=7.81\cdot e^{0.021\cdot \left(\text{LBW}-52\right)}\text{\hspace{0.17em}*}(1\hspace{0.17em}+\hspace{0.17em}\text{CI}\cdot (-0.407))$$

$$V_2(L)=9.06\cdot e^{0.0214\cdot (\text{LBW}-52)}$$

$${\text{CL}}_{\text{NR}}\left(\text{L/h}\right)=1.39$$

(iii) Aztreonam. A 2-compartment model with proportional error model in plasma and proportional error in urine was selected as the structural model for ATM (Appendix SA). Given that 16.1% of plasma samples were BQL/outliers, the M1 (discard all BQL samples) and M3 (likelihood-based) methods were used to evaluate model performance. The parameter estimates and model predictions of the 2-compartment with proportional error models were comparable between the M1 and M3 methods. The M1 method was selected due to better model stability with the lowest OFV (3,017.7) and AIC (3,033.7) compared with the M3 method OFV (3,128.1) and AIC (3,142.1). The precision of PK parameters for the base model was good overall (RSE, 3.3 to 45.9%), and the residual proportional error was 15.6% for plasma and 37.4% for urine (Table S3). The ETA versus covariate plots suggested that CL_CR was a significant covariate on CL_R, body surface area (BSA) was a significant covariate on V₁, and CZA-ATM administration was a significant covariate on CL_NR. After the covariate step (Appendix SA), the final ATM PopPK model included the CL_CR as a statistically significant covariate for CL_R, BSA as a statistically significant covariate for V₁, and CZA-ATM administration as a statistically significant covariate for CL_NR (Table 4).

The final ATM PopPK model was as follows:

$${\text{C}}_{\text{LR}}\left(\text{L/h}\right)=4.79\cdot {\left(\frac{{\text{CR}}_{\text{CL}}}{118.69}\right)}^{0.622}$$

$$V_1\left(\text{L}\right)=3.77\cdot {\left(\frac{\mathit{BSA}}{1.85}\right)}^{4.11}$$

$$Q\left(\text{L/h}\right)=28.7$$

$$V_2(\text{L})=9.87$$

$${\text{CL}}_{\text{NR}}\left(\text{L/h}\right)=2.66\cdot \left(1\hspace{0.17em}+\hspace{0.17em}\text{concomitant CZA}/\text{ATM}\cdot \left(-0.391\right)\right)$$

(iv) Model diagnostics. Diagnostic plots of the final PopPK models for plasma and urine show overall good model fit without misspecification for all agents (Fig. 2A to C). The prediction-corrected visual predictive checks (pcVPCs) for CAZ, AVI, and ATM indicated that the final model described plasma and urine PK data adequately (Fig. 3A to C). Plots of the observed and individual post hoc predictions for each subject for each drug further demonstrate the accuracy and precision of the PK models (Appendix SB). The relationship between the parameter estimation with interaction covariance matrix of the estimate is presented in Appendix SA, which indicates that the final model parameter estimation was successful. The post hoc PK parameters for all PopPK are summarized in Table 5, and the PK exposure parameters calculated using simulated concentration data based on empirical Bayesian estimates (EBE) and individualized dosing histories are summarized in Table 6.

FIG 2.

(A to C) Population PK goodness of fit plots for the final models in plasma and in urine for ceftazidime (A), avibactam (B), and aztreonam (C). In plasma: observed versus population (A1) and individual (B1) predictions; conditional weighted residuals versus population predictions (C1) and time after first dose (D1). In urine: observed versus population (A2) and individual (B2) predictions; conditional weighted residuals versus population predictions (C2) and time after first dose (D2).

FIG 3.

Prediction-corrected visual predictive check of observation versus time after dose (TALD) by matrix for ceftazidime, avibactam, and aztreonam. (A) Ceftazidime in plasma; (B) ceftazidime in urine; (C) avibactam in plasma; (D) avibactam in urine; (E) aztreonam in plasma; (F) aztreonam in urine.

TABLE 5.

Mean (standard deviation) post hoc individual PK parameters from final population PK models by study drug and cohort^a

Drug and cohortb	CL (L/h)	CLR (L/h)	CLNR (L/h)	V1 (L)	Q (L/h)	V2 (L)
Ceftazidime
Cohort 1	8.70 (0.95)	5.62 (0.89)	3.07 (0.29)	11.40 (2.34)	7.08 (2.84)	8.84 (2.72)
Cohort 2	8.94 (1.68)	6.07 (0.97)	2.87 (0.89)	11.80 (2.24)	4.93 (1.47)	9.90 (2.31)
Cohort 5	7.70 (0.86)	7.42 (0.85)	0.28 (0.01)	9.72 (1.93)	6.40 (2.12)	8.31 (2.28)
Cohort 6	8.19 (1.02)	7.92 (1.02)	0.27 (0.01)	10.90 (2.17)	7.50 (2.38)	8.64 (2.40)
Avibactam
Cohort 1	14.60 (1.48)	11.60 (1.15)	2.97 (1.29)	13.40 (3.08)	8.09 (2.06)	9.45 (2.42)
Cohort 2	14.80 (2.36)	12.00 (1.35)	2.80 (1.56)	14.30 (2.97)	5.29 (1.03)	10.40 (2.10)
Cohort 5	12.30 (1.41)	11.30 (1.16)	0.97 (0.46)	11.60 (2.72)	7.70 (1.75)	8.91 (2.04)
Cohort 6	13.20 (2.05)	12.20 (1.87)	0.96 (0.29)	12.60 (2.80)	7.96 (1.80)	9.22 (2.16)
Aztreonam
Cohort 3	7.41 (0.99)	4.70 (0.78)	2.71 (0.28)	3.59 (2.06)	28.70	10.30 (0.82)
Cohort 4	6.70 (0.29)	4.06 (0.28)	2.64 (0.09)	3.70 (0.92)	28.70	9.90 (1.29)
Cohort 5	6.30 (0.69)	4.67 (0.66)	1.63 (0.13)	3.76 (1.50)	28.70	9.77 (1.25)
Cohort 6	6.42 (0.98)	4.80 (0.93)	1.62 (0.08)	4.25 (1.93)	28.70	9.75 (1.12)

^aCL, total clearance; V₁, volume of the central compartment; Q, intercompartmental clearance; V₂, volume of the peripheral compartment; CL_R, renal clearance; CL_NR, nonrenal clearance.

