Authors: Emma Rademaker, Lisette M. Vernooij, Tom van der Poll, Marc J. M. Bonten, Helen Leavis, Olaf L. Cremer, Lennie P. G. Derde
Categories: Research, Critical illness, Hypogammaglobulinemia, Intravenous immunoglobulins, Personalized medicine, Pneumonia
Low endogenous immunoglobulin(Ig)-levels are common in critically ill patients with sepsis, but it is unknown whether low Ig-levels are associated with poor outcome, and in which patients Ig-replacement therapy (IgRT) improves outcome. Given the crucial role of immunoglobulins in eliminating certain encapsulated pathogens, we examined the relationship between serial Ig-levels and disease course in critically ill patients with community acquired pneumonia (sCAP) caused by encapsulated or other pathogens.
We included a cohort of consecutive critically ill patients with CAP, and PaO2/FiO2-ratio < 200 with or without septic shock, from an existing biorepository where microbiological causes of infection had been adjudicated in a protocolized manner. We used generalized linear mixed models to assess the association between IgG and IgM (measured on admission days 1, 3 and 7) and disease course (Sequential Organ Failure Assessment (SOFA)-score on day 2, 4, and 8) for all-cause sCAP and for episodes caused by Streptococcus (S.) pneumoniae or Haemophilus (H.) influenzae.
We included 255 eligible patients admitted with CAP, of which 82 (32%) episodes were caused by S. pneumoniae or H. influenzae. 151 (59%) patients had low IgG (< 7.0 g/L), 77 (30%) had low IgM (< 0.4 g/L), and 56 (22%) had both. A lower IgG-level was related to a slightly higher SOFA-score at admission (β = − 0.07 per 1 g/L IgG, p = 0.029), but an IgG-level decline over time was not associated with a SOFA-score increase (β = − 0.04, p = 0.564). IgM-levels were not associated with changes in SOFA-score over time. Neither association was affected by the presence or absence of S. pneumoniae and H. influenzae.
In critically ill patients with CAP, IgG and IgM dynamics in the first week of ICU stay are not associated with clinically relevant changes in disease course, regardless of the causative pathogen.
The online version contains supplementary material available at 10.1186/s13054-024-05197-3.
Keywords: Pneumonia, Critical illness, Hypogammaglobulinemia, Intravenous immunoglobulins, Personalized medicine
Low endogenous immunoglobulin (Ig-) levels (hypogammaglobulinemia) are common in patients admitted to the intensive care unit (ICU) with sepsis [1–5]. Hypogammaglobulinemia in sepsis is caused by increased consumption, decreased production, altered distribution due to capillary leak, or hemodilution due to iatrogenic fluid resuscitation [6]. As immunoglobulins are crucial in opsonizing pathogens to facilitate phagocytosis, and in neutralizing exotoxins [7], hypogammaglobulinemia may contribute to higher morbidity and mortality rates. Augmenting immune resistance (i.e., immunostimulation) using Ig-replacement therapy (IgRT)—regardless of immunoglobulin levels—has been proposed in patients with sepsis. However, while recent meta-analyses of IgRT trials have reported significant benefits [8, 9], such effects were mainly observed in studies with a high risk of bias, while low risk of bias studies showed no survival benefits. Furthermore, while some subgroup analyses have shown promising results for IgRT in sepsis patients with hypogammaglobulinemia [5], findings from other subgroup analyses were inconsistent [10–15]. Therefore, IgRT is generally not recommended for sepsis [16].
Currently, IgRT is considered standard of care as a preventive measure in patients with inherent hypogammaglobulinemia (e.g. common variable immune deficiency), who are at high risk of infections, like community acquired pneumonia (CAP). From these patients, we know that opsonization is particularly important in the defense against certain encapsulated pathogens in the sinopulmonary tract. These pathogens, including S. pneumoniae and H. influenza, have polysaccharide capsules that prohibit phagocytosis, thus causing recurrent sinopulmonary infections in these patients [17]. In patients with inherent hypogammaglobulinemia, immunostimulation with IgRT reduces the occurrence of infections and organ damage [18]. Similarly, it may be useful to give IgRT to patients with hypogammaglobulinemia and sCAP, especially if caused by encapsulated bacteria.
