Splenectomy sequelae: an analysis of infectious outcomes among adults in Victoria

All patients who underwent a splenectomy in Victoria from 1 July 1998 to 31 December 2006 were identified using linked hospital discharges from the Victorian Admitted Episodes Dataset and death registration data.

The Victorian Admitted Episodes Dataset is maintained by the Victorian Department of Health and is based on hospital data compiled by all public and private hospitals in the state. The dataset contains demographic and clinical information on each episode of patient care, with information coded according to the International statistical classification of diseases and related health problems, 10th revision, Australian modification (ICD-10-AM).5 Death registration data are compiled by the Victorian Registry of Births, Deaths and Marriages.

There are no unique identifiers in these datasets and as such the linked longitudinal dataset of hospital episodes and death registrations has been developed through a process of sequential deterministic linkage. Data released for research are stripped of any potentially identifying data element and in their place is an encrypted unique personal identification number, which serves to bring together data from the same patient over different years of admission. The analysis data are thus personally non-identifiable.

The sampling frame included all patients who underwent splenectomy (defined by the ICD-10-AM procedure codes for splenectomy) between 1998 and 2006. The authors classified the indications for splenectomy into six mutually exclusive groups, with cases having multiple indications ordered hierarchically, regardless of other indications: (i) trauma; (ii) therapeutic malignancy (splenectomy performed specifically to manage malignant disease; usually planned); (iii) therapeutic haematological (splenectomy performed specifically to manage haematological disease; usually planned); (iv) therapeutic other (includes treatment of underlying diseases, such as local infections of the spleen or congenital abnormalities of the spleen); (v) iatrogenic malignancy (splenectomy performed as an unintended accompaniment to surgery to manage malignant disease; generally unplanned); and (vi) iatrogenic non-malignant (splenectomy performed as an unintended accompaniment to surgery to manage non-malignant disease; generally unplanned).

Severe infections were defined using ICD-10-AM codes for infections from the hospitalisation data. The authors categorised the codes into groups according to organ system. Those infections occurring in the first 30 days after splenectomy were excluded in order to avoid inclusion of complications potentially resulting from the primary operation or related admission.

Survival was calculated from the date of splenectomy to either the date of death or to 31 Dec 2006 for those alive at the end of follow-up.

The time to first severe infection was calculated from the date of splenectomy to the date of the same patient’s first admission for an infectious diagnosis. For those without any infections, it was calculated until the end of follow-up.

Incidence rates of first severe infection per 100 person-years and 95% confidence intervals were calculated for sex, age group and indication for splenectomy. Multivariate Cox proportional hazards regression models were fitted to compute hazard ratios (HRs), adjusted for age, sex and indication for splenectomy. Hazard proportionality was assessed using analysis of scaled Schoenfeld residuals. Paediatric cases were defined as aged less than 15 years at the time of splenectomy.

Survival curves/time to event curves were calculated for time to death, first severe infection and first episode of sepsis by sex, age group and indication for splenectomy.

All reported P values are two-tailed and, for each analysis, P < 0.05 was considered significant. We used Stata, version 11.0 (StataCorp, College Station, Tex, USA) for all analyses.

The Victorian Department of Health granted ethics approval for this study.

A total of 644 (26.0%) splenectomised patients had a severe infection requiring hospitalisation with an overall incidence of first severe infection (FSI) of 8.0 per 100 person-years (95% CI, 7.2–8.4).

The rate of FSI differed significantly according to age, sex and indication for splenectomy, with female sex, age < 50 years and trauma as the indication for the splenectomy having significantly lower unadjusted rates of first infection. These relationships did not change after adjustment (Box 2).

The mean incidence rate of FSI in the 12 months after splenectomy from 1998 to 2003 was 7.0 per 100 person-years. Following the introduction of the Victorian Spleen Registry in November 2003, the median incidence rate of FSI for the first 12 months after infection was 6.2 per 100 person-years (P = 0.047).

Eighty-three per cent of infections fell into nine categorised body sites; the remainder were uncategorisable (Box 3). Respiratory infections and sepsis (proven or presumed bloodstream infections) were the most common sites of infection. Gram-negative infections represented the most frequent causative organism group, accounting for 698 (51%) of bacterial pathogens. Staphylococcus aureus accounted for 414 (31%) of infections. Streptococcus pneumoniae, meningococcal species and Haemophilus influenzae accounted for 7 (5%), 4 (3%) and 10 (7%) infections, respectively. Compared with splenectomised trauma patients, the adjusted HR of first severe sepsis was 16.5 (95% CI, 6.1–44.6; P ≤ 0.001) for patients splenectomised for therapeutic malignancy; 6.2 (95% CI, 2.4–16.3; P < 0.001) for patients splenectomised for therapeutic haematological indications; 4.9 (95% CI, 1.7–14.3; P = 0.003) for patients splenectomised for iatrogenic malignancy; and 2.9 (95% CI, 0.7–12.2; P = 0.149) for those splenectomised for therapeutic other indications (Box 4).

