Pandrug-resistant Gram-negative bacteria: a systematic review of current epidemiology, prognosis and treatment options

Abstract

Background

The literature on the epidemiology, mortality and treatment of pandrug-resistant (PDR) Gram-negative bacteria (GNB) is scarce, scattered and controversial.

Objectives

To consolidate the relevant literature and identify treatment options for PDR GNB infections.

Methods

A systematic search in MEDLINE, Scopus and clinical trial registries was conducted. Studies reporting PDR clinical isolates were eligible for review if susceptibility testing for all major antimicrobials had been performed. Characteristics and findings of retrieved studies were qualitatively synthesized.

Results

Of 81 studies reviewed, 47 (58%) were published in the last 5 years. The reports reflected a worldwide dissemination of PDR GNB in 25 countries in 5 continents. Of 526 PDR isolates reported, Pseudomonas aeruginosa (n=175), Acinetobacter baumannii (n=172) and Klebsiella pneumoniae (n=125) were most common. PDR GNB were typically isolated in ICUs, but several studies demonstrated wider outbreak potential, including dissemination to long-term care facilities and international spread. All-cause mortality was high (range 20%–71%), but appeared to be substantially reduced in studies reporting treatment regimens active in vitro. No controlled trial has been performed to date, but several case reports and series noted successful use of various regimens, predominantly synergistic combinations, and in selected patients increased exposure regimens and newer antibiotics.

Conclusions

PDR GNB are increasingly being reported worldwide and are associated with high mortality. Several treatment regimens have been successfully used, of which synergistic combinations appear to be most promising and often the only available option. More pharmacokinetic/pharmacodynamic and outcome studies are needed to guide the use of synergistic combinations.

Introduction

Mathematical prediction models have estimated that thousands of extra deaths are attributable to MDR bacterial infections every year in Europe.¹ However, the mortality attributable to XDR and pandrug-resistant (PDR) infections appears to be much lower based on real-life data from a recent study in France.¹ In contrast, an alarming increase in the incidence of PDR infections, most frequently caused by PDR Acinetobacter baumannii, has been detected in the authors’ region² and has been associated with high mortality.³ In view of such controversial findings, studying the worldwide epidemiology of pandrug resistance becomes especially important. Furthermore, studies on the treatment of PDR infections are scarce and scattered,^4,5 and clinicians often resort to ‘salvage treatments’, including the use of antimicrobial combinations or antibiotics for non-approved indications, with questionable dosing and administration routes.⁶

This review aims to systematically search and consolidate the literature on the epidemiology, mortality and treatment of PDR Gram-negative bacteria (GNB) infections. In particular, considering the controversies described above, we examined the geographical dissemination of PDR GNB and the mortality associated with PDR GNB infections. Furthermore, treatment options that have been reported to be effective against PDR GNB infections are summarized and their potential to reduce mortality is assessed.

Methods

This systematic review complies with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.⁷

Search strategy

We conducted the following search in MEDLINE (PubMed) and Scopus, from inception to May 2019: panresistant OR panresistance OR pan-resistant OR pan resistant OR pandrug OR pan-drug OR pandrug-resistant OR pan-drug-resistant OR therapeutic dead end OR therapeutic impasse. Retrieved articles were screened for relevance based on their titles and abstracts. The full texts of potentially eligible articles were then reviewed. The search was supplemented by reference tracking of included papers. Additionally, we used the same search terms in clinical trial registries (the International Clinical Trial Registry Platform, the Cochrane Central Register of Controlled Trials, ClinicalTrials.gov, the Australia and New Zealand Clinical Trials Registry, the International Standard Randomised Controlled Trial Number and the European Clinical Trials Database) to identify trials assessing the management of PDR infections.

Eligibility criteria

We included all types of study providing information regarding the epidemiology, prognosis and treatment of PDR GNB isolated from clinical samples in healthcare settings. We used the definition and list of antimicrobial agents proposed by Magiorakos et al.⁸ plus tigecycline for A. baumannii to define PDR bacteria. We included studies reporting at least one PDR isolate in which susceptibility testing was performed for at least one agent in each of the following antimicrobial categories (with the exception of intrinsic resistance to these agents): carbapenems, polymyxins, aminoglycosides and glycylcyclines (tigecycline). Eligibilities of non-English language studies were assessed based on their English titles and abstracts. When necessary, the full text was translated into English using Google Translate.