^bDrug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5; CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h (cohort 6).

TABLE 6.

Geometric mean (geometric coefficient of variation) post hoc individual exposure estimate from final population PK models by study drug and cohort

Drug and cohorta	Cmax,Day1 (mg/L)	Cmin,Day1 (Css,Day1) (mg/L)	AUC0–24 (mg*h/L)	Cmax,Day3 (mg/L)	Cmin,Day3 (css,Day1) (mg/L)	AUC48–72 (mg*h/L)	AUC0–72 (mg*h/L)
Ceftazidime
Cohort 1	69.3 (14.9)	5.7 (17.0)	687 (9.0)	76.4 (14.8)	5.7 (18.5)	706 (8.2)	2100 (8.4)
Cohort 2b	68.8 (10.1)	32.1 (8.8)	762 (6.8)	27.8 (6.4)	27.8 (6.4)	661 (7.7)	2090 (6.9)
Cohort 5	75.7 (16.0)	6.34 (24.3)	752 (9.3)	80.3 (13.7)	7.2 (28.8)	777 (9.1)	2310 (9.1)
Cohort 6	71.5 (13.9)	6.06 (22.0)	708 (14.6)	75.2 (15.6)	6.6 (25.6)	732 (14.6)	2170 (14.5)
Avibactam
Cohort 1	12.7 (11.9)	0.6 (35.9)	112 (7.0)	13.7 (11.8)	0.5 (33.4)	114 (6.9)	340 (6.9)
Cohort 2b	12.4 (8.5)	5.1 (10.8)	124 (8.7)	4.4 (8.3)	4.4 (8.4)	105 (9.6)	334 (8.8)
Cohort 5	13.0 (11.3)	0.5 (55.4)	112 (8.9)	13.4 (9.4)	0.5 (56.7)	113 (9.4)	338 (9.2)
Cohort 6	12.3 (11.6)	0.5 (35.8)	105 (12.4)	12.6 (13.3)	0.5 (39.0)	106 (12.5)	316 (12.4)
Aztreonam
Cohort 3	86.6 (9.6)	11.2 (10.3)	1030 (9.0)	90.8 (10.3)	12 (11.0)	1090 (7.9)	3230 (7.9)
Cohort 4b	91.6 (4.4)	52 (5.7)	1250 (3.9)	50.6 (4.0)	47.8 (4.3)	1140 (6.2)	3590 (3.4)
Cohort 5	71.6 (9.4)	11.6 (12.4)	957 (7.7)	76.9 (7.9)	14.5 (22.2)	995 (8.3)	2950 (8.1)
Cohort 6	90.9 (10.5)	15 (16.8)	1210 (10.7)	99.5 (11.0)	17.3 (20.7)	1250 (11.4)	3710 (11.1)

^aDrug cohorts: CZA 2.5 g i.v. over 2 h every 8 h (cohort 1); CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a continuous infusion (CI) (cohort 2); ATM 2 g i.v. over 2 h every 6 h (cohort 3); ATM 2 g i.v. × 1 and then 8 g/daily as a CI (cohort 4); CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h (cohort 5); CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h (cohort 6).

^bC_min values reflected the C_ss concentrations in the continuous infusion cohorts (cohorts 2 and 4).

ATM plasma exposure and highest observed ALT/AST response analyses.

The relationships between day 1 and 3 ATM exposure parameters and the highest observed ALT/AST values are graphically displayed in Fig. S1 to S4. No notable associations were identified in the overall locally weighted scatterplot smoothing (LOWESS) analyses. In the analysis of subjects who received ATM as 2-h intermittent infusions, the LOWESS curves suggested that there were potential ATM exposure-ALT/AST relationships for the day 1 maximum simulated concentration at steady state (C_max,ss), day 3 C_max,ss, AUC_0–24, and AUC_48–72. In the curvilinear regression analyses that included only subjects who received 2-h intermittent infusion of ATM, significant associations between the day 1 C_max,ss, day 3 C_max,ss, AUC_0–24, and AUC_48–72, expressed as power functions, and the highest observed ALT/AST values were observed (Fig. 4 and 5). In the general linear model (GLM) analysis (log-link function with a GAUSSIAN distribution) of the ATM 2-h intermittent infusion cohorts, significant associations were observed between the ATM day 1 C_max,ss, day 3 C_max,ss, AUC_0–24, and AUC_48–72 and the highest observed ALT/AST values (Fig. 6 and 7 and Table S4). Coadministration of CZA-ATM was not associated with ALT/AST in the overall GLM analyses (ALT: exponentiated [exp] beta-coefficient of 0.51, 95% confidence interval [CI] of 0.16 to 1.63, P = 0.26; AST: exp beta-coefficient of 0.50, 95% CI of 0.16 to 1.59, P = 0.24) and 2-h intermittent infusion ATM cohort analyses (ALT: exp beta-coefficient of 0.78, 95% CI of 0.44 to 1.41, P = 0.41; AST: exp beta-coefficient of 0.82, 95% CI of 0.50 to 1.35, P = 0.44).

