|Home | About | Journals | Submit | Contact Us | Français|
Listeriosis is a rare but severe food-borne disease, affecting unborn or newly delivered infants, the elderly, and the immunocompromised. The epidemiology of listeriosis in England and Wales changed between 2001 and 2007, with more patients ≥60 years old presenting with bacteremia (but without central nervous system [CNS] involvement). In order to explain this increase and understand the altered disease presentation, clinical, microbiological, and seasonal data on bacteremic cases of Listeria monocytogenes infection identified through national surveillance were compared with those for patients with CNS infections. Logistic regression analysis was applied while controlling for age. Bacteremic patients, who presented more frequently with gastrointestinal symptoms, were more likely to have underlying medical conditions than CNS patients. This was most marked in patients with malignancies, particularly digestive organ malignancies. Treatment to reduce stomach acid secretion modified the effect of nonmalignant underlying conditions on outcome, i.e., patients with an underlying condition who were not taking acid-suppressing medication were equally likely to have a bacteremic or a CNS infection. However, this type of therapy did not modify the effect of malignancies on the likelihood of having a bacteremic or a CNS infection. The increase in the incidence of human listeriosis among patients ≥60 years old in England and Wales between 2001 and 2007 appears to have occurred in those with cancer or other conditions whose treatment included acid-suppressing medication. Therefore, this vulnerable patient group needs specific dietary advice on avoiding risk factors for listeriosis.
Listeria monocytogenes is an opportunistic bacterial pathogen that causes listeriosis and most often affects the immunocompromised, the elderly, pregnant woman, and their unborn or newly delivered infants. The disease is transmitted predominantly via contaminated food and is estimated to be the greatest cause of food-related deaths in the United Kingdom (7). A large outbreak of listeriosis, affecting mostly pregnant women and associated with the consumption of imported pâté, occurred in the United Kingdom in the late 1980s (16). Consequently, specific advice provided to pregnant women and immunocompromised individuals on foods to avoid in order to minimize the risk (Department of Health, Advice to vulnerable groups on pâté stands, press release 189/369, 1989; Department of Health and Social Security, Advice from the Chief Medical Officer: listeriosis and food, DHSS PL/CMO 89, 1989), has subsequently been reiterated and preferentially targeted at pregnant woman (Food Standards Agency, http://www.eatwell.gov.uk/agesandstages/pregnancy/?lang=en).
The epidemiology of listeriosis in England and Wales changed between 1990 and 2007 (9). The incidence almost doubled (an average of 191 cases were reported annually between 2001 and 2007 versus 110 between 1990 and 1999), with the increase occurring mainly among patients aged ≥60 years presenting with bacteremia in the absence of central nervous system (CNS) infection (Fig. (Fig.1).1). These changes are independent of recognized outbreaks, gender, season, ethnicity, socioeconomic status, region, or L. monocytogenes subtype and are not thought to be artifactual. Similar patterns have been reported subsequently in other European countries (4, 10).
The purpose of this study was to identify clinical and epidemiological factors that might explain this increased incidence and altered disease presentation by interrogating surveillance data for listeriosis cases reported in England and Wales between 2001 and 2007.
The Health Protection Agency Centre for Infections coordinates the surveillance of listeriosis in England and Wales. Cases are ascertained by the voluntary electronic reporting of laboratory-diagnosed cases from microbiology laboratories (approximately 80% of instances), by the voluntary referral of cultures for identification and subtyping (approximately 80% of instances) (5, 19), or by both means (approximately 60% of instances). Epidemiological and microbiological data from the two systems are combined, deduplicated, and stored in a bespoke electronic database. Additional clinical data are sought from the consultant medical microbiologist responsible for each case, using a standard questionnaire (12) modified in 2005 to include symptom data.
Cases of listeriosis were defined as those with clinically compatible illnesses where L. monocytogenes was isolated from normally sterile sites, usually blood or cerebrospinal fluid. Pregnancy-associated cases (all maternal-fetal and neonatal patients) were not considered in this analysis. Non-pregnancy-associated cases (illness in patients >1 month old) were categorized further into those with CNS infections (isolation of L. monocytogenes from cerebrospinal fluid or brain tissue, clinical evidence of CNS infection, or both), those with bacteremia in the absence of CNS infections (isolation of L. monocytogenes from blood but without microbiological or clinical evidence of CNS infection), and those with other conditions not included in these two categories.