Data items and collection process

The following data were collected: study design, country, time period, bacterial species, list of antibiotics used for susceptibility testing, criteria used to define susceptibility breakpoints, method of susceptibility testing for colistin and tigecycline, breakpoint used for susceptibility to tigecycline, number of PDR isolates, proportion of PDR bacteria (of all examined isolates), and treatment regimens and outcomes (both clinical and microbiological) of infections caused by PDR GNB. Reports of the same isolate in more than one study were counted only once. Data extraction was carried out by the first author.

Exploration of the activities of newer agents

Because few relevant studies were found in our main search, we expanded the search in MEDLINE with the following terms: plazomicin OR eravacycline OR vaborbactam OR (avibactam AND aztreonam) OR (ceftazidime AND avibactam) OR ceftolozane. Only studies reporting activities of these agents against GNB resistant to all other antimicrobials were included.

Risk of bias

We assessed the completeness of the lists of antimicrobials used to define PDR bacteria and the methods of susceptibility testing focusing on polymyxins, taking into account that broth microdilution is the only currently recommended method.⁹ In interpreting the proportion of bacteria that were PDR, we took into account the populations studied and the fact that studies that did not report any PDR isolates were excluded from the review. We also considered the geographical distributions of possible PDR isolates from non-eligible studies. For studies reporting patient outcomes, we examined the definition of outcome, the length of follow-up and whether mortality was attributable to the infection. In studies reporting treatment regimens, we recorded their design and whether patients with polymicrobial infections were included.

Synthesis of results

We conducted a qualitative presentation, and synthesis of the characteristics and findings of retrieved studies. Because of substantial clinical heterogeneity, a meta-analysis was not pursued.

Results

Study selection

A flow chart of our review is depicted in Figure 1. A flow chart focused on the non-English language literature is depicted in Figure S1 (available as Supplementary data at JAC Online). Non-eligible studies reporting PDR isolates are summarized in Tables S5 to S7.

Figure 1.

Flow chart of the review.

Study characteristics

A total of 81 studies were eligible for this review; their main characteristics are summarized in Tables S1 to S3. The earliest study was published in 2004. Most studies (n=73; 90%) were published within the last decade; more than one-half (n=47) were published in the last 5 years and about a one-third (n=28) in the last 2 years. The proportion that were PDR among clinical isolates was available in 44 (54%) studies.^1,10–51 All-cause mortality was reported in 33 (41%)^{1,4,5,20,24,35–37,51–75} studies. Treatment regimens were reported in 26 (32%) studies.^{4,5,24,37,49,51–55,57,58,60,63–66,68–72,75,76}

Definition of PDR

Susceptibility testing for the full list of antimicrobial agents as recommended in consensus definitions⁸ was reported in only six (7%) studies.^{32,34,40,59,77–79} Antimicrobials for which susceptibility testing was most frequently not reported were: fosfomycin, older-generation tetracyclines (such as minocycline), aminoglycosides (missing at least one aminoglycoside), trimethoprim/sulfamethoxazole and ampicillin/sulbactam (for A. baumannii) (Table S4). Isolates with intermediate susceptibility were interpreted as non-susceptible in most studies (n=53, 65%). In 14 (17%) studies,^{49,52,54–59,61,63,64,69,71,80} none of the isolates had intermediate susceptibilities, while in 14 (17%) other studies,^{16,17,28,37,38,43,46,48,50,51,62,68,74,81} the interpretation of intermediate susceptibility was not clarified. Broth microdilution to assess susceptibility to colistin was used in only 26 (32%) studies.^{4,10–12,14,15,18,22,27,29–32,37,42,50,56,59,61,66,68,69,78,80,82,83} The EUCAST contemporary susceptibility breakpoints for tigecycline (MIC>1 mg/L) were used in 13 (16%) studies,^{11,13,18,22,29,30,33,42,45,49,52,67,69} while FDA breakpoints (MIC>2 mg/L or FDA disc diffusion breakpoints) were used in 11 (14%) studies.^{4,16,19–21,23,27,35,38,43,84} In 18 (22%) studies exact MICs were reported,^{5,49,50,56,57,59,61,63,64,66,67,69,71,77–80,82} with almost all isolates having MICs>2 mg/L.

