Nosocomial infections (NIs) constitute a major public health concern, as they are common and associated with both high morbidity and mortality rates and high healthcare costs 1 2 . Risk factors for NIs vary with the infection site, healthcare setting, and patient age. The National Nosocomial Infections Survey done in the US in 1986-1990 showed that 54% of all NIs occurred in people aged 65 or over 3 . In elderly individuals, the increased susceptibility to severe infections and decreased efficacy of vaccination may reflect ageing of the immune system, or immunosenescence, which involves nearly all the components of the immune system 4 5 6 .

Among the changes characteristic of immunosenescence, those believed to play a major role include the progressive decline in naive circulating T-cell counts (paralleling involution of the thymus), expansion of memory T-cells, and accumulation of terminally differentiated effector CD8 T-cells characterised by loss of CD28 expression 7 . Peripheral CD28^- CD8 T-cell counts are elevated in elderly individuals 8 9 10 . The expansion of this T-cell subset is related to the clonal T-cell expansions often detected in elderly individuals and leads to competition for CD8 T-cell survival niches, thereby hampering the immune responses to infections, most notably those of viral origin 10 .

Cytomegalovirus (CMV) is considered crucial among the repeated antigenic stimuli responsible for the accumulation of oligoclonal effector CD8 T cells 11 12 . Actively keeping CMV in its latent state via the anti-viral immune response places an enormous burden on the entire adaptive immune system 13 and makes a substantial contribution to the phenotype changes observed in T-cell subsets in elderly people 14 .

Geriatric studies have identified several alterations in immunological parameters that are associated with shorter survival in elderly individuals. The term "immune risk phenotype" (IRP) was defined by Pawelec et al. 15 as the combination of high CD8 and low CD4 T-cell counts with a poor proliferative response to concanavalin A and low percentages of B-cells. The IRP predicted a 2- to 4-year decrease in survival among free-living healthy Swedes over 85 years of age (OCTO study, 16 ). Subsequently, the definition of the IRP was simplified to inversion of the CD4/CD8 T-cell ratio 17 . Serological evidence of persistent CMV infection was then found to be strongly associated with CD28^- CD8 T-cell expansion and was consequently added to the definition of the IRP 18 . In the NONA longitudinal study of IRP parameters in a cohort of elderly out-patients not selected for good health, the IRP was associated with a higher mortality rate 19 . To the best of our knowledge, immunosenescence as a possible risk factor for NIs in elderly patients has not been evaluated previously.

The aims of this study were to prospectively assess the prevalence of NIs in elderly in-patients and to look for potential relationships between the occurrence of NIs and the immunophenotype (i.e., main lymphocyte subsets and naive and memory/effector T-cells), with special attention to the IRP as defined in the NONA study 19 .

We included 252 patients aged 70 years or over (181 [71.8%] females; mean age, 85.2 [± 6.2]) years). These patients were transferred from acute medical units (n = 224) or orthopaedic surgical units (n = 28) to the geriatric rehabilitation unit of our internal medicine department. Table 1 reports the main clinical characteristics at admission. The most common diagnoses were cardiopulmonary disease (25.8%), dementia (24.2%), and osteoporosis or fracture (20.2%).

At least one NI was diagnosed in 97 patients, yielding a prevalence rate of 38.5% (95%CI, 32.5-44.5%); 18 patients each had two NIs; thus the total number of NIs was 115. The most common sites of NI were the respiratory tract (21%) and urinary tract (17.1%).

Table 2 compares the groups with and without NIs (NI+ and NI-, respectively). Age and sex did not differ significantly between the groups. The CD4/CD8 ratio was significantly lower in the NI+ group than in the NI- group (p = 0.027). Accordingly, although CD4 T-cell counts did not differ significantly between the groups, the absolute CD8 T-cell count tended to be higher in the NI+ group. Naive CD4 T-cells were significantly decreased in the NI+ group when expressed as a percentage, but the difference was no longer significant when the counts were expressed as absolute values. The NI+ group had significantly lower naive CD8 T-cell counts expressed as both the absolute value and the percentage, compared to the NI- group. Counts of peripheral memory CD8 T-cells and CD28^-CD8 T-cells were significantly higher in the NI+ group, both as absolute values and as percentages. However, in both groups, the median values of all parameters were within the normal age-specific range.