FIG 4.

Curvilinear associations between day 1 aztreonam exposures, expressed as power functions, and highest observed ALT/AST values in the regression analyses among subjects who received 2-h intermittent infusions of aztreonam. (A) ALT versus AUC_0–24; (B) AST versus AUC_0–24; (C) ALT versus C_max; (D) AST versus C_max; (E) ALT versus C_min; (F) AST versus C_min.

FIG 5.

Curvilinear associations between day 3 aztreonam exposures, expressed as power functions and highest observed ALT/AST Values, in the regression analyses among subjects who received 2-h intermittent infusions of aztreonam. (A) ALT versus AUC_48–72; (B) AST versus AUC_48–72; (C) ALT versus C_max; (D) AST versus C_max; (E) ALT versus C_min; (F) AST versus C_min. ALT, alanine aminotransferase; AST, aspartate aminotransferase; ATM, aztreonam; AUC, area under the curve; D1, day 1.

FIG 6.

Log linear associations between day 1 aztreonam exposures and highest observed ALT/AST values in the generalized linear model analyses among subjects who received 2-h intermittent infusions of aztreonam. (A) ALT versus AUC_0–24; (B) AST versus AUC_0–24; (C) ALT versus C_max; (D) AST versus C_max; (E) ALT versus C_min; (F) AST versus C_min. ALT, alanine aminotransferase; AST, aspartate aminotransferase; ATM, aztreonam; AUC, area under the curve; D1, day 1.

FIG 7.

Log linear associations between day 3 aztreonam exposures and highest observed ALT/AST values in the generalized linear model analyses among subjects who received 2-h intermittent infusions of aztreonam. (A) ALT versus AUC_48–72; (B) AST versus AUC_48–72; (C) ALT versus C_max; (D) AST versus C_max; (E) ALT versus C_min; (F) AST versus C_min. ALT, alanine aminotransferase; AST, aspartate aminotransferase; ATM, aztreonam; AUC, area under the curve; D3, day 3.

DISCUSSION

The plasma PK parameters and exposure profiles of CAZ, AVI, and ATM in the present study were highly consistent with those observed in previous phase 1 healthy subject clinical trials of CZA and ATM (15, 16, 18–24). In this phase 1 study, all drugs exhibited linear PK, best described by a 2-compartment model linked to a separate compartment for urine elimination, and there was no appreciable accumulation observed with multiple doses. Consistent with previous CAZ, AVI, and ATM PopPK models, a body size descriptor (i.e., LBW and BSA) was found to be a significant covariate on V₁, and an estimate of renal function (i.e., CL_CR and a combination of S_CR and LBW) was a significant covariate on CL_R. All drugs are primarily eliminated through urinary excretion, and it was reasonable that CL_CR and S_CR accounted for a large percentage of interindividual variability (IIV) in CL_R. The observed modest relationships for all drugs between body size descriptors and V₁ were physiologically consistent with drugs that are largely distributed in extracellular fluid (25, 26).

The plasma PK parameters and exposure estimates of CAZ, AVI, and ATM in each cohort were similar on days 1 and 7, but some variations in PK parameter and exposure estimates were noted between the single and combination study drug cohorts. While this may have been a reflection of the typical between-subject PK variability observed in phase 1 studies (27, 28), the day 1 and 7 plasma ATM AUCs, when dose normalized, were ~15% higher, and the day 7 ATM CL estimates were ~15% lower in the CZA-ATM cohorts than in the ATM-alone cohorts in the NCA analyses. Coadministration of CZA-ATM was found to reduce the CL_NR of ATM by 1 L/h (16% of overall CL) in the final PopPK ATM model, resulting in higher daily post hoc individual ATM AUC estimates. Although the mechanism(s) for altered ATM CL with concurrent CZA-ATM administration is unclear and merits further investigation, ATM is known to undergo fecal excretion (12%), and approximately 6 to 16% is metabolized to inactive metabolites through hydrolysis, and it is possible that coadministration with CZA may affect one of these nonrenal clearance mechanisms (15).

In contrast to the reductions in ATM CL with CZA-ATM coadministration, the urine NCA analyses suggested that CAZ-ATM coadministration may result in increased CL_R of CAZ and AVI. However, the urine NCA analyses should be interpreted with caution, as some subjects in the combination cohorts displayed an the fraction of dose excreted in urine from the time of dosing to time t (Fe%) above 100%, suggesting that there may have been some inaccuracies in the urine data resulting from the inherent challenges in recording/collecting precise urine timing and volume. Furthermore, the day 1 and 7 CAZ and AVI AUC values in the plasma NCA analyses were higher in the CZA-ATM combination cohorts than in the CZA intermittent infusion cohort, suggesting that the increased CL_R with CZA-ATM coadministration does not result in clinically meaningful changes in the overall CL of CAZ and AVI. In the PopPK CAZ model, CZA-ATM coadministration was a significant covariate on both CL_R and CL_NR, but the net change of CZA-ATM coadministration on total CL was negligible, and CZA-ATM coadministration was not identified as a significant covariate on the CL parameters in the final AVI PopPK model. These findings further support the notion that CZA-ATM coadministration has minimal impact on the PK profiles of CAZ and AVI. As this was a phase 1 study of healthy subjects, further investigation is warranted, especially in patients with active infections and various degrees of renal function (25).