To date, few studies investigating Ig plasma concentrations and disease course have evaluated potential high-risk subgroups, instead most studies focused on sepsis as a whole. In sepsis patients, depletions in IgG and/ or IgM levels have been associated with disease severity and mortality [2–4, 19–21], though this relation is not consistently observed, likely due to sepsis heterogeneity [1, 6, 22]. Some studies focusing solely on CAP patients have also reported an association between low Ig levels and poor outcomes [23, 24]. However, in these studies, the bidirectional relationship between Ig-levels and disease course—where critical illness can lower Ig-levels and low Ig-levels can worsen disease course—is not well studied. As a result, it remains unclear if these findings should be considered prognostic or causal. And lastly, past studies have not clarified whether the observed hypogammaglobulinemia in sepsis patients represents a persistent production deficit. For sepsis it remains, thus, unclear if hypogammaglobulinemia is associated with harm and if this is extra detrimental if the underlying infection is caused by encapsulated bacteria.
We assessed the relationship between Ig-levels and disease course in a curated cohort of critically ill patients with CAP caused by various pathogens. We hypothesized that persistently low Ig-levels are associated with worsening disease course; and that this effect is more pronounced if pneumonia is caused by S. pneumoniae or H. influenzae.
Patients with an expected ICU stay of > 24 h were prospectively enrolled in the Molecular Diagnosis and Risk Stratification of Sepsis (MARS) cohort in two tertiary ICUs in the Netherlands from 2011 onwards. The local ethical committee approved the study with opt-out consent (Institutional Review Board number 10-056C; TCBio 21/131). Trained observers prospectively recorded clinical data on all infectious events and performed a detailed and prespecified post-hoc adjudication of the likelihood of infection during ICU admission, based on CDC criteria [25]. For the current analysis, we selected consecutive patients with CAP as their main admission diagnosis, with an adjudicated infection likelihood of ‘probable’ or ‘definite’ [25]. In addition patients had to meet the following three disease severity an ICU length of stay of at least 72 h, invasive mechanical ventilation with a PaO2/FiO2-ratio < 200, and/or had Sepsis-3 defined septic shock [26]. These criteria were established to ensure the study population reflected patients who, despite 48 h of antibiotics on the ICU have not shown sufficient recovery (and IgRT could thus be considered). Patients were excluded if they had opportunistic infections, had Coronavirus disease 2019, had received antibiotics > 48 h before ICU admission, or received IgRT within the first ICU week. Patients with a known prior immune deficiency were not excluded, as they represent a group of interest.
Data on demographics, comorbidities, disease severity, and laboratory results up to one year after ICU admission, had been prospectively collected within the MARS cohort. All microbiological evidence was thoroughly reviewed to determine the causative pathogen(s) responsible for CAP. In cases where the causative pathogen was uncertain, the case was discussed by a panel of three informed reviewers. If S. pneumoniae or H. influenzae (SpHi +) was identified as (one of the) causative pathogen(s), the patient was considered at higher risk for a poorer disease course in relation to low Ig-levels, as elimination of these pathogens may rely more on Ig. While other causative pathogens (e.g., Staphylococcus aureus, Klebsiella species and Escherichia coli) may also have polysaccharide capsules and similarly depend on immunoglobulins for elimination*,* they are not typically associated with recurrent infections in patients with persistent hypogammaglobulinemia and were thus not classified as high-risk for the primary analyses.
We measured IgG, IgM and albumin levels in plasma samples collected upon admission and on ICU days 3 and 7. We used daily leftover heparin plasma samples, processed within 4 h of collection and stored at − 80 °C. IgG and IgM concentrations were measured using a PEG-enhanced immunoturbidimetric method (Atellica CH, Tarrytown, NY). Albumin was measured by bromocresol purple (BCP) dye-binding (Atellica CH, Tarrytown, NY).
Hypogammaglobulinemia was defined as either a plasma IgG-concentration < 7.0 g/L, and/or a plasma IgM-concentration < 0.4 g/L. Hypoalbuminemia was defined as a plasma albumin concentration of < 35 g/L. We calculated Ig-level to albumin ratios (Ig/albumin-ratio), to estimate the extent to which an Ig-level decrease could be explained by increased consumption or a lack of production of Ig, versus hemodilution and vascular leakage (using albumin levels as a measure for hemodilution and vascular leakage). In the primary analyses, Ig-levels were treated as continuous data. The primary outcome was the disease course during the first week of ICU admission, as measured by the Sequential Organ Failure Assessment (SOFA) score. We excluded the central nervous system component of SOFA as it was frequently not observable in our predominantly sedated study population [27]. We analyzed trends in PaO2/FiO2-ratio as a secondary outcome.