Epidemiology of PDR

A total of 526 PDR GNB clinical isolates were reported in 25 countries in five continents (Table 1), reflecting a worldwide distribution. Although the ward of isolation was not reported for 59% (n=311) of all PDR isolates, at least 37% (n=194) were isolated from ICU patients. Nevertheless, several studies demonstrated the potential of PDR pathogens to cause outbreaks, involving both intra-^38,65,73 and inter-hospital dissemination,⁶¹ spread between hospitals and long-term care facilities,^38,67 and international spread even to countries with a low rate of resistant pathogens.^56,67,79

Table 1.

Geographical distribution of PDR isolates by species

	Total (n=526)a	A. baumannii (n=172)	K. pneumoniae (n=125)	P. aeruginosa (n=175)a	Other (n=54)
Europe	280	107	71	80	22
Greece	181	100	47	17	P. stuartii n=16
					Enterobacteriaceae NS n=1
Spain	50	5	1	43	B. cepacia n=1
Italy	15	1	9	5	0
France	13	0	7	4	B. cepacia n=2
Slovakia	8	0	0	8	0
Germany	3	1	0	0	Serratia marcescens n=2
Belgium	2	0	0	2	0
Serbia	1	0	0	1	0
Netherlands	6	0	6	0	0
Portugal	1	0	1	0	0
Americas	24	2	12	8	2
USA	9	0	5	2	Enterobacteriaceae NS n=2
Brazil	12	2	7	3	0
Mexico	2	0	0	2	0
Canada	1	0	0	1	0
Asia	180	63	41	52	24
India	101	11	28	40	22b
Turkey	11	1	7	1	Enterobacteriaceae NS n=2
Russia	3	3	0	0	0
Pakistan	29	19	2	8	0
Iran	18	17	0	1	0
Thailand	12	12	0	0	0
United Arab Emirates	4	0	4	0	0
Taiwan	1	0	0	1	0
Japan	1	0	0	1	0
Australia	35	0	1	34	0
Australia	35	0	1	34	0
Africa	6	0	0	0	6
South Africa	6	0	0	0	Serratia marcescens n=6

Continent totals are shown in bold. NS, not specified.

a

The source (country) of one PDR P. aeruginosa isolate from a multicentre study was not reported.⁴²

b

E. coli n=1, Enterobacter spp. n=6, Burkholderia spp. n=3, E. meningoseptica n=3, C. indologenes n=5, M. morganii n=2 and unclear n=2.

The most common PDR species were Pseudomonas aeruginosa (n=175, 33%), A. baumannii (n=172, 33%) and Klebsiella pneumoniae (n=125, 24%). Other less-common PDR pathogens included Providencia stuartii (n=16),^27,65,Serratia marcescens (n=8),^22,77,Enterobacter spp. (n=6),^21,Burkholderia spp. (n=6),^1,21,55,Chryseobacterium indologenes (n=5),^21,54,Elizabethkingia meningoseptica (n=3),^21,Morganella morganii (n=2),²¹ and Escherichia coli (n=1).²¹

The proportions of the examined isolates that were PDR were highly variable (ranging from 0.01% to 21%) among the reviewed studies, reflecting their heterogeneity in terms of bacterial species, geographical locations, patient populations, and types of infection or culture sites. For example, PDR proportions ranged from 0.01% to 0.20% in large (>3000 patients) multicentre studies involving diverse patient populations,^{13,23,27,29,31,42,83} but were much higher (0.65% to 11%) in studies of ICU patients^{15,19,28,36,44} and in studies that focused on MDR (3%⁴⁵), carbapenem-resistant (5.7%,¹¹ 13.3%,²² 2%³² and 7.7%⁴⁹), XDR (10.8%,²⁰ 4.7%,¹⁶ 6.5%,³⁵ 17.7%³³ and 3.7%³⁷) or colistin-resistant isolates (13.9%¹⁸).

Prognosis of infections by PDR

All-cause mortality was examined in 142 patients with PDR GNB infection.^{1,4,5,20,24,35–37,49,51–74,76} To assess mortality in these studies, the patients were followed for variable lengths of time; 28 or 30 days,^1,4,5,35,36 or until discharge from the ward or the hospital.^{24,37,52–57,59,60,63–72,75} Summing the data from all studies, all-cause mortality in PDR GNB infection was 53% (n=75 of 142) but highly variable; excluding studies with less than five patients, all-cause mortality ranged from 20% to 71%.^{4,5,20,35,36,51,65,73–75} Mortality was high irrespective of infecting pathogen or site of infection, ranging from about 20% in urinary tract infections to >40% in other sites (Table 2). Among the 75 reported deaths, mortality was judged by the authors to be directly attributable to the infection in 11 cases,^{20,52,56,61,62,67,71,73,75} but there were at least 6 deaths known to be unrelated to the PDR infection^{5,57,60,68,72} and for the rest of the cases this was not discussed.

