In this register-based cohort study describing the surveillance of MDROs among persons diagnosed with cancer, we found that 0.3% were registered with MRSA, V/LRE, or CP-GNB. In a sub-cohort of persons diagnosed or treated at OUH, we found that 6% of Enterobacterales cultured from the urine or blood of persons with cancer were resistant to third-generation cephalosporins. We found that persons with cancers of the lymphoid or haematopoietic tissues had the highest incidence of MDRO infection/colonisation and the highest proportion of resistance in Enterobacterales infection/colonisation compared to persons with other types of cancer.

In our work, we have demonstrated how it is possible to expand the register-based surveillance at the national level to include both potentially susceptible patient populations using existing register data and organisms like 3GCR-E using the richer microbiological laboratory data from a hospital. 3GCR-E is not nationally notifiable in Norway, yet is responsible for a large proportion of the total disease burden of AMR [3]. As the capture and linkage of existing regional laboratory data and national register data may be automated, it may be possible to expand surveillance without increasing the resource requirements [15]. Previous work has shown how it is possible to expand surveillance to detect outbreaks and clusters in the healthcare setting through the use of automated algorithms to capture and link existing register data [16, 17]. The challenges posed to such an expansion of surveillance include the need for standardised protocols to capture heterogenous laboratory data from multiple sources and ensure the privacy of the patient [15].

The proportion of colonisation and the development in the trends of notifications were similar to a previous study of the epidemiology of MRSA, V/LRE, and CP-GNB in the whole population of Norway, 2006–2017 [2]. Elstrøm and colleagues concluded that incidence rates were increasing up until 2017. It is uncertain whether this trend is carried forward due to disruptions from the pandemic. Compared to the incidence rates of Elstrøm et al. and publicly available surveillance data found when selecting the same categories of MDROs and years on the website of the Norwegian Institute of Public Health [18], our cohort of persons diagnosed with cancer has more than three times the rate of V/LRE, but only about a fourth the rate of MRSA [2, 7]. The most likely reasons for these discrepancies could be that V/LRE in Norway has predominantly been associated with large outbreaks in healthcare institutions with extensive contact tracing and screening [2, 19,20,21].

One could hypothesise that the higher V/LRE incidence was not only a consequence of heightened susceptibility to infection, but also of the time spent in healthcare. This may have led to both an increased probability of being implicated in ongoing nosocomial outbreaks, and as such of being tested and found positive, and a selection for V/LRE colonisation in patients who receive frequent antibiotic treatment, in particular those with cancers of the gastrointestinal tract, as supported by the incidence rate ratios [22]. The higher rates of discovered V/LRE colonisation in persons with cancers of the lymphoid and haematopoietic tissues and persons with cancers of unknown origin (often metastasised at the time of diagnosis) are in accordance with this hypothesis, due to their frequently prolonged hospital stays. The increased finding of resistant bacteria directly after cancer diagnosis may also reflect screening on admittance or prior to procedures, rather than acquisition at that time. We also tested a secondary hypothesis that persons with cancer and who were healthcare workers were more often infected or colonised with resistant bacteria than persons with cancer and other occupations, as they were potentially more exposed to the bacterial flora in the healthcare setting but found no indication of this. As V/LRE colonisation is a risk factor of the development of V/LRE bacteraemia, and many persons with cancer are vulnerable to infections, the increased rate of colonisation in this group is concerning and underlines the importance of reducing the prevalence of such bacteria in the healthcare setting [23, 24].

MRSA, on the other hand, has previously been shown to be predominantly acquired in the community or associated with travel from abroad [2, 21, 25]. If people living with cancer defer from travelling and shield themselves, either consciously to avoid complicating infections during cancer treatment or because they are too ill to participate in the community, their exposure to situations associated with MRSA (and CP-GNB) acquisition may decrease. The fact that a lower proportion of notifications were reported with a place of transmission outside of Norway for persons with cancer compared to the general population supports that persons with cancer defer from travelling to some extent [18, 25]. Although there are few studies on the epidemiology of resistant Gram-positive bacteria in national cancer cohorts, compared to studies using clinical data, the incidence rates of in our cohort were low [26].