Generalized linear mixed models were used to analyze separately the association of IgG and IgM plasma concentrations measured on days 1, 3, and 7 with the change in SOFA scores on day 2, 4, and 8. Including Ig concentration as a time-varying covariate for SOFA score, allowed us to distinguish the effect of average Ig concentration differences between patients (i.e., between-patient effect) from the effect of changes in Ig concentration over time within individual patients (i.e., within-patient effect). In practice, the patient-mean Ig concentration reflects the between-patient effect, while deviations from this capture the within-patient effect [28] (Supplementary methods). To assess if the relation between IgG- or IgM-levels and disease course was affected by SpHi + , we included an interaction term for SpHi ± and the person-mean Ig-level. To obtain our final mixed-effects models, we examined if a quadratic term for time, and random intercepts and slopes improved model fit. We used the likelihood ratio test and Akaike’s information criterion to select the best fitting mixed-effects model. Furthermore, we utilized restricted maximum likelihood estimation to obtain unbiased variance estimates for the final models [28]. To adjust for potential confounding, age and SOFA score on ICU admission were included as covariates in all models.
We performed subgroup analyses for patients with and without septic shock, as well as for immunocompromised and non-immunocompromised patients, as the association between Ig-levels and our primary outcome may vary across these groups. We performed sensitivity analyses to assess the robustness of our findings. First, all patients infected with pathogens for which some strains are encapsulated (and thus may rely more on Ig for elimination) were reclassified into the SpHi + group (Table S1). Second, we analyzed the reciprocal association between SOFA scores on days 0, 2 and 6 and Ig-concentrations (days 1, day 3 and day 7) to quantify potential bias due to a bidirectional relation between exposure and outcomes. Specifically, instead of Ig-levels influencing SOFA scores, SOFA scores may impact Ig-levels (Supplementary methods).
We used R version 4.0.2 for all statistical analyses. Between-group differences were assessed using a Fisher’s exact tests or Mann–Whitney U tests, as appropriate. Correlations were assessed using Pearson’s correlation test or Spearman’s rank test, as appropriate. Data are presented as medians (interquartile range) or absolute numbers (%). A p-value < 0.05 was considered statistically significant.
Out of 1374 consecutive critically ill patients admitted with CAP between January 2011 and June 2024, 255 were eligible for the study (Figure S1). The median age of included patients was 62 years [Interquartile range (IQR): 48–70] (Table 1). 111 patients (44%) had septic shock at ICU admission, and 47 patients (18%) died in the ICU (Figure S2). The most frequently detected bacterial pathogens of CAP were S. pneumoniae (n = 58, 23%), S. aureus (n = 36, 14%) and H. influenzae (n = 29, 11%). Overall, 82 patients (32%) had CAP caused by either S. pneumoniae or H. influenzae, whereas 25 patients (9.5%) had viral CAP without bacterial coinfection and in 43 patients (17%) no causative pathogen was identified (Table S1). Measurements of disease severity at ICU admission were significantly higher in patients with an infection caused by S. pneumoniae or H. influenzae, including a higher median SOFA score (8 [IQR: 7–9] and 7 [IQR: 5–9], for SpHi + and SpHi-, respectively) (Table S1).
Hypogammaglobulinemia (either low IgG or low IgM concentrations) was present in 132 (55%) patients at day 1 of ICU admission, 166 patients (63%) at day 3 and 82 patients (46%) at day 7 (Fig. 1), primarily driven by low IgG levels. In 45 patients (19%), 52 patients (20%), and 16 patients (9%), both IgG and IgM levels were below the reference ranges on day 1, 3 and 7, respectively. In 162 patients (64%) Ig-levels declined from day 1 to day 3 in ICU. However, at day 7, 79% of those still in ICU had higher Ig-levels than measured at day 3 or 1 (Table 1). Ig-levels at day 1, 3 and 7 did not differ statistically significant between critically ill patients with CAP caused by S. pneumoniae or H. influenzae and those with CAP caused by other pathogens (Table S1).