Data manipulation and statistical analysis were undertaken using Stata, version 10, and were restricted to non-pregnancy-associated cases reported in England and Wales between 2001 and 2007, where a clinical questionnaire was returned and where the patient presented with a CNS infection or bacteremia. The descriptions of patients’ underlying condition were grouped by two authors (I.A.G. and C.P.), and cancers were classed further according to the International Statistical Classification of Diseases and Related Health Problems, version 10 (22). More than one underlying condition can be reported for each patient. The study date was derived using onset and specimen dates as available (70% and 30%, respectively, with a median lag of 1 day between these dates [range, 0 to 33 days]). The season (winter [December to February], spring [March to May], summer [June to August], or autumn [September to November]) was defined using the study date. The lag time to clinical investigation was calculated from the date of onset to the date of the first specimen collection or the admission date as available, and these lag times were grouped as either as ≤1 day or ≥2 days, based on a median of 1 day (range, 0 to 28 days). Differences in proportions were assessed using the chi-square test and Fisher's exact test, and the nonparametric test was used for the comparison of medians. Individuals with missing data were omitted from the analyses involving those data.
An outcome variable was created to compare bacteremic infections with CNS infections, and single-variable analysis was initially used to investigate the associations between this outcome and the reported clinical features. Because the increase in the incidence of bacteremia occurred predominantly in patients aged ≥60 years, logistic regression was applied to calculate odds ratios (ORs), 95% confidence intervals (CIs), and significance tests while controlling for age. Variables associated with the outcome at a level of >90% (i.e., P < 0.1) were considered for further investigation. Logistic regression was applied to obtain maximum-likelihood estimates of the effect of exposures on the outcome while controlling for potential confounding factors. The model was simplified in a stepwise manner using the likelihood ratio test. Interactions between the factors included in multiple-variable analyses were investigated in the same way. Patient age was retained in the model throughout.
Clinical questionnaires were received for 925 of 1,167 non-pregnancy-associated listeriosis cases reported in England and Wales between 2001 and 2007 (79% response rate). The response rate was high (i.e., >70%) in each study year except 2004, when only 99 of 190 questionnaires were received (52% response). Data from this year were excluded from further analysis, giving an 85% response rate (826/977) for the remaining years. Twenty-seven cases (3%) reported during this time were linked to recognized clusters or outbreaks.
Bacteremia (693 cases) and CNS infections (224 cases) made up 94% of all cases; for the remaining 60 cases, either L. monocytogenes was isolated from other sites (ascitic fluid, joint aspirates, and other sites) (40 cases) or the cases could not be classified (20 cases). Questionnaire receipt was independent of age (means, 67 and 66 years) and gender (57% and 50% male; P = 0.13 by the χ2 test), but questionnaires were received less frequently for bacteremic infections (571/693 [82%]) than for CNS infections (207/224 [93%]) (P < 0.001 by the χ2 test). Among cases of bacteremic infection, questionnaire receipt was independent of age (means, 70 versus 67 years; P = 0.06 by the t test). Mortality data were available for 748/778 patients with clinical questionnaires (96%), and death rates were similar for bacteremic infections (216/546 [40%]) and CNS infections (89/202 [44%]) within this group (P = 0.27 by the χ2 test).
The ages of patients with bacteremia ranged from 3 to 102 years (median, 72 years; interquartile range, 61 to 80 years), while those of patients with CNS ranged from <1 year to 97 years (median, 65 years; interquartile range, 52 to 74 years)—a significant difference (P < 0.001). The single CNS patient less than 1 year old was a 57-day-old infant, and therefore the infection did not meet the criteria for being considered pregnancy associated. Within the bacteremic group where mortality data were available (n = 546), there was no difference between those aged <60 years and those aged ≥60 years with regard to mortality (36/109 [33%] versus 180/437 [41%], respectively; P = 0.12 by the χ2 test). Where mortality data were available for CNS cases, however (n = 202), patients aged ≥60 years experienced higher mortality than those aged <60 years (71/132 [54%] versus 18/70 [26%]) (P < 0.001 by the χ2 test).
