A total of 216 patients had pyogenic liver abscesses during the study period. However, after analyzing all medical charts, we found that 199 pyogenic liver abscesses were definitively identified; these abscesses were further studied. Among the 199 abscesses, a microbiological diagnosis was available for 158 abscesses, which constitute the study cohort (Figure 1). Pyogenic liver abscesses were cryptogenic in 27 (17%) of 158 patients. K. pneumoniae was isolated from 31 (20%) of 158 patients and was more frequent in patients with cryptogenic pyogenic liver abscesses (14/27, 52%) than in those with noncryptogenic pyogenic liver abscesses (17/131, 13%; p = 0.00003). K. pneumoniae was the most common causative microorganism in the group with cryptogenic pyogenic liver abscesses (Figure 2).

We analyzed characteristics of patients having K. pneumoniae cryptogenic and noncryptogenic liver abscesses (Table 1). A total of 13/17 patients with K. pneumoniae noncryptogenic pyogenic liver abscesses had a biliary tract origin, and 2/17 showed development of pyogenic liver abscesses after liver transplantation (1 patient after cyst punction and 1 patient after chemoembolization). Healthcare-related infections (12/17 vs. 0/14; p<0.00001) and cancer (7/17 vs. 0/14; p<0.01) were significantly more frequent for patients with noncryptogenic pyogenic liver abscesses than for patients with cryptogenic pyogenic liver abscesses. We found a tendency toward higher C-reactive protein levels for patients with K. pneumoniae cryptogenic pyogenic liver abscesses. We observed 7 relapses (41%) for the 17 patients with K. pneumoniae noncryptogenic pyogenic liver abscesses, compared with none for the 14 patients with cryptogenic pyogenic liver abscesses (p<0.01) (Table 1).

In the cryptogenic group, 3 patients were of Asian origin, compared with 1 in the noncryptogenic group. Three patients had severe septic metastases. The first of these patients was a 64-year-old woman with diabetes from Sri Lanka who had K1-type K. pneumoniae pyogenic liver abscesses, endophthalmitis, and brain abscesses. She survived after enucleation treatment. The second patient was a 57-year-old man with diabetes from Morocco who had K2-type K. pneumoniae pyomyositis associated with pyogenic liver abscesses. He was cured after 6 weeks of treatment with antimicrobial drugs. The third patient was a 33-year-old man from France who did not have any previous medical condition, but in whom K1-type K. pneumoniae acute tibial periostitis associated with pyogenic liver abscesses developed. He was cured after 6 weeks of treatment with antimicrobial drugs.

We studied 199 patients with liver abscesses managed in a single hospital during 2010−2015 and analyzed in detail 158 patients with microbiologically proven pyogenic liver abscesses. Although our study was retrospective, it had the advantage of particularly homogeneous management in a monocentric university hospital specializing in gastroenterology, liver diseases, and abdominal surgery. The main finding was that hypervirulent K. pneumoniae are now responsible for most cryptogenic liver abscesses in patients in the Paris region in France.

Pyogenic liver abscesses caused by hypervirulent K. pneumoniae represented 14 (8.9%) of 158 liver abscesses, which is more frequent than previously described in Europe (16). Although hypervirulent K. pneumoniae have only rarely been described in Europe and North America (18,21,22,28,31), our results indicate that hypervirulent K. pneumoniae pyogenic liver abscesses might be considered as an emerging infection in the Paris area and should always be considered in the diagnosis of patients with pyogenic liver abscesses of no obvious origin. This factor is a key point because infections are often severe, metastatic locations might impair prognosis (especially for meningeal or ocular involvement), and immediate treatment with antimicrobial drugs is required.

We found a relatively low proportion (17%) of cryptogenic pyogenic liver abscesses, which might be related to the conditions of our hospital, which is a highly specialized tertiary care hospital. However, similar rates of cryptogenic abscesses were reported in the United Kingdom (7/42, 16.7%) (32). This finding suggests that epidemic diffusion of hypervirulent K. pneumoniae in Europe has not yet reached its peak, in contrast to what was observed in other regions, because cryptogenic pyogenic liver abscesses now represent as many as 34% of pyogenic liver abscess cases in Australia (33) and 65% of pyogenic liver abscess cases in Asia (34).