Fig. 1 Immunoglobulin levels in critically ill patients with CAP. The red sections represent patients with immunoglobulin levels below the normal reference ranges (IgG < 7.0 g/L or IgM < 0.4 g/L). IgG: Immunoglobulin G; IgM: Immunoglobulin M. Encapsulated bacteria are limited to Streptococcus pneumoniae and Haemophilus influenzae
The median SOFA-score at ICU admission was 7 [6–9]. On average, SOFA scores decreased by 0.6 (95%CI −0.8 to −0.4) points daily, suggesting overall clinical recovery. We observed a crude association between IgG-levels and SOFA at baseline with a β of −0.17 (95%CI −0.27 to −0.06) per 1 g/L IgG increase (p < 0.01). This suggests that patients with lower IgG-levels were in poorer condition at ICU admission. However, further decreases in IgG-levels within individual patients were not associated with disease progression (i.e. increasing SOFA) (β = −0.08 (95%CI −0.28 to 0.11) per g/L, p = 0.39).
The analysis adjusted for age and SOFA score at ICU admission yielded similar results. Lower IgG-levels were associated with higher admission SOFA scores, but the effect size was small (β = −0.07 (95%CI −0.13 to −0.01) per g/L, p = 0.03) (Fig. 2, Table S3A) and this was only apparent in patients with septic shock (Table S4A). IgM levels were not significantly associated with SOFA scores at admission or over time (Figure S3, Table S3B). There was no significant interaction between Ig-levels and SpHi + (Fig. 2, Table S6B), meaning low or declining Ig-levels were not more detrimental if CAP was caused by S. pneumoniae or H. influenzae, compared to other causative pathogens. Ig-levels were not associated with PaO2/FiO2 trajectories (our secondary outcome) (Table S3A), regardless of causative pathogen type (Table S6B). The relation between Ig-levels and SOFA scores did not vary between immunodeficient and non-immunodeficient patients. Sensitivity analyses were consistent with the main analyses (Table S5A, S5B).
Fig. 2 Estimated disease progression across different clinical scenarios. Mean changes in SOFA-score estimated with a generalized linear mixed-effects model for repeated measures, reported for A high vs. low baseline IgG which remains constant over time, B increasing vs decreasing IgG over time, from an equal baseline level, and C patients with encapsulated vs. non-encapsulated infections in the context of a high or low constant IgG-level. The I bars represent 95% confidence intervals. Encapsulated + : Patients with infections caused by Streptococcus pneumoniae or Haemophilus influenzae; Encapsulated-: Patients with infections caused by other pathogens
94% of critically ill patients with CAP had hypoalbuminemia at ICU admission and, like Ig-levels, albumin levels reached a nadir at day 3 (Table 1). Unlike Ig-levels, albumin levels did not recover by day 7, resulting in increasing Ig/albumin-ratios (Fig. 3). This suggests that the observed recovery of Ig-levels results from restored production rather than from normalization of hemodilution and protein loss. This is further supported by the increasing strength of correlations between albumin levels and cumulative fluid balance from day 1 to day 7 (r = −0.42, p < 0.0001; r = −0.43, p < 0.0001; r = −0.49, p < 0.0001 for days 1, 3, and 7, respectively), while the correlation between IgG levels and cumulative fluid balance weakened by day 7 (r = −0.25, p < 0.01; r = −0.27, p < 0.0001; r = −0.18, p = 0.02). IgM levels did not correlate with cumulative fluid balance (r = 0.07, p = 0.39; r = −0.11, p = 0.147; r = −0.09, p = 0.23, day 1, day 3 and day 7 respectively).
Fig. 3 Ig/albumin ratios. Plasma Ig/albumin ratios on ICU admission day, day 3, and day 7; Between-group differences in Ig/albumin ratios were assessed using the Mann–Whitney U test for continuous data. *p < 0.05; **p < 0.01; ***p < 0.001; Encapsulated + : Patients with infections caused by Streptococcus pneumoniae or Haemophilus influenzae; Encapsulated -: Patients with infections caused by other pathogens; ICU: Intensive Care Unit
In this study, we assessed the relationship between Ig-levels and disease course in critically ill patients with CAP. We found a statistically significant but clinically negligible association between low IgG levels and the SOFA score at ICU admission, while declining IgG-levels over time were not linked to worsening SOFA scores. IgM levels showed no association with disease course, and neither relationship was influenced by the presence of encapsulated bacteria (i.e., S. pneumoniae and H. influenzae).