On single-variable analysis controlling for age, bacteremic cases were no different from CNS cases in terms of gender, season, L. monocytogenes serotype, or lag period between the onset of symptoms and clinical investigation (Table (Table1).1). Bacteremic patients were more likely to have an underlying condition (most noticeably malignancies, and digestive-organ malignancies in particular) than CNS patients but were less likely to have alcohol-related conditions. Bacteremic patients were more likely to be treated with cytotoxic drugs or to receive treatment to reduce stomach acid secretion. There was no difference between bacteremic and CNS cases with regard to treatment with immunosuppressive drugs or steroids.
Initial multiple-variable logistic regression analysis (“model 1”) revealed that the effect of alcohol-related conditions was no longer significant (OR, 1.5; 95% CI, 0.9 to 2.8; P = 0.16) when the analysis was controlled for all other nonmalignant underlying conditions (OR, 2.4; 95% CI, 1.4 to 4.0; P < 0.001) and for malignancies (OR, 3.0; 95% CI, 1.7 to 5.4; P < 0.001). Malignancies were examined further by comparing malignancies of the digestive organs with other “known” malignancies (i.e., excluding ill-defined, secondary, and unspecified neoplasms, or those of uncertain or unknown behavior) (“model 2”); both remained independently associated with the outcome when the analysis was controlled for nonmalignant underlying conditions and patient age (ORs, 17.0 [95% CI, 3.9 to 74.3; P < 0.001] and 3.0 [95% CI, 1.8 to 5.0; P < 0.001], respectively). When included in this model, neither the effect of cytotoxic drugs (OR, 1.6; 95% CI, 0.9 to 3.0; P = 0.11) nor treatment to reduce stomach acid secretion (OR, 1.4; 95% CI, 0.8 to 2.4; P = 0.23) remained significant. However, the latter interacted significantly with malignancies, and therefore its effect on each stratum was investigated further (“model 3”) (Table (Table2).2). Treatment to reduce stomach acid secretion modified the effect of nonmalignant underlying conditions on the outcome of interest, in so far as patients who had an underlying condition but did not have treatment to reduce stomach acid secretion were equally likely to have a bacteremic or a CNS infection. However, treatment to reduce stomach acid secretion did not modify the effect of malignancies on the likelihood of having a bacteremic or a CNS infection.
Fever was the single most common symptom reported by both patient groups (77% and 83% for bacteremic and CNS infections, respectively [Fig. [Fig.2]).2]). Bacteremic patients were more likely to experience abdominal pain, respiratory or cardiovascular symptoms, and diarrhea. Patients with CNS infections were more likely to have headache, confusion, impaired consciousness, nuchal rigidity, seizures, and myoclonus. Logistic regression analysis revealed that the associations with abdominal pain, diarrhea, respiratory or cardiovascular symptoms, nuchal rigidity, and seizures remained independently significant when the analysis was controlled for underlying conditions and malignancies as described above (“model 2”).
Reported mortality for bacteremia patients who experienced abdominal pain, nausea, or diarrhea did not differ from that for patients who did not report these symptoms (Table (Table3).3). The same was observed for CNS patients, although lower numbers were available for comparison.
This study highlights hitherto unrecognized differences between the clinical presentations of L. monocytogenes infections in England and Wales. In the absence of point source exposures, it is impossible to measure the incubation period in individuals, but these data suggest that disease presentation occurs contemporaneously for bacteremic and CNS infections. Given this, the CNS infection is unlikely to be a sequela of bacteremia for most patients in this study population. While it is acknowledged that less severe presentations of the disease (e.g., influenza-like illness or gastrointestinal disturbance) might be underreported, this study suggests that the severity of disease for patients who experienced gastrointestinal symptoms was similar to that for patients who did not.
Bacteremic patients were more likely to have underlying conditions than those with CNS infections, and specific malignancies (especially digestive-organ malignancies) were especially prevalent in the bacteremia group. The former observation concurs with Danish (8) and French (11) studies of listeriosis mortality, while the latter concurs with previous United Kingdom (15) and French (11) studies, and with more-recent findings from France (10), although leukemic patients predominated in the latter French study. The disparity in presentation by underlying pathology might reflect the degree of medical surveillance, the rapidity of onset, or the severity of infection in this patient group, since bacteremia in previously healthy individuals will be recognized only under unusual circumstances when blood cultures are collected (e.g., during outbreaks ), or when the disease is of high severity.