Table 2.

All-cause mortality^a of patients with PDR infections by bacterial species and site of infection^b

	Deaths (%), number of deaths/total number of patients
	all sites	BSIc	RTId	UTI without BSI	CNSe	osteomyelitis	IAI
Total	53, 75/142	50, 21/42	44, 19/43	23, 3/13	75, 3/4	67, 2/3	50, 1/2
P. aeruginosa	58, 30/52	56, 5/9	31, 4/13	0, 0/1	50, 1/2	no data	0, 0/1
K. pneumoniae	47, 16/34	31, 5/16	50, 1/2	29, 2/7	100, 1/1	67, 2/3	100, 1/1
A baumannii	56, 20/36	71, 5/7	50, 14/28	no data	100, 1/1	no data	no data
P. stuartii	47, 7/15	60, 6/10	no data	20, 1/5	no data	no data	no data

BSI, bloodstream infection; IAI, Intra-abdominal infection; RTI, Respiratory tract infection; UTI, urinary tract infection

a

Mortality was variably defined in included studies (28 or 30 day mortality, mortality up to discharge from the ward or mortality up to discharge from the hospital).

b

In some studies it was not possible to extract data for each site of infection.

c

BSIs including primary BSI, catheter-related BSI, BSI secondary to UTI and BSI secondary to cellulitis.

d

RTIs including pneumonia and ventilator-associated pneumonia.

e

CNS infections.

Comparison of the mortality in PDR GNB infections to that in XDR GNB infections was possible in four small-scale studies, and mortality was substantially higher for the former in all of them (71% versus 55%,²⁰ 67% versus 57%,³⁶ 67% versus 30%³⁵ and 36% versus 23%⁵). However, differences were not statistically significant in individual studies, and confounding factors such as the severity of the underlying conditions or comorbidities, the sites of infection and use of different treatment regimens were not considered.

All-cause mortality was very high (43/61 patients, 70%) with treatment regimens inactive in vitro,^{24,28,39,40,55,58,73,74,77} but was substantially lower (19/61 patients, 31%) with regimens that were confirmed to be active in vitro, mainly involving synergistic combinations.^{4,5,37,52,53,55,57,58,60,63–66,68–70,73}

Treatment regimens

We did not find any completed or ongoing clinical trials on the treatment of PDR GNB infections. The only available data came from 26 studies, exclusively case series or case reports, involving n=105 patients: K. pneumoniae n=31,^{5,49,52,53,57,63,64,66,68,69,71,}P. aeruginosa n=45,^{37,51,58,60,70,72,73,75,76,}A. baumannii n=11,^4,24,Pr. stuartii n=15,^65,C. indologenes n=2⁵⁴ and Burkholderia cepacia n=1.⁵⁵ In 8 of the 26 studies, polymicrobial infections or patients with concurrent infections by other bacteria in other sites were included.^{35–37,57,70,73–75}

Treatment of PDR K. pneumoniae

Treatment regimens for PDR K. pneumoniae infections were reported in n=11 case series or case reports, and are summarized in Table 3. Four studies reported successful use of double-carbapenem combinations, either alone^5,57,68 or combined with colistin,^57,66 against KPC-producing PDR K. pneumoniae. It is notable that in two of the studies, successful clinical and microbiological outcomes were reported despite high MICs (128–256 mg/L) of carbapenems,^5,68 and the synergism was confirmed in vitro with time–kill assays.⁶⁸ Other effective treatment regimens were ceftazidime/avibactam,^53,63,64 and high-dose (200 mg once daily) tigecycline combined with colistin and amikacin, a synergistic combination in vitro⁶⁹ (Table 3).

Table 3.