No identifiable relationships were observed in the overall ATM exposure-highest observed ALT/AST response analyses. Coadministration of CZA with ATM was also not found to be associated with an augmented risk of ALT/AST elevations. Since the shape of the concentration-time curve has been found to modulate the observed exposure-toxicity profile for other antibiotics (29–42), separate analyses were performed in subjects who received ATM as 2-h intermittent infusions and CI. Despite the occurrences of 2 severe ALT/AST adverse events (AEs) in the ATM CI cohort, no notable ATM exposure-ALT/AST associations were observed. The reasons for the 2 severe ALT/AST AEs in the subjects in the ATM CI cohort are unclear. These were likely idiosyncratic drug reactions that were potentially due to unrecognized intrinsic patient factors, as only one other subject in the ATM CI cohort had an ALT AE, and it was mild in severity (highest observed ALT of 57 U/L; upper limit of normal [ULN], 54 U/L) (43, 44). Of note, the 2 subjects with severe ALT/AST elevations in the ATM CI cohort were African American males, and individuals of African American descent have been reported to be at increased risk for more severe idiosyncratic drug-induced liver injury (45, 46). Furthermore, the degree of ALT/AST elevations for the 2 subjects in the ATM CI with severe ALT/AST elevation AEs was considerably higher than those observed in the subjects in the ATM 2-h intermittent infusion cohorts with serum aminotransferase AEs. This suggests that the pathogenesis for the drug-induced liver injury may have been different for the two subjects in the ATM CI cohort (e.g., immune mediated) relative to the subjects in the ATM 2-h intermittent infusion cohorts who had ALT/AST AEs (e.g., hepatocellular injury due to the adaptation process) (43, 44). Further study is needed to determine the patient populations at greatest risk for severe ALT/AST elevations with receipt of ATM alone and in combination with CZA. For now, some caution should be exercised with use of CI ATM, especially in males of African American ancestry.

Modest associations, best described by a power or linear log-link function, were observed between daily C_max and AUC values and ALT/AST elevations in subjects who received ATM as 2-h intermittent infusions. The most notable ALT/AST elevations in the ATM intermittent infusion cohorts were observed in subjects who received CZA with ATM 2 g intravenously (i.v.) over 2 h every 6 h (8/g daily) in cohort 6. While this suggests that administration of 2-h intermittent infusions of ATM with CZA exacerbates ALT/AST elevations, CZA-ATM administration was not identified as a significant covariate in the GLM analyses. The higher ALT/AST values observed in cohort 6 were most likely reflective of mild hepatocellular injury that increased as a function of the higher ATM exposures observed in subjects who received CZA with ATM 2 g i.v. over 2-h every 6 h (8/g daily) (43, 44). Coadministration of CZA-ATM was found to reduce the overall ATM CL by 16% in the final ATM PopPK model, resulting in higher daily ATM exposure estimates and potentially more pronounced ALT/AST elevations. While there is a potentially higher risk for ALT/AST elevations with CZA with 2-h intermittent infusion of ATM 8 g/daily versus 6 g/daily, this must be balanced against the observed increased bacterial killing with CZA with ATM 2 g i.v. over 2-h every 8 h in the HFIM study (13). Furthermore, the lack of an identifiable exposure-ALT relationship in the CI ATM cohort (cohort 4) may have been due to its lower post-day 1 C_max,ss and daily AUC values with CI ATM relative to the 2-h intermittent infusion cohorts. For many drugs, C_max and AUC are the PK/pharmacodynamics (PD) indices linked to exposure-associated AEs (33–35), and this may have accounted for the observed differences in exposure-ALT/AST response curves between the CI and 2-h intermittent infusion ATM cohorts. Since the pathogenesis of the observed ATM-induced liver injury was likely different for the two subjects with severe ALT/AST AEs in the ATM CI cohort, (43, 44), post hoc exposure-response regression analyses that excluded the two subjects with severe ALT/AST AEs in the ATM CI cohort were performed. In the post hoc curvilinear and GLM regression analyses, day 3 C_max and AUC, both expressed as power functions, were significantly associated with the highest observed ALT/AST values (Fig. S5 and 6). As this was a phase 1 study of a limited sample size, caution should be exercised in interpreting the results. Additionally, it is important to note that the higher incidence of ALT/AST AEs observed in cohort 6 than in cohort 5 may have been due to missed ATM doses on days 6 and 7 in several subjects in cohort 5. More study is needed in a more diverse patient population to discern if there are ATM exposure-related differences in ALT/AST AE profiles between ATM administered as a CI or an intermittent infusion. As part of future studies, plasma pharmacokinetic data should be collected, and PK/PD and Monte Carlo simulation modeling should be performed to delineate the therapeutic window for CZA-ATM combination therapy and the risk of ALT/AST elevations associated with the CZA-ATM regimens evaluated in this study (33).

From a clinical perspective, the aggregate findings from the ATM exposure-ALT/AST response analyses suggest that coadministration of CZA-ATM does not exacerbate AST/ALT elevations relative to ATM alone. However, coadministration of CZA-ATM was found to reduce ATM overall CL by 16% in the final ATM PopPK model, resulting in higher daily ATM exposures and potentially more pronounced ALT/AST elevations. This suggests that the benefits versus risks of using 2-h intermittent infusions of ATM at a dose of 8 g/daily with CZA should be considered. The results also suggest that CI ATM should be used with some caution given the 2 severe ALT/ASTs in the CI ATM cohort. Of note, none of the subjects with AST/ALT elevations were symptomatic, there were no clinical findings suggestive of liver damage (e.g., jaundice or impaired synthetic function), and AST/ALT normalized in all subjects after cessation of ATM. Regardless of the CZA-ATM regimen administered, liver function tests should be measured at baseline and daily, especially in patients receiving ≥4 days of treatment. Future comparator-controlled randomized clinical trials are required to better define the safety and PK of the CZA-ATM regimens evaluated in this study, especially among patients with active infections who have unstable renal function or baseline renal impairment. The range of ATM exposures observed in this phase 1 study likely were a conservative estimate of true exposures in patients, and the ATM exposure threshold(s) associated with clinically meaningful ALT/AST elevations are more likely to be delineated in a more diverse patient population with differing physiological states (27, 28). More data are also needed on the safety of the CZA-ATM combination regimens evaluated in this study for >7 days. Elevations of ALT/AST observed with ATM increased as a function of treatment duration (17), and there is the potential that ATM exposure-related ALT/AST elevations may be more pronounced among subjects receiving longer courses of CZA-ATM.

