Before the discovery of insulin and antibiotics, infections contributed substantially to diabetes-associated morbidity and death. It has been estimated that infections killed 1 in 5 diabetic patients in the 1920s compared to <1 in 20 in the late 1960s 1. Despite recent advances in the management of both diabetes and infectious diseases, diabetic patients remain at increased risk of infection 2. Although intensive blood glucose control significantly reduces vascular complications in both type 1 and type 2 diabetes 3, 4, the relationship between glycaemia and infections is less well established. Other factors, including attenuated immune responsiveness and diabetic tissue damage, may play important roles. The present review assesses the current understanding of the relationship between diabetes, immune function and the management of common community- and hospital-acquired infections.
Search strategy and selection criteria
We searched PubMed, Excerpta Medica Database (EMBASE) and Medline using ‘diabetes’ in combination with keywords including ‘infection’, ‘sepsis’ and ‘immunity’. The reference lists in the articles generated were used to identify other publications of interest. The primary criterion for inclusion in the present review was clinical relevance.
Components of host defense that may be impaired in diabetes have been studied in detail 5, 6, especially the innate immune system and the polymorphonuclear neutrophil (PMN). Defects in adaptive immunity are less well characterized, but are likely to be important cofactors given the diversity of organisms, including fungi and Mycobacterium spp., to which diabetic patients show increased susceptibility.
The steps involved in pathogen elimination by PMN are (1) PMN adhesion to vascular endothelium, initially via the cell surface adhesion molecule L-selectin and then integrins, (2) transmigration through the vessel wall down a chemotactic gradient, and (3) phagocytosis and microbial killing. Although an increase in PMN adhesion to vascular endothelium has been documented in patients with diabetes 7, the significance of this change in mediating a predisposition to infection is unclear. Such increased adhesion does not relate HbA1c levels and does not persist after PMN stimulation 7, but could predispose to vascular complications given the potential for endothelial injury.
Transmigration of PMN, as measured through endothelial cell monolayers, is impaired in patients with diabetes in parallel with an increase in the concentration of advanced glycosylation end products 8. Several studies have also shown that PMN chemotaxis may be reduced in diabetic patients irrespective of glycaemia 9–12 but the data are difficult to interpret because of variations in experimental methodology. In one well-standardized study 7, chemotaxis was reduced in PMN from diabetic versus control subjects, especially in the presence of glucose concentrations >12 mmol/L and when PMN were isolated from patients with vascular complications 7. Similar results have been reported for monocytes from diabetic patients 13. In a more recent study, hyperinsulinaemia in the presence of euglycaemia significantly improved PMN chemotaxis, raising important questions about the direct role of insulin in the immunity of diabetes 14.
Studies of phagocytic function are also confounded by variations in experimental technique. Initial evidence for reduced phagocytosis of opsonized Staphylococcus aureus6, 15 and pneumococcus 5 by PMNs from diabetic subjects has not been confirmed by more recent investigations using the opsonized latex bead ingestion test 7 or other bacteria and fungi 16, 17. As with chemotaxis, PMN phagocytosis has recently been shown to improve in the presence of hyperinsulinaemic euglycaemia 14, adding support to the suggestion that insulin is an important influence on PMN function.
The most convincing evidence of PMN dysfunction in diabetic patients relates to microbial killing 7, 18–21. Superoxide production is reduced in parallel with increasing glycaemic exposure 20–22, especially at ambient glucose concentrations >12 mmol/L 7. The reduced superoxide production may reflect an increase in polyol pathway activity as a consequence of raised intracellular glucose 23, 24. As an electron donor, NADPH is a necessary part of the polyol pathway and its consequently increased utilisation may be at the expense of reduced levels of superoxide production.
In an attempt to improve PMN function in diabetic patients, several investigators have studied the role of immune-modulating agents. Firstly, improving PMN superoxide production has been studied using aldose reductase inhibitors 22–25, agents that inhibit the polyol pathway. However, the role of such therapy in the prevention or treatment of infection in diabetes remains to be established. Secondly, and probably of greater clinical relevance, granulocyte-colony stimulating factor (G-CSF) has been used as adjuvant therapy in diabetic patients with foot infections. G-CSF induces terminal differentiation and release of PMN from the bone marrow and improves PMN function, particularly superoxide production 26–28. In one trial 26, diabetic patients with lower limb infections who received G-CSF showed earlier microbial eradication from wound swabs, quicker resolution of cellulitis, and shorter duration of intravenous antibiotic administration and hospitalisation. Unfortunately, these findings were not confirmed in a similar, subsequent study 28. Given the cost and paucity of data, G-CSF cannot yet be recommended as an adjunctive therapy for infections in diabetic patients.