Studies of treatment options for PDR K. pneumoniae

Study description		Treatment regimen	Outcomes
Double-carbapenem combinations (±colistin)
Oliva et al. 201468	Case series of three patients in Italy with BSI.	Cases 1 and 3: 2 g of meropenem q8h plus 1 g of ertapenem q24h.	Case 1: ‘complete recovery’ after 21 days of treatment.
		Case 2: 500 mg of ertapenem q24h and 1 g of meropenem q12h (doses adjusted to creatinine clearance).	Case 2: ‘The patient became afebrile after 48 h of treatment and blood cultures were sterile; however, he died 2 days later due to ‘acute heart failure’.
			Case 3: complete recovery after 24 days of treatment.
Souli et al. 20175	Case series of 14 patients in Greece. UTI n=3, sBSI due to UTI n=2, sBSI due to PN n=1, BSI n=5, VAP n=1, CRBSI n=1, EVD n=1.	Ertapenem (1 g, 1 h infusion) q24h administered 1 h prior to the first dose of meropenem, which was given at a dose of 2 g q8h (3 h infusion) or equivalent renally adjusted doses.	Clinical and microbiological outcome was evaluated on days 14 and 28, and patients were followed up to discharge; n=11 responded clinically and n=10 responded both clinically and microbiologically. At last follow-up, n=9 were alive.
Oliva et al. 201566	Case report of a bloodstream infection (both urine and central venous catheter cultures were also positive). Italy.	Ertapenem 1 g/day plus meropenem 2 g q8h plus IV colistin (loading 6 MIU, then 4.5 MIU q12h). The triple combination (ertapenem, meropenem and colistin) was found to be more rapidly bactericidal compared with double carbapenem alone in time–kill assays.	‘After 96 h she became afebrile. Laboratory analyses showed a reduction of the ESR and CRP. Blood and urine cultures did not grow any organism’. The patient was discharged after 14 days.
Emre et al. 201857	Report of two patients in Turkey. One with soft tissue infection and one with catheter-related bacteraemia.	Case 1: meropenem 1 g×3 plus ertapenem 1 g×1 plus colistin 9 MIU loading dose followed by 4.5 MIU q12h.	In both patients repeat cultures were sterile. The first patient died at day 32 (not attributable to the infection). The second patient was followed up to day 77 (cured).
		Case 2: meropenem 1 g×3 + ertapenem 1 g×1.
Regimens based on ceftazidime/avibactam
Parruti et al. 201953	Case report. Italy. Recurrent bacteraemia secondary to vertebral osteomyelitis associated with prosthetic material.	The final regimen included: ceftazidime/avibactam (2 g/8 h), tigecycline (loading 100 mg then 50 mg/12 h), meropenem (2 g/8 h) and gentamicin (loading 7 mg/kg, then 5 mg/kg/24 h)	Repeated recurrences despite transient response and removal of the prosthetic material. The patient finally responded to treatment including ceftazidime/avibactam.
Mandrawa et al.  201663	Case report from Australia. Severe pancreatitis complicated by IAI by PDR K. pneumoniae.	Ceftazidime/avibactam (plus metronidazole and teicoplanin). The isolate (KPC2-producing) was susceptible in vitro to ceftazidime/avibactam.	The patient demonstrated a clinical, biochemical and radiological response with no development of in vitro resistance after 6 weeks of treatment. However, microbiological clearance was not achieved, surgical management was not possible and the patient died soon after.
Camargo et al. 201564	Case report from USA. BSI. Also isolated from urine at >105 cfu/mL and the central venous catheter tip (>102 cfu/mL).	Ceftazidime/avibactam (1 g/250 mg q8h) and ertapenem (1 g q24h) (doses adjusted for creatinine clearance). Of note is that the patient had previously failed a triple regimen (meropenem, ertapenem and colistin). The isolate was susceptible in vitro to ceftazidime/avibactam alone and synergy was noted with carbapenems.	The patient responded well, with sterilization of blood cultures within 24 h. She was discharged from the ICU after 2 weeks of treatment.
Other treatment regimens
Humphries et al. 201069	Case report. USA. CRBSI. MICs: colistin >8 mg/L, carbapenems >16 mg/L, amikacin 32 mg/L, tigecycline 2 mg/L.	IV (10.5 MIU q24h) and inhaled (2.25 MIU q12h) colistin, high-dose tigecycline (200 mg IV daily) and iv amikacin (500 mg q12h for 10 days). Of note is that the patient had not responded to a combination regimen including standard-dose tigecycline (50 mg twice daily). Both tigecycline plus amikacin and tigecycline plus colistin were found to be synergistic in chequerboard assays.	The patient improved clinically and was discharged after 75 days. The bacteraemia resolved but the patient remained colonized (rectal swab and sputum cultures).
Treviño et al. 201149	Case report (only n=1 PDR infection). Spain. Bacteraemia.	Case 1: first tigecycline 50 mg q12h (MIC=1 mg/L), then fosfomycin 4 g q8h (MIC=192 mg/L) plus amikacin 500 mg q12h (MIC=32 mg/L).	‘Failure and still in hospital’.
Alho et al. 201952	Case report. Portugal. BSI secondary to osteomyelitis. Patient with ALL.	Colistin 5 MIU q12h (dose increased from 4 MIU, which had been given for 12 days), tigecycline 200 mg loading followed by 100 mg q12h, meropenem 2 g q8h and amikacin 1 g q24h. MICs were not reported.	The patient died with persistent bacteraemia.
Elemam et al. 200971	Case report. USA. Case 1: catheter-related UTI.	Case 1: catheter removal and tigecycline for 10 days (MIC>8 mg/L).	Case 1: persistent dysuria at discharge. Spontaneous resolution of symptoms but persistent bacteriuria 1 year later.
	Case 2: sBSI associated with a post-Whipple hepatic abscess	Case 2: abscess drainage and tigecycline plus colistin (doses not reported).	Case 2: died of septic shock at day 14.