In conclusion, this was the first controlled trial to evaluate CZA in combination with ATM. The findings suggest that CZA-ATM coadministration does not result in clinically meaningful changes in the PK profiles of CAZ and AVI but is associated with an ~16% reduction in total ATM CL, resulting in higher ATM daily exposures. No notable relationships were identified between ATM exposure and ALT/AST elevations in the overall analyses, but modest ATM exposure-ALT/AST response associations were observed in subjects who received 2-h intermittent infusions of ATM. Although the findings from the ATM exposure-ALT/AST response analyses suggest that administration of CZA-ATM does not exacerbate AST/ALT elevations relative to ATM alone, the benefits versus risks of using CZA with 2-h intermittent infusions of ATM at a dose of 8 g/daily should be considered given the modest observed associations between the extent of daily ATM exposure and ALT/AST elevations. The results also indicate that some caution should be exercised with the use of CI ATM at 8 g/daily. Future controlled trials of longer treatment durations are warranted in patients with active infections and various degrees of renal function to better define the safety and PK of the CZA-ATM regimens evaluated in this study. Further study is also needed to determine the patient populations at greatest risk for severe ALT/AST elevations with receipt of ATM alone and in combination with CZA.

MATERIALS AND METHODS

Study selection and design.

COMBINE was an open-label single center phase 1 study that was conducted with healthy volunteers aged between 18 to 45 years to investigate the safety and PK of AVYCAZ combined with ATM, AVYCAZ alone, and ATM alone (ClinicalTrials.gov identifier NCT03978091). Following subject screening, eligible subjects were assigned to one of the six dosing cohorts, and investigational product(s) were administered for 7 days. Subjects were admitted to the study site on day −1 to confirm that they continued to meet eligibility criteria, after which they were admitted to the study site for dosing and observation. The study was approved by the Duke Health Institutional Review Board and conducted in accordance with good clinical practice principles as established by the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH).

Study safety was closely monitored using daily assessments of AE’s, vital signs, and clinical laboratory safety tests. A final visit (day 11 + 3) was scheduled with each subject for a final safety evaluation and collection of clinical laboratory safety tests. The type, incidence, relatedness, and severity of AEs and serious adverse events (SAEs) were recorded from the start of infusion of the first dose of study product on day 1 through to the final study visit. Adverse events were assessed by the investigator using the protocol-defined grading system that was based on the FDA Guidance for Industry for grading of AEs (47), and all AEs after the first dose of study product(s) were coded using Medical Dictionary for Regulatory Activities (MedDRA) terminology (dictionary version 23.1). The reference ALT range for males was 17 to 63 U/L; grade 1 (mild) AE was >63 to 105 U/L; grade 2 (moderate) AE was 106 to 175 U/L; grade 3 (severe) was >175 U/L. The ALT reference range for females was 14 to 54; grade 1 AE was >54 to 105 U/L; grade 2 AE was 106 to 175 U/L; grade 3 was >175 U/L. The AST reference range for all subjects was 15 to 41; grade 1 AE was >42 to 105 U/L; grade 3 AE was 106 to 175 U/L; grade 3 was >175 U/L. All AEs and laboratory abnormalities reported as AEs were followed until resolution, return to pretreatment status, or stabilization even if the duration of follow-up went beyond the final study visit.

Eligible subjects (48) were sequentially assigned to one of six cohorts (8 subjects per cohort), with study drug(s) administered for 7 days. Four of the six cohorts were single-study-drug cohorts (CZA 2.5 g i.v. over 2 h every 8 h [cohort 1], CZA 2.5 g i.v. over 2 h × 1 and then 7.5 g/daily as a CI [cohort 2], ATM 2 g i.v. over 2 h every 6 h [cohort 3], and ATM 2 g i.v. × 1 and then 8 g/daily as a CI [cohort 4]), and two were combination-study-drug cohorts (CZA 2.5 g i.v. over 2 h every 8 h with ATM 1.5 g i.v. over 2 h every 6 h [cohort 5] and CZA 2.5 g i.v. over 2 h every 8 h with ATM 2 g i.v. over 2 h every 6 h [cohort 6]). The two CZA-ATM combination cohorts included regimens that were found to be efficacious in the HFIM studies (13). Cohorts 1 to 4 were completed prior to cohorts 5 and 6.