We found that although Enterobacterales found in the blood or urine of persons with cancer at OUH have a higher proportion of resistance to the third-generation cephalosporins cefotaxime and ceftazidime than the national average of Klebsiella spp. and Escherichia spp. found in blood or urine as reported in the NORM-NORM/VET surveillance, the resistance proportion was similar to the mean of the whole hospital [7, 27]. This could be because OUH is Norway’s only tertiary hospital, and only specialist cancer hospital and transplantation centre. This means that the hospital admits patients with higher or more complex comorbidity than other regional hospitals and that patients may be transferred to OUH after lengthy stays in other institutions, which may increase the risk of antimicrobial resistance. Furthermore, we found a higher resistance proportion of 3GCR-E in patients with cancers of the lymphoid or haematopoietic tissues, which may reflect that the nature of the immunosuppression or specific antibiotic regimens given to these patients selects for this type of antimicrobial resistance. While resistance to third-generation cephalosporins appears to be steadily increasing in Norway, we found a lower proportion of such resistance overall in our study cohort compared to the averages in the European general population found in the EARS-NET reports [1, 7, 28].

CP-GNB was also increasing throughout the study period. However, there were no reported CP-GNB in our study cohort in 2021, during the COVID-19 pandemic. This may not have been an effect of the pandemic alone, but also due to the gradual depletion of observation time in 2021, as only persons diagnosed with cancer in 2018 could be followed into 2021. It could also be related to the limited travel during the pandemic, as the sporadic cases of CP-GNP in Norway are commonly associated with import, demonstrated by the fact that half of the CP-GNB in our study were reported to be acquired abroad. This association was also reflected by the sharp decline in CP-GNB notifications during the COVID-19 pandemic in the general population [18]. Other countries have reported an increase in typically healthcare-associated or emerging carbapenem-resistant Gram-negative bacteria during the pandemic, like Acinetobacter baumannii [29, 30]. However, a study performed in an endemic setting indicated a decline in carbapenem-resistant Klebsiella pneumoniae infections, potentially as a result of COVID-19 prevention measures [31]. The potential of endemicity may be due to both successful clones and the potential for rapid, horizontal transferring of plasmid-mediated resistance such as carbapenemases and extended-spectrum beta-lactamases, with several examples globally [32]. With this potential in mind, prevention of resistant Gram-negative bacteria with mobile genetic elements should be intensified in the coming years, particularly with the reopening of international travel. When such bacteria become widespread in the healthcare setting, it may prove impossible to reverse the situation.

The strengths of our study are that we present notifications of resistant microbes in a nationwide and complete cohort of persons diagnosed with cancer. For persons diagnosed or treated at OUH, we have complete microbiological data, allowing us to include bacteria that are not nationally notifiable. In addition, we retrieved occupational data, allowing us to test a secondary hypothesis. However, we could not know whether persons with cancer were actively working as healthcare workers in the full observation time, or if they quit or were on extended sick leave. It is also important to consider our open cohort structure when interpreting the data, in which observation time gradually accumulates at the beginning and gradually depletes towards the end of the study period. The incidence rates in those years are therefore calculated in a few persons, meaning interpretation should be done with care. We have also only included the infectious diagnosis closest in time to the cancer diagnosis, as the same microbe was often found multiple times. Further, we were not able to present results and discuss the clinical outcomes of colonisation or infection with resistant microbes, as we did not retrieve clinical data. In the OUH sub-cohort, we did not have a complete denominator allowing us to calculate the incidence rate of Enterobacterales among all persons with cancer who had any part of their diagnostic workup or treatment at the hospital. Although exceedingly rare before that year, CP-GNB were not notifiable prior to 2012, leading to a possible underestimation early in the study period. Finally, the data on resistant microbes are only representative of the tests performed and were thus sensitive to changing guidelines and routines for screening, or special situations like ongoing outbreak investigations.

In a cohort of persons diagnosed with cancer in Norway between 2008 and 2018, we found that the incidence rate of resistant Gram-positive bacteria has not increased in recent years although we observed more of the often healthcare-associated V/LRE and less of the often community-associated MRSA compared to previously published data on the general population. It may seem like infection prevention and control measures aimed at containing (resistant) Gram-positives are succeeding. However, we observed an unfavourable trend in both incidence rates and resistance proportions of Gram-negative bacteria. Furthermore, we found both the highest incidence rates and resistance proportions in patients with cancers of the lymphoid or haematopoietic tissues, possibly reflecting specific healthcare exposures like prolonged hospitalisation or antibiotic treatment. Regardless of what is causing the increase, infection prevention efforts aimed at controlling Gram-negative bacteria with the potential for acquiring extensive drug resistance should be intensified to protect potentially susceptible groups like persons living with cancer. We have demonstrated that it is possible to expand the surveillance of infectious diseases at the national level to both include such potentially susceptible populations and non-notifiable pathogens by combining existing data sources.