BSI, bloodstream infection; CRBSI, catheter-related bloodstream infection; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; EVD, external ventricular drainage; IAI, intra-abdominal infection; MIU, million international units; PN, pneumonia; sBSI, secondary bloodstream infection; UTI, urinary tract infection; VAP, ventilator-associated pneumonia.

Treatment of PDR P. aeruginosa

Reports of successful treatment regimens against infections caused by PDR P. aeruginosa included: ceftolozane/tazobactam in a patient with ventilator-associated pneumonia,⁶⁰ high-dose IV amikacin (25–50 mg/kg) in one patient with intra-abdominal infection and one patient with pneumonia (amikacin MIC 16 mg/L),⁷⁰ and high-dose intraventricular amikacin in a case of ventriculitis (amikacin MIC=32 mg/L).⁵⁸ It is notable that both patients treated with high-dose IV amikacin had renal failure and continuous venovenous haemodiafiltration was concomitantly performed to prevent amikacin nephrotoxicity, allowing trough concentrations <5–10 mg/L.⁷⁰

Synergistic combinations may also be useful according to two studies.^73,75 Amikacin (1 g/day) with meropenem (2 g q8h infused over 3 h) resulted in both microbiological and clinical success in four cases of ventilator-associated pneumonia.⁷³ The combination was confirmed to be synergistic in vitro.⁷³ However, all isolates were intermediately susceptible to both amikacin (MIC 16 mg/L) and meropenem (MIC 8 mg/L),⁷³ and the high-dose prolonged infusion of meropenem may explain the efficacy of the regimen.⁸⁵ In another case series, four of the five patients with various infections were cured (both microbiological and clinical cure), three of which were treated with different potentially synergistic colistin-based combinations.⁷⁵ However, synergism was not confirmed.⁷⁵

Treatment of PDR A. baumannii

We found only one case series regarding the treatment of PDR A. baumannii.⁴ In 10 patients with ventilator-associated pneumonia, the combination of IV colistin, high-dose IV tigecycline (200 mg loading dose followed after 12 h by 100 mg q12h), high-dose IV ampicillin/sulbactam (6/3 g q8h) and inhaled colistin resulted in clinical success in nine patients, and microbiological eradication in seven.⁴ All patients were concurrently receiving empirical MRSA coverage (linezolid n=8, vancomycin n=1 and ceftaroline n=1), an important consideration as synergism between colistin and these agents has been described.^86–88

Treatment of other PDR GNB

The combination of ceftazidime/avibactam (MIC=16 mg/L) plus meropenem (MIC ≥256 mg/L) plus high doses of nebulized colistin successfully treated a post-transplant cystic fibrosis patient with PDR bacteraemic B. cepacia infection.⁵⁵ Ceftazidime/avibactam was synergistic in vitro with meropenem,⁵⁵ which is in agreement with another case report in PDR K. pneumoniae.⁶⁴