The rate of FSI after splenectomy declined over time. The highest risk of infection was seen within the first 2 years after splenectomy, with 25% of all FSIs occurring during this time, 50% of infections by 3.7 years, 75% by 5.7 years, and 95% by 7.6 years. The cumulative incidence at 3, 6, 12 and 24 months was 73.4, 53.9, 38.3 and 22.1 per 100 person-years, respectively. However, first infections did continue to occur late, with the cumulative incidence at 8 years being 7.8 per 100 person-years (Box 5).

Overall, 562 splenectomised patients (22.7%) died during the follow-up period, with 395 (15.3%) dying within the first 2 years (Box 6). The median survival for patients who died during the follow-up period after splenectomy was 3.1 years (interquartile range [IQR], 1.3–5.3). There was an increased adjusted hazard of death for individuals aged ≥ 50 years compared with those aged < 50 years (HR, 4.8; 95% CI, 3.5–6.3). Women had a decreased hazard of death compared with men (HR, 0.8; 95% CI, 0.6–0.9) (Box 7).

Compared with all other reasons for splenectomy, patients who underwent splenectomy for trauma had the lowest mortality. Among patients who died, the median survival for those splenectomised for trauma was 3.7 years (IQR, 1.7–5.8 years), compared with 2.1 years (IQR, 0.8–4.3 years) for iatrogenic malignancy, 3.2 years (IQR, 1.1–5.7 years) for iatrogenic non-malignancy, 3.6 years (IQR, 1.6–5.5 years) for therapeutic haematological indications, 2.6 years (IQR, 0.9–4.6 years) for therapeutic malignancy, and 3.4 (IQR, 1.9–5.5 years) for therapeutic other indications. The adjusted HR of patients undergoing splenectomy for therapeutic malignancy was 2.4 (95% CI, 1.9–3.4; P < 0.001) and, for iatrogenic malignancy, 2.5 (95% CI, 1.7–3.4; P < 0.001) .

Vaccination, education and antibiotic prophylaxis are recognised as reducing the risk of postsplenectomy sepsis and death.6 Although the Australasian Society for Infectious Diseases has published guidelines,7 the appropriate duration of antibiotic prophylaxis in otherwise immunocompetent hosts remains ambiguous. Consequently, the questions remain of who should receive antibiotic prophylaxis and for how long? Committing patients to lifelong antibiotics is difficult; therefore, more information on how to tailor therapy can assist decision making. Our study has identified patient groups at highest risk for adverse outcomes and can be used to guide management decisions.

Based on one of the largest population-based cohorts of splenectomised patients in the literature, our results show that 26% of splenectomised patients had at least one severe infection requiring hospitalisation during the period of observation, with these occurring at a mean of 3.5 years after splenectomy. Our rate of postsplenectomy infection of 8.0 per 100 person years is consistent with international studies (using the same case-definition for severe infection), which have reported rates of 7.0,8 7.2,9 and 7.710 per 100 person-years. A Western Australian study reported a substantially lower rate of severe infection of 0.4 per 100 person-years, but included only septicaemia, meningitis and pneumococcal pneumonia.11 In contrast, a Danish civil registry system reported a risk of severe infection in the community of 2.0 per 100 person-years.10

A quarter of severe infections occurred in the first 2 years after splenectomy, with the highest risk in the first 6 months, consistent with the findings in other studies.8,12-15 A Danish population-based cohort study reported that 10.2% of infections occurred within 90 days of splenectomy, with a relative risk of 18.1 compared with the general population.10 In our cohort, 31.8% of FSI occurred more than 5 years after splenectomy. Late postsplenectomy sepsis is a less well recognised problem,11 but fulminant infection has been documented more than 20 years after splenectomy.11,16,17 Our data support the statement that asplenic patients should be considered to have a risk of developing severe late postsplenectomy infection up to at least 5 years after splenectomy.

The rates of infection were lowest among patients splenectomised for trauma and highest among those splenectomised for malignant disease, in keeping with other literature.8,9,12,13 Notably, trauma patients may have a partial protective effect from splenosis, owing to self-implanted remnants of the traumatised spleen.18 Partly because of the problems associated with using an administrative dataset, it is difficult to determine whether infection or death was due to the splenectomised state or to the intrinsic risk conferred by the underlying disease process as well as immunosuppressive therapies. This represents an important limitation of our study.

Late in the period of observation (2003) the Victorian Spleen Registry was established. One of its major roles is in providing education to splenectomised patients.19 The median incidence of FSI in the first year after splenectomy was higher than in the 4 years before the Registry began, compared with the 2 years after introduction. However, the short follow-up periods make these data difficult to interpret and unlikely to have a major influence over our findings.

Our study found that encapsulated organisms were less common than gram-negative organisms and S. aureus. It is difficult to determine if this represents uptake of vaccination and antibiotic prophylaxis in Victoria or a changing pattern of postsplenectomy infections. Targeted studies are warranted to determine the microbiology of infecting organisms and whether penicillin remains the most appropriate agent for antibiotic prophylaxis. Prospective studies are needed to analyse infectious outcome data relating to the use of prophylaxis, both with antibiotics and vaccinations. The use of such data to develop a risk assessment tool, similar to the CHADS2 score for anticoagulation in atrial fibrillation,20 and other individualised guidelines for postsplenectomy care would greatly assist clinicians managing asplenic patients.