In the initial design, the intent was to evaluate the safety and PK of CZA 2.5 g i.v. as 2-h infusions every 8 h combined with ATM 2 g i.v. as 2-h infusions every 6 h, and CZA in combination with ATM, each administered as CIs (CZA 7.5 g/day CI combined with ATM 8g/day CI). These CZA-ATM regimens were originally selected for evaluation in this phase 1 study because they were associated with maximal bacterial killing and resistance suppression over 7 days in the HFIM experiments (13). In the ATM-alone cohorts, two subjects in cohort 4 experienced severe AST/ALT adverse events (AEs). All subjects with ALT/AST AEs in cohort 4 were asymptomatic, and none had clinical findings suggestive of liver damage (e.g., jaundice or impaired synthetic function); all ALT/AST AEs resolved after discontinuation of ATM. In response to the occurrence of two severe AEs in cohort 4, the study was halted, and a safety monitoring committee (SMC) meeting was convened. The SMC did not recommend continuing CI ATM due to the lack of established safety with CI ATM. The SMC recommended continuing the study but reducing the dose of ATM in cohort 5 to ATM 1.5 g i.v. as 2-h infusions every 6 h (ATM 6 g/daily) and to administer it in combination with CZA 2.5 g i.v. as 2-h infusions every 8 h, a regimen found to be highly effective in the HFIM study (13), and to reconvene after completion of cohort 5 to review safety data. After review of the safety data from cohort 5, the SMC recommended administering ATM 2 g i.v. as 2-h infusions every 6 h (ATM 8 g/daily) in combination with CZA 2.5 g i.v. as 2-h infusions every 8 h in cohort 6 since no halting rules were met and no other safety concerns were observed in cohort 5. The SMC was in favor of escalating the ATM daily dose because the HFIM experiments demonstrated that there was increased bacterial killing with ATM 2 g i.v. over 2 h every 6 h (8 g/daily) versus ATM 1.5 g i.v. as 2-h infusions every 6 h (6 g/daily) when administered with CZA 2.5 g i.v. over 2 h every 8 h (13).

Blood and urine PK samples.

Venous blood (plasma) samples were collected in K₂EDTA collection tubes on the following study days and time points ±5 min for predose, during, and end of short infusion samples and then ±15 min for other samples in each cohort. Day 1: predose (10 min prior to the start of the infusion), 0.5 h, 1 h, 1.5 h, 2 h, 2.5 h, 3 h, 4 h, 5 h, 6 h, 7 h, and 8 h after the start of the 1st infusion; days 3 and 5: 10 min prior to the start of the 1st infusion; day 7: predose (10 min prior to the start of the last infusion), 0.5 h, 1 h, 1.5 h, 2 h, 2.5 h, 3 h, 4 h, 5 h, 6 h, 7 h, 8 h, 10 h, 12 h, 18 h, and 24 h after the start of the day 7 infusion. Serial urine samples were collected for all cohorts on days 1 and 6 at the following intervals: >0 to 4 h, >4 to 8 h, >8 to 12 h, and >12 to 24 h after the start of the morning dose administration(s) of the day. The total volume of urine was collected to determine the amount of ceftazidime (CAZ), AVI, and ATM in urine. Plasma and urine PK specimens were processed and stored in a −80°C freezer until time of shipment to the central laboratory for analysis.

Analytical methods.

Plasma and urine PK samples were analyzed for concentrations of CAZ, AVI, and ATM using validated liquid chromatography–tandem mass spectrometry (LC-MS/MS) assays. For the CAZ and AVI plasma assay, the lower limit of quantification (LLOQ) was 0.04 μg/mL for CAZ and 0.01 μg/mL for AVI. Intra-assay precision (percent CV) was 1.87 to 14.73% and 2.26 to 9.67% for CAZ and AVI, respectively. Intra-assay accuracy (percent relative error [RE]) was −4.58 to 6.85% and −7.78 to 9.83% for CAZ and AVI, respectively. Inter-assay precision (percent CV) was 3.57 to 10.10% and 3.38 to 11.87% for CAZ and AVI, respectively. Inter-assay accuracy (percent RE) was 1.19 to 6.98% for CAZ and –4.37 to 4.22% for AVI. For the urine assay, the LLOQ was 0.25 μg/mL for CAZ and 0.0625 μg/mL for AVI. Intra-assay precision (percent CV) was 3.85 to 8.05% for CAZ and 2.99 to 7.69% for AVI. Intra-assay accuracy (percent RE) was 2.30 to 9.30% for CAZ and –3.91 to 9.80% for AVI. Inter-assay precision (percent CV) was 6.35 to 8.33% for CAZ and 3.19 to 8.24% for AVI. Inter-assay accuracy (percent RE) was –6.22 to 4.44% and –4.34 to 3.76% for CAZ and AVI, respectively. For ATM, the LLOQ was 0.2 μg/mL for the plasma and urine assays. For the assay in plasma, intra- and interassay precision (percent CV) were 1.00 to 7.81% and 3.91 to 5.48%, respectively, and intra- and interassay accuracy (percent RE) were –1.88 to 4.08% and 1.67% to 2.92%, respectively. For the urine PK assay, intra- and interassay precision (percent CV) were 1.75 to 3.92% and 2.70 to 7.43%, respectively. Intra-assay and interassay accuracy (percent RE) were –4.83 to 8.56% and –3.92 to –0.17%, respectively.

PK population and data set.

To be included in the PK evaluable population, subjects had to receive CZA alone, ATM alone, or CZA in combination with ATM and have at least one quantifiable plasma or urine concentration of CAZ, AVI, or ATM. Actual dosing (i.e., infusion start and stop) and PK sampling times were included in the time-ordered analysis data set and used in all PK computations. Drug infusion rates and volume of drug remaining in the bag for agents in which the full infusion was not completed were also used in the PK calculations. Nominal PK sampling times were only used in the mean plasma concentration (standard deviation [SD]) aggregation. For the calculation of PK parameters in the noncompartmental analyses (NCA), a value of zero was assigned for PK samples below the quantification limit (BQL) that were collected prior to the time at maximal concentration (C_max). For samples occurring after the C_max, half the lower limit of quantification (LLOQ) was assigned for the first BQL sample, and a value of zero was assigned for all subsequent BQL samples. For the primary population PK modeling, all BQL samples were excluded. Additionally, outlier concentrations were excluded in both the NCA and population PK analyses based upon visual inspection of concentration-time data as recommended by the U.S. FDA guidance given the potential for these observations to negatively impact model convergence and/or the final parameter estimates (49). For ATM, a sensitivity analysis using different BQL rules (M1 and M3 methods) were compared with these observations excluded/included from the population PK analysis (50–53).