Another case series of ICU patients with bloodstream or urinary tract infections caused by PDR P. stuartii evaluated the combination of piperacillin/tazobactam (4.5 g q8h) plus amikacin (1 g q24h), a synergistic combination in vitro.⁶⁵ Follow-up cultures were sterile in all but one patient, but mortality was high (6 of the 10 patients with bacteraemia and 1 of the 5 patients with urinary tract infections died).⁶⁵

In vitro activity of newer agents

In vitro activity of newer agents against PDR isolates was reported in some studies; ceftolozane/tazobactam was active against all 7 PDR P. aeruginosa isolates in three studies,^12,42,60 whereas all 14 isolates in four other studies were resistant to ceftolozane/tazobactam.^10,31,83,89 Aztreonam/avibactam was active against two PDR and all XDR (n=111) Enterobacteriaceae in one study.²³ Ceftazidime/avibactam was active against selected PDR strains based on case reports (n=4, KPC-producing K. pneumoniae^53,59,63,64 and n=1, B. cepacia⁵⁵). Meropenem/vaborbactam was not active against the single PDR isolate (P. stuartii) identified in one study.²⁷ Finally, plazomicin was initially reported to have good activity (MIC≤2 mg/L) against 8 of 9 PDR Enterobacteriaceae in a study in Greece,⁴⁵ but in a subsequent study in the same region, 7 of 17 PDR K. pneumonia isolates were highly resistant to plazomicin (MIC>256 mg/L); n=7 were susceptible (≤2 mg/L) and n=3 had an MIC of 4 mg/L.¹¹

Discussion

Summary of main findings

Despite the rarity of pandrug resistance, the reviewed studies reflect increasing worldwide dissemination of PDR GNB in at least 25 countries in five continents. PDR GNB were mostly reported in ICU patients, but significant outbreak potential and dissemination were demonstrated, including international spread. Among the PDR pathogens detected by this review, A. baumannii, K. pneumoniae and P. aeruginosa are the most common species reported to be PDR, whereas PDR E. coli remain exceedingly rare.

All-cause mortality in patients with PDR GNB infection is high. Although the extent of attributable mortality was unclear, reviewed studies indicated that mortality might be substantially reduced by treatment regimens active in vitro. Newer agents, such as ceftolozane/tazobactam, ceftazidime/avibactam and plazomicin, appear to be active against some GNB strains that are resistant to all older antimicrobials.^{11,12,42,45,53,59,60,63,64} However, strains that are PDR even to the newer agents have been reported.^{10,11,31,45,83,89} Other options for the treatment of PDR GNB infections include synergistic combinations^{4,5,55,57,65,68,69,73,75} and increased exposure treatment regimens to achieve pharmacokinetic/pharmacodynamic (PK/PD) targets.^58,70 However, current evidence remains limited, the risk of bias is high and head-to-head trials of different treatments are lacking.

Synergistic combinations for PDR infections

Several synergistic combinations have been successfully used in PDR GNB infections, such as double carbapenem^5,57,68 or double carbapenem with colistin^57,66 for KPC-producing K. pneumoniae, ceftazidime/avibactam with carbapenems for K. pneumoniae⁶⁴ and B. cepacia,⁵⁵ and high-dose ampicillin/sulbactam with meropenem and colistin for A. baumannii.^4,90,91 It is notable that neither double carbapenem nor ceftazidime/avibactam are active against MBLs. This has important implications considering that a significant percentage of carbapenem-resistant isolates in some areas are MBL producers.^11,92–94 In contrast, the combination aztreonam/avibactam can restore activity against MBL-producing isolates,^95,96 since aztreonam is not hydrolysed by MBLs and avibactam effectively inhibits other β-lactamases (including ESBLs, KPC and OXA-48), therefore restoring the activity of aztreonam. Aztreonam/avibactam is not currently available; however, the combination of ceftazidime/avibactam plus aztreonam has been used successfully against infections caused by MBL-producing bacteria.^97–99

Other less-studied combinations may also be useful. The in vitro synergy of colistin with agents such as linezolid, vancomycin and teicoplanin in colistin-resistant GNB strains is notable.^86,87 Fosfomycin combined with meropenem also appears promising and high cure rates (7/10 patients) have been reported even for fosfomycin-resistant isolates.¹⁰⁰ Despite the availability of several in vitro studies on synergistic combinations, in vivo studies, such PK/PD and outcome studies, are lacking.¹⁰¹