The higher prevalence of gastrointestinal symptoms in the bacteremia group is of note. It is not clear, however, if these symptoms result from the listeriosis, pathogenicity changes in the microorganism, another infectious agent, or patients’ concurrent pathologies or treatments. Not all L. monocytogenes strains can cause gastrointestinal disturbance (2, 14), and since the current increase is caused by many different strains (9), it is unlikely that a single genetic change in L. monocytogenes has occurred. It has been suggested previously that another infectious or toxic agent leads to increased susceptibility to L. monocytogenes infection (21). Surveillance data from England and Wales provide little evidence for this: of 614 cases of L. monocytogenes infection reported electronically between 2004 and 2007, none had a reported gastrointestinal coinfection as part of the listerial episode or an antecedent gastrointestinal infection in the previous 90 days (I. A. Gillespie, unpublished observations). Many enteric pathogens are underdiagnosed in fecal specimens (1), however, and feces from listeriosis cases are rarely investigated; hence, a gastrointestinal coinfection would be detected less frequently in this patient group. Finally, existing pathologies (or their treatments) may alter the alimentary canal in such a way as to predispose to gastrointestinal symptoms, which may or may not have an infectious etiology. Such phenomena might explain the observed association between digestive-organ cancers and the group experiencing more gastrointestinal symptoms.
In comparison to patients’ underlying conditions, information on patients’ treatments was reported less frequently, and steps are required to improve the completeness of this information in routine national surveillance of listeriosis. Nevertheless, the observed associations with patient treatments warrant further comment. The association between cytotoxic drug use and bacteremia is perhaps unsurprising given the association with malignancies described above. This effect diminished when malignancies were controlled for, suggesting either that, in this instance, the malignancy is more important than the treatment itself, that a wide variety of cytotoxic treatments was described, or that missing data played a part.
Treatment to reduce stomach acid secretion was associated with bacteremic infections on single-variable analysis and modified the effect of underlying conditions on logistic regression analysis, missing data aside. Stomach acid inhibitors are used to treat a variety of conditions (13), but this drug class increases the risk of acquiring Campylobacter, Giardia, and Salmonella species infections (3, 17, 18) and increases susceptibility to listeriosis in animal models (20). Furthermore, the incidence of L. monocytogenes bacteremia in England closely mirrors overall prescribing patterns for proton pump inhibitors in England (Fig. (Fig.3);3); the number of such prescriptions increased by 257% between 1998 and 2007. Further work to investigate the role of proton pump inhibitors in human listeriosis is required. If an effective stomach acid barrier is absent, however, lower L. monocytogenes levels in foods may increase the likelihood of infection, and efforts should be made to reduce the exposure of vulnerable groups. The current European legal limit for L. monocytogenes in ready-to-eat foods not intended for infants or for special medical purposes (≤100 CFU/g during shelf life ) may not be stringent enough to reduce the risk of infection to acceptable levels for those undergoing therapy to reduce stomach acid secretion.
Patients with CNS infections had fewer underlying conditions than the bacteremia group and presented with various neurological symptoms. The comparative lack of gastrointestinal symptoms in the CNS group suggests that the initial stages of the disease process in the alimentary tract might differ from those for the bacteremia group.
In conclusion, the increase in the incidence of L. monocytogenes bacteremia among people aged ≥60 years in England and Wales between 2001 and 2007 appears to relate to patients with cancer and, to a lesser extent, to patients with other conditions whose treatment leads to stomach acid suppression. Individuals are living longer with chronic conditions requiring immunocompromising treatments; hence, the population at risk of developing infections with opportunistic pathogens such as L. monocytogenes is increasing. Cancer cases represent a large proportion of listeriosis cases in England and Wales; therefore, the provision of targeted food safety advice to this group prior to, during, and after treatment might reduce the impact of subsequent infection. National surveillance of listeriosis in England and Wales must capture information on patients’ treatments, in addition to their underlying pathologies, more effectively.
We gratefully acknowledge the contributions of hospital microbiology, public health, and environmental health colleagues to the effective working of this surveillance system. Thanks are extended to Hilary Kirkbride for advice on symptom classification.
The authors declare no competing interests.
Published ahead of print on 12 August 2009.