Moreover, in our study, increasing Ig/albumin ratios over time suggested effective Ig production. While decreased albumin production could offer an alternative explanation for increasing Ig/albumin ratios, albumin declines in sepsis are primarily attributed to enhanced clearance from the circulation rather than reduced production [29]. The strong correlation between albumin levels and cumulative fluid balance further supports the former explanation of effective Ig production. This seems to contrast findings from previous studies in sepsis, which reported lower levels of B-cells, and an exhausted phenotype—characterized by decreased major histocompatibility complex class II expression and increased production of the anti-inflammatory cytokine IL-10 [30–32]. In another study, in vitro IgM production by lymphocytes of sepsis patients was impaired [4]. However, a meta-analysis reported that Ig concentrations tend to increase over time in most studies [1], consistent with our observations.
While some previous studies align with our findings [6, 22], others have reported significant associations between low Ig-levels and clinical outcomes in both sepsis [2–4, 19–21] and severe CAP patients [23, 24]. Some factors may explain this discrepancy. First, in our study, we had access to an extensive ICU biorepository with daily plasma samples of over 17,000 ICU patients, which included detailed microbiological data. This allowed us to include a homogeneous high-risk subgroup of critically ill patients with CAP who might have been considered candidates for immunotherapy (i.e., those with insufficient recovery after 48 h of antibiotics) with accurate classification based on pathogen type. In this large, homogeneous subgroup of patients with sepsis we could not demonstrate a clinically meaningful relation between low Ig-levels and disease course.
Second, we designed our statistical methods to account more effectively for the bidirectional relationship between Ig-levels and disease course. That is, as patients clinically deteriorate, Ig-levels may also deteriorate through disease severity related mechanisms such as protein loss and hemodilution. Thus, the observed association between low Ig-levels and mortality may not reflect that Ig-levels influence disease progression, but rather the opposite—where Ig-levels serve as a marker of disease severity. To avoid this issue, we modelled the effect of temporal Ig-levels on disease course directly, rather than on other outcomes for which disease course is in the causal pathway (e.g., mortality).
Results from the CIGMA II trial indicate that certain subgroups—particularly severe CAP patients with high CRP and low lymphocyte counts, and especially when accompanied by low IgM—did benefit from treatment with Trimodulin (a human polyvalent immunoglobulin containing 23% IgM, 21% IgA, and 56% IgG) [5, 33]. Trimodulin had important immunomodulatory effects (i.e. dampening of inflammatory cytokines and quicker normalization of neutrophile levels), that may explain these observed benefits [33–35]. So while our findings do not support the rationale for using IgRT in sepsis, such additional properties may still be a reason to investigate IgRT as a therapeutic approach for sepsis. Moreover, as this was an observational study, we could not definitively assess IgRT efficacy.
Our study has several limitations. First, we did not measure pathogen specific Ig titers, but used IgG and IgM as proxies. Additionally, free light chains, which have been associated with mortality in previous studies, were not included in our analysis [36]. Second, the sample size limited our ability to employ models that more effectively address bi-directional and auto-regressive effects, such as multi-state models or cross-lagged panel models. Third, we focused on critically ill patients with CAP, with inadequate recovery after two days of antibiotic therapy, as we considered this subset theoretical candidates for IgRT therapy. Therefore, our findings are generalizable only to this specific patient group, not the broader severe CAP population at ICU admission. Lastly, our study began in 2011, and standards of care may have shifted since then—for example, hydrocortisone use for septic shock increased in our ICU, potentially impacting our results.
In a curated cohort of critically ill patients with CAP admitted to the ICU, who had been treated with antibiotics for 48 h, low Ig-levels had no clinically meaningful association with disease course, regardless of whether the causative pathogen was encapsulated (pathogens that rely on immunoglobulins for elimination) or not.
Not applicable.
All authors designed the study. ER, OC, TvP and LD participated in data collection. ER, OC and TvP accessed and verified all the data. ER and LV performed the data analysis. ER and LD drafted the manuscript. All authors participated in revision of the manuscript. All authors gave final approval for the submission of this version. All authors agree to be accountable for all aspects of the work.
No external funding was received for this study.
The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at the UMC Utrecht.
The local ethical committee approved the Molecular Diagnosis and Risk Stratification of Sepsis (MARS) cohort and biobank with opt-out consent (Institutional Review Board number 10-056C). Additionally, the use of selected plasma samples from the MARS biobank for this study was approved by the designated committee (TCBio 21/131).
Not applicable.
The authors declare no competing interests.
The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at the UMC Utrecht.