Revival of old antibiotics

The emergence of MDR/XDR bacteria has led to the revival of older antibiotics, with colistin being a good example.¹⁰² In PDR infections, colistin is often used in synergistic combinations with other antibiotics.^4,57,66,72 Nebulized colistin, allowing higher epithelial lining fluid concentrations than IV colistin,^103,104 may be useful for PDR respiratory infections and has been used as part of synergistic combinations.⁴ However, data on PDR infections to allow comparison of inhaled versus IV colistin are lacking. Based on the available data from colistin-susceptible infections, the addition of nebulized colistin to IV colistin has been associated with improved outcomes^105–107 and the efficacy of nebulized colistin (without concomitant IV colistin), either as monotherapy or in combination with other antibiotics, has been found to be similar to IV colistin and has been associated with lower nephrotoxicity.^108,109

Another old antibiotic of renewed interest is IV fosfomycin. Although largely unavailable outside Europe, IV fosfomycin appears to be an effective treatment option for antibiotic-resistant Enterobacteriaceae^110,111 and has been used successfully for infections resistant to all other options.^101,112 However, we noted a lack of reporting of fosfomycin susceptibility in several studies in this review (Table S4). Furthermore, there are concerns regarding the in vivo activity of fosfomycin against P. aeruginosa (even when susceptible in vitro) and the risk of emergence of resistance during treatment.¹¹³ Minocycline has also been proposed as an option for resistant bacteria¹¹⁴ but susceptibility data for minocycline were not reported in most of the reviewed studies (Table S4). To guide the use of these old antimicrobials in clinical practice, more evidence is required from modern PK/PD studies and randomized controlled trials.^102,115,116

Reconsideration of the PDR definition

Until recently, isolates with intermediate susceptibilities were interpreted as non-susceptible.⁸ In 2019, EUCAST replaced the term ‘intermediate’ with ‘susceptible, increased exposure’ to indicate that there is high likelihood of achieving therapeutic success by increasing exposure to the antimicrobial. For example, high-dose prolonged infusion meropenem may be effective for K. pneumoniae with an MIC 4–8 mg/L.⁸⁵ Additionally, two studies on PDR P. aeruginosa infection demonstrated successful treatment using higher doses of amikacin.^58,70 Another point to consider when defining PDR is the potential synergism between agents that are inactive alone but active when used in combination.

Limitations

Reported PDR proportions in this review may be overestimated because we excluded studies reporting no PDR isolates and because the lists of antimicrobials tested for susceptibility were incomplete in several studies (Table S4). In contrast, the use of methods other than broth microdilution for the evaluation of susceptibility to colistin may result in underestimation of the proportion of PDR because these methods result in a high rate of false susceptibility.^117,118

Furthermore, although our findings indicate the worldwide spread of PDR GNB, they do not accurately reflect their geographical distribution. Most non-eligible studies originated from Asia, mainly from China (Tables S5 and S6), resulting in under-representation of these areas in this review. Additionally, we cannot exclude the possibility of over- or under-reporting of PDR GNB in some countries.

Finally, despite our efforts to decrease language bias using Google Translate, data extraction may have been inaccurate.¹¹⁹ Nevertheless, for studies requiring full-text review using Google Translate we believe that their exclusion was reliable as it was based on the list of agents used for susceptibility testing.

Conclusions

PDR GNB isolates are increasingly being reported worldwide, and several studies have demonstrated their potential for intra- and inter-institutional, and even international, dissemination. All-cause mortality following PDR GNB infection appears to be high irrespective of the infecting organism, but the extent to which mortality is attributable to the infection remains unclear. Despite the lack of controlled trials, several treatment regimens have been reported to be effective against PDR GNB infections and the reviewed studies indicated that mortality might be substantially reduced by treatment regimens active in vitro. These include newer agents and increased exposure regimens, but most studies reporting successful treatment of PDR GNB infections used synergistic combinations. Synergistic combinations are often the only treatment option for PDR GNB infections and therefore more research is required, including PK/PD and outcome studies. Considering the rarity of PDR GNB, multicentre studies are necessary.

Funding

This study was supported by internal funding

Transparency declarations

None to declare.

Author contributions

Conception and design of the study: S.K., E.I.K., A.G. Literature searches and data extraction: S.K. Analysis and interpretation of data: S.K., E.I.K., A.G. Drafting the article: S.K. Critical revisions of the article: E.I.K., A.G. Approval of the final version of the article: S.K., E.I.K., A.G.