Detailed study of cell-mediated immunity (CMI), predominately T-lymphocyte function, has identified specific defects in poorly controlled type 1 patients 29, 30. However, in a recent in vitro study involving type 1 patients, all with a HbA1c <8.0% 31, there were impaired proliferative CD4+ cell responses to primary protein antigens, perhaps due to altered expression of cellular adhesion molecules and/or reduced cytokine release independent of glycaemia 32. By contrast, another study involving patients with relatively good metabolic control showed a robust secondary immune response to standard antigens, suggesting normal T memory cell and CD4+ lymphocyte function 33. Thus, although tight blood glucose control may help to normalize CMI in diabetic patients, other factors may be important.
With regard to humoral immunity, it is possible that glycosylation may impair the biological function of antibodies, whether generated by natural exposure or vaccination. IgG glycosylation occurs in patients with diabetes in proportion to HbA1c34. In one study, the antigen-binding fragment of IgG was glycosylated in preference to the effector fragment 34. The clinical relevance of these observations is unclear since the antibody response and protection after vaccination against the common infections, influenza, pneumococcal infection and hepatitis B show adequate responses in patients with diabetes 35–44. Thus, we support the recommendation that adult diabetic patients receive the influenza vaccine before every epidemic season and the polysaccharide pneumococcal vaccine on one occasion followed by a booster at 5 years 45–49.
Common community-acquired infections
Contemporary management of common infections in diabetes is summarized in Table 1, with an emphasis on diagnosis and antimicrobial therapy. In the following sections, features of these infections in diabetic patients are discussed.
|Diagnostic tests||Organisms||Empirical antimicrobial treatmenta||Other treatments|
|Skin and soft tissueb|
|Cellulitis 50||Clinical diagnosis +/− wound swab for culture||S. aureus, S. pyogenes, less common gram-negative aerobes and anaerobes||Nafcillin/flu/dicloxacillin/1–2 g/IV 4–6 hourly||Cephazolin 1 g IV 8 hourly or clindamycin 600 mg IV 8 hourly or 300–450 mg PO 8 hourly or vancomycin 1g IV 12 hourly|
|Diabetic foot infection 51||Culture of tissue specimen from base of debrided ulcer, ‘probe to bone’ testc, operative biopsy for culture, plain radiograph||S. aureus, S. pyogenes, gram-negative aerobes including Pseudomonas aerigunosa, anaerobes. Often polymicrobial.||Mild–moderate||Cephalexin 500 mg PO 6 hourly plus metronidazole 400 mg PO 8–12 hourly, broad-spectrum fluoroquinolone, or ciprofloxacin 500 mg PO 12 hourly plus clindamycin 600 mg IV 8 hourly or 300–450 mg PO 8 hourly.||Surgical review, assess requirement for revascularisation, wound debridement, avoidance of pressure-induced ischaemia with orthotic devices.|
|Usually an oral regimen to cover gram + ve and common gram-negative organisms. e.g. Amoxicillin-clavulanate 875/125 mg PO 12 hourly OR ampicillin/sulbactam 3 g IV 6 hourly|
|Severe||Ciprofloxacin 750 mg orally 12 hourly plus clindamycin 600 mg IV 8 hourly or lincomycin 600 mg IV 8 hourly Add vancomycin 1g IV 12 hourly if MRSA isolated|
|Ticarcillin-clavulanate 3.0–0.1 g IV 6 hourly or Piperacillin-tazobactam 3.375 g IV 8 hourly or Mero/imipenem-cilastatin 500 mg–g 8/6 hourly|
|Necrotizing fasciitis 50||Clinical appearance/surgical findings/imaging (plain radiography/CT/MRI)/gram stain and culture of operative specimens||S. pyogenes or Clostridium sp. or polymicrobial||Meropenem 1 g IV 8 hourly plus clindamycin 600 mg IV 8 hourly or lincomycin 600 mg IV 8 hourly||Ampicillin-sulbactam 1.5–3 g 6–8 hourly or piperacillin-tazobactam 3.375 g 6–8 hourly, plus ciprofloxacin 400 mg IV 12 h plus clindamycin 600–900 mg IV 8 hourly||Surgical removal of devitalized tissue is the basis of treatment.|
|Community-acquired pneumonia52||Chest radiography, sputum gram stain and cultures, blood cultures, urinary antigen testing, serology||Streptococcus pneumoniae, Mycoplasma pneumoniae, Chlamydia pneumoniae, Legionella spp., Haemophilus