All K. pneumoniae isolates from these cryptogenic abscesses were hypermucoviscous (defined as a positive string test result) and belonged to K1 or K2 serotypes, thus having the characteristics of hypervirulent K. pneumoniae. Similar to Ye et al. (35), we observed a clonal diffusion of ST23 K1-type hypervirulent K. pneumoniae, in contrast to K2-type hypervirulent K. pneumoniae, which belong to different sequence types (ST65, ST86, ST380, and ST679) (35).

One patient with noncryptogenic pyogenic liver abscesses was also infected with a strain harboring phenotypic and genotypic characteristics of hypervirulent K. pneumoniae. This finding indicated that hypervirulent K. pneumoniae might also be involved in noncryptogenic pyogenic liver abscesses.

From a clinical point of view, our description of hypervirulent K. pneumoniae−related pyogenic liver abscesses is consistent with previous reports. These abscesses were mostly monomicrobial, rarely harbored organisms that showed antimicrobial drug resistance, occurred in persons with few concurrent conditions, were not healthcare-related, and were more likely responsible for aggressive inflammatory disease with metastatic infectious locations (17,36,37). Unlike a previous report (17), diabetes did not appear to be a risk factor for hypervirulent K. pneumoniae pyogenic liver abscesses, and Asian patients did not represent the main ethnic group (only 21% in our study vs. 50% in cohorts in the United States). In addition, hypervirulent K. pneumoniae−related pyogenic liver abscesses were usually single and located in the right hepatic lobe (17).

Although hypervirulent strains were involved and metastatic locations were frequent, outcomes were more favorable than for patients with noncryptogenic K. pneumoniae pyogenic liver abscesses, probably because patients with hypervirulent K. pneumoniae−related pyogenic liver abscesses were less likely to be immunosuppressed or have cancer. Thus, host factors seem to outweigh bacterial determinants in the prognosis of pyogenic liver abscesses (38). However, a substantial mortality rate (4%–10%) caused by these hypervirulent strains occurred and should not be underestimated (17,36–38). As found in our study, antimicrobial drug resistance was rarely an issue for these hypervirulent strains. Nevertheless, ESBL- and carbapenem-resistant hypervirulent K. pneumoniae have been reported in Taiwan, China, India, and France and might represent an emerging problem in the future (39–42).

The microbiological definition of hypervirulent K. pneumoniae is still ambiguous because none of the phenotypic (string test) or genotypic (capsular serotype and virulence genes) tests alone is specific for hypervirulence (43–46). However, the combination of clinically aggressive monomicrobial cryptogenic pyogenic liver abscesses and isolation of positive string test−positive K. pneumoniae isolates with K1/K2 capsular serotypes and these genes involved in hypervirulence (e.g., prmpA and iutA) is highly suggestive of hypervirulent K. pneumoniae−related disease. We believe that it is the combination of this typical clinical and microbiological presentation that defines what might be called the hypervirulent K. pneumoniae syndrome.

Our study had several limitations. First, this study was monocentric and retrospective, which might limit its epidemiologic scope. Second, information concerning patients’ travel history was missing. Third, because our center is highly specialized in liver diseases, cryptogenic hypervirulent K. pneumoniae pyogenic liver abscesses might be underestimated, and cancer or postsurgical pyogenic liver abscesses are probably overrepresented.

As the main microorganism responsible for cryptogenic pyogenic liver abscesses, hypervirulent K. pneumoniae represents an emerging microorganism in the Paris area. The clinical and molecular differences between K. pneumoniae implicated in cryptogenic and noncryptogenic pyogenic liver abscesses confirm the correlation between clinical and microbiological virulence and define a hypervirulent K. pneumoniae syndrome. This finding is a major epidemiologic shift that should be considered in the diagnostic and therapeutic management of patients with pyogenic liver abscesses without underlying or obvious causes. Given the rapid increase in the prevalence of hypervirulent K. pneumoniae pyogenic liver abscesses in other parts of the world, it is possible that France and Europe might experience the same epidemic evolution as occurred in Asia in the early 2000s. Thus, clinicians need to be prepared for this possibility.