Noncompartmental analysis (NCA).

A NCA was performed using Phoenix WinNonLin (version 8.2; Certara, Princeton, NJ, USA). The PK parameters for CAZ, AVI, and ATM in plasma and urine were summarized using the geometric mean and geometric CV percentage. Actual dosing and PK sampling times were used in all computations. After administration of the first dose on day 1, the following plasma PK parameters were calculated for study product(s) received in each cohort: C_max, minimum concentration (C_min) (cohorts 1, 3, 5, and 6 only), steady-state concentration (C_ss) (cohorts 2 and 4 only), and area under the concentration versus time curve over one dosing interval (AUC_0–Tau) (cohorts 1, 3, 5, and 6 only). For CAZ and AVI in cohorts 1, 5, and 6 on day 1, Tau was set to 8 h after the first dose. For ATM in cohorts 3, 5, and 6, Tau was set as 6 h after the first dose on day 1. Due to NCA data requirements, it was not possible to calculate AUC_0–Tau for the CI cohorts (2 and 4) since there were multiple dosing inputs (bolus and CI) and no interdose Tau. After the last dose on day 7, the following PK parameters were calculated: C_max, AUC_0–Tau (Tau = 8 h for CAZ and AVI; Tau = 6 h for ATM), AUC_inf, terminal elimination rate constant (K_el), terminal-phase disposition half-life (t_1/2), total body clearance (CL), and volume of distribution at steady state (V_ss). In addition, the accumulation ratio for AUC_0–Tau (RAUC_0–Tau) for CAZ, AVI, or ATM, was calculated after the first dose on day 1 and last dose on day 7. Like the other AUC PK parameters, the accumulation ratio parameters could not be estimated for the CI cohorts.

The amount of drug excreted during each urine collection interval was calculated by multiplying the urine drug concentration measured in a sample from a urine collection interval by the volume of urine collected for the corresponding collection interval. The amount of drug excreted into the urine (Ae_0–Tau) on days 1 and 6 from time of dosing (0 h) to 4 h, >4 to 8 h, >8 to 12 h, and >12 to 24 h and the cumulative amount (Ae_0–t) from 0 to 8 h and 0 to 24 h were calculated. The cumulative amounts (Ae_0–t) excreted into the urine from the time of dosing up to time t were obtained by adding the amounts excreted over each collection interval up to t. The fraction of dose excreted in urine (Fe_0–Tau) on days 1 and 6 from 0 to 4 h (ATM), 0 to 8 h (CAZ and AVI), and 0 to 24 h were determined based on actual samples collected from each patient. Due to the timing of urine collection and dosing, the renal CL (CL_R) for CAZ and AVI was calculated as CL_R = Ae_0–8/AUC_0–8, and ATM CL_R was calculated as CL_R Ae_0–4/AUC_0–4. Because no plasma samples were collected on day 6, the CL_R was calculated using the day 1 data only.

Population PK analysis.

CAZ, AVI, and ATM plasma and urine PK data following administration of CZA in combination with ATM, CZA alone, or ATM alone were analyzed with a nonlinear mixed effects modeling approach using the software NONMEM (Version 7.4; Icon Solutions, Ellicott City, MD, USA). Base population PK (PopPK) models were developed separately for CAZ, AVI, and ATM. The first-order conditional estimation method with interaction was used for all model runs. Based on visual inspection of the plasma PK data, 1- and 2-compartment base PopPK models were evaluated for CAZ, AVI, and ATM. The best model structure for plasma PK for each drug was then linked to a separate compartment for urine PK via CL_R and fit to the plasma and urine data simultaneously. Interindividual variability (IIV) was assessed for PK model parameters using an exponential relationship. Estimation of a covariance matrix for IIV on CL_R, volume of the central compartment (V₁), intercompartmental clearance (Q), volume of the peripheral compartment (V₂), and nonrenal clearance (CL_NR) was attempted, and IIV was removed when shrinkage was high (i.e., 30% or greater) in base PK models. Different error models (i.e., additive, proportional, and combined additive and proportional error) were tested to describe residual variability (σ²).

After developing a base structural PopPK model for each drug, the potential effects of clinical covariates on PK parameters were evaluated if relationships were suggested by visual inspection of scatter and box plots (continuous and categorical variables, respectively) of the individual ETA against covariates. The following covariates were explored: age, sex, weight (WT), body mass index (BMI), body surface area (BSA), lean body weight (LBW), baseline serum creatinine (S_CR), baseline creatinine clearance (CL_CR) estimated by the Cockcroft-Gault method (54), coadministration of CZA-ATM, and CI. A forward inclusion (P <~0.05 and Δ objective function value [OFV] of >3.84) and backward elimination (P <~0.01 and ΔOFV of >6.64) with 1 degree of freedom were used to evaluate statistical significance of the relevant covariates. Due to the relationship between CL_CR with body weight, if a body size descriptor was added to a model prior to CL_CR, S_Cr versus CL_CR was considered prior to deciding which additional covariate effects to include to avoid adding body weight twice to the same covariate.