CMV serological status was determined in 246 patients and IRP status in 240 of these 246 patients. CMV serology was positive in 193 (78.5%) patients. In these CMV+ patients, we found immunological changes known to be associated with positive CMV serology (increased peripheral memory and effector CD4 and CD8 T-cells, data not shown). However, CMV status was not significantly associated with NI occurrence (p = 0.87). The rate of IRP+ patients was slightly higher in the NI+ group than in the NI- group, but the difference was not statistically significant (p = 0.11). The rate of nosocomial pneumonia was significantly higher in the IRP+ group (n = 60) than in the IRP- group (17/60, 28.3% versus 28/180, 15.6%; p = 0.036). Table 3 compares the IRP criteria in the 97 NI+ patients according to the initial site of infection. No significant differences were found across sites for a CD4/CD8 ratio < 1, CD8T-cells > 600/μL, CD28-negative T-cells > 300 cells/μL, or positive CMV serology. In-hospital mortality was 6% (n = 15/252) and did not differ between the IRP+ and IRP- groups (3/60, 5% versus 12/180, 6.6%, respectively; p = 0.76). However, in-hospital mortality was significantly higher in patients with nosocomial pneumonia than in the other patients (8/45,17.8% versus 7/207, 3.4%; p = 0.001).

In our elderly population, the immunological parameters associated with ageing of the immune system were more severely altered at admission in the patients who subsequently experienced one or more NIs than in those who remained free of NIs. More specifically, at baseline, patients with subsequent NIs had lower CD4/CD8 ratios, higher CD8 T cell counts, lower naive CD4 and CD8 T-cell counts, higher proportions of memory C8 T-cells, and higher counts of CD28^-CD8 T-cells, compared to patients without NIs. CMV status was not associated with NIs. Moreover, patients exhibiting the IRP had a higher rate of nosocomial pneumonia than did the other patients.

To evaluate the immune system in our elderly patients, we looked at two sets of abnormalities associated with ageing: (i) classical immunology variables, which showed decreases in naive T-cell subsets and increases in memory/effector cells; (ii) and the IRP, a combination of immunological alterations specifically identified as associated with shorter survival in free-living healthy Swedes over 85 years of age 15 16 17 . In subsequent studies, such as the NONA study 19 , the IRP was used in cohorts of individuals not selected for good health and was found to be unrelated to morbidity 20 . Both sets of abnormalities were present in our patients with NIs.

A limitation of our study is that only peripheral lymphocytes were counted. We did not evaluate other immune cells such as antigen-presenting cells or phagocytes, and we did not assess local factors at sites of infection.

In this study, we endeavoured to determine whether the risk of NI was predicted by immunological parameters. To our knowledge, the ability of immunological abnormalities reflecting immunosenescence to predict infections had not been studied previously in elderly in-patients. Wikby et al. first established that the IRP was a valuable predictor of mortality in very elderly people, a finding later confirmed in the NONA study 17 19 . However, the NONA study did not specify the causes of death and assessed the ability of the IRP to predict relatively long-term survival (6 years in patients aged 90 years or over). Interestingly, immunosenescence as assessed based on alterations in blood T-cell subsets has been described in Alzheimer's disease 21 , and CMV infection is believed to contribute to the increased susceptibility of elderly individuals to many types of chronic antigenic stress such as cancer and autoimmunity 22 .

Our data indicate a greater degree of immunosenescence in patients with NIs, compared to patients without NIs, despite the similar mean ages in the two groups. Moreover, patients with a positive CMV status and at least one immunological criterion for IRP were at increased risk for bacterial or viral pneumonia but not for bacterial infections of the urinary tract or other sites. This finding is consistent with the hypothesis that senescent CD8 T cells cannot respond properly to viral infections 10

The immunological alterations that characterise immunosenescence are more pronounced in elderly in-patients who subsequently experience NIs. The IRP profile is associated with nosocomial pneumonia in the relatively short term. The IRP parameters, which can be assessed easily on a routine basis and are known to predict higher mortality in elderly, have been shown by our study to also indicate a higher risk of NI, which is a clear manifestation of dysimmunity.

CMV: cytomegalovirus; Ig G: Immunoglobulin G; IQR: interquartile range; IRP: immune risk phenotype; NI: nosocomial infection; NK cell: natural killer cell.

The authors declare that they have no competing interests.

EP conceived and designed the study and is the guarantor for the study.

SBG contributed to the design, performed the statistical analysis, and interpreted the data.

AP performed the flow cytometry analysis and wrote the initial draft of the article, to which all the authors contributed subsequently.

FT, CB and KL collected the clinical data.

JPF participated in data interpretation and provided helpful comments about the manuscript.

All authors had full access to the study data and take responsibility for the integrity of the data and accuracy of the data analysis.

All authors read and approved the final manuscript.