During the CAZ, AVI, and ATM base and final PopPK model building processes, successful minimization, diagnostic plots, plausibility, and precision of parameter estimates, as well as objective function and shrinkage values, were used to assess model appropriateness. The diagnostic plots generated included individual predictions (IPRED) and population predictions (PRED) versus observations, conditional weighted residuals (CWRES) versus PRED and time, and individual weighted residuals versus IPRED. Scatterplot matrices of random parameters were used to screen for potential covariance among the interindividual error terms for each parameter in the full multivariable model prior to estimation of covariance terms. Prediction-corrected visual predictive checks (pcVPCs) for the popPK models were performed for the final model by generating 1,000 Monte Carlo simulation replicates per time point of CAZ, AVI, and ATM exposure. Simulated results were compared at the participant level with those observed in the study by calculating and plotting the percentile of each observed concentration. The dosing and covariate values used to generate the simulations in the pcVPCs were the same as those used in the study population.

Summary statistics of the individual post hoc PK parameters (empirical Bayesian estimates [EBE]) for each PopPK model were calculated. The final PopPK models and individual dosing histories were used to simulate the individual predicted (IPRED) plasma concentration-time data for each study subject. The simulations included individual patient covariates who received each study drug(s) and the IIV (55). Simulations included 25 replicates per time point, and the mean plasma concentration-time value from the simulations were then analyzed using NCA PK methods as described above to estimate the following on days 1 and 3 for each subject who received each study drug(s): the maximum simulated concentration at steady state (C_max,ss), the minimum simulated concentration at steady state (C_min,ss) for cohorts 1, 3, 5, and 6, and steady-state concentration (C_ss) for cohorts 2 and 4. The daily AUC values on days 1 and 3 (AUC_0–24 and AUC_48–72) for each subject who received each study drug(s) were estimated using numerical integration of the simulated mean concentration-time value from 0 to 72 h postdose.

ATM plasma exposure and ALT/AST response analyses.

The associations between day 1 ATM or day 3 ATM exposures (estimated for each subject from the final ATM PopPK model, EBEs, and individual dosing histories) and highest observed ALT/AST values during the study were assessed in all subjects who received ATM alone or CZA-ATM. Day 1 and 3 exposures were selected for the ATM exposure-ALT/AST response analysis for several reasons. Day 1 and 3 exposures preceded the occurrence of ALT/AST elevations. Intensive PK sampling was performed on day 1, and the ATM PK parameter estimates and EBEs of the PopPK models were heavily informed by day 1 plasma PK time points. Due to the administration of a loading dose in the ATM CI cohort (cohort 4), day 3 exposures were examined to minimize the potential influence of the loading dose on the observed day 1 exposure estimates for the ATM CI cohort. ATM plasma PK parameter and exposure estimates on days 1 and 7 were found to be nearly identical in the NCA PK analyses, suggesting that it was highly reasonable to use the ATM PopPK model to estimate time-concentration profiles on day 3 when there was sparse sampling. The highest observed ALT/AST values observed during the study were selected as the primary outcomes for the exposure-response analyses since they are the serum aminotransferase values used for grading the severity of liver injury in clinical trials and in clinical practice (56–58).

For several antibiotics (e.g., aminoglycosides, vancomycin, tedizolid, iclaprim, amphotericin, daptomycin), different administration strategies have been found to alter the observed safety profile, and the PK/PD driver or exposure threshold associated with toxicity varied as a function of the frequency of dosing and duration of infusion (29–42). Given the different shape of the ATM concentration-time profiles for subjects who received 2-h intermittent infusions and CI and its potential effect on the observed safety profile (29–42), the associations between ATM day 1 and day 3 exposures and highest observed ALT/AST values were evaluated separately for subjects who received ATM as either a 2-h intermittent infusion (cohorts 3, 5, and 6) or CI (cohort 4). Relationships between the ATM day 1 and day 3 exposures and highest observed ALT/AST values were first evaluated graphically by plotting the highest observed ALT/AST value versus each ATM drug exposure variable. A locally weighted scatterplot smoothing (LOWESS) function was fit to each scatterplot to visually assess the presence of any potential ATM exposure-ALT/AST response associations. Based on the observed LOWESS curves, curvilinear regression was then used to assess the potential fit of different functions (e.g., power, compound, exponential, sigmoid maximum effect [E_max], etc.) to ATM exposure-ALT/AST scatterplots that suggested the presence of a potential exposure-response association, with β coefficients having a P value of less than 0.05 considered significant. Generalized linear model (GLM) analyses were also conducted to evaluate exposure-ALT/AST associations for ATM exposure variables that were identified by the LOWESS curves as being potentially related to the highest observed ALT/AST values. Additionally, the associations between coadministration of CZA-ATM and the highest observed ALT/AST values were assessed in the GLM analyses.

Data availability.

Researchers interested in accessing the clinical trial data presented here are encouraged to submit a research proposal and publication plan. The proposal and plan will be reviewed by the Antibacterial Resistance Leadership Group (ARLG) publications committee and/or appropriate study team members. If approved and upon receipt and approval of a signed data access/use agreement, individual participant data necessary to complete the proposed analysis will be made available. Related documents, including the study protocol, statistical analysis plan, and data dictionary, may also be shared. Access to data will only be granted to researchers who provide a methodologically and scientifically sound proposal. Proposed analyses that are duplicative of ongoing or proposed analyses may not be supported. To submit a proposal, please complete a proposal at https://arlg.org/how-to-apply/protocol-concept. Alternatively, visit dcri.org/data-sharing. There may be costs associated with data sharing that researchers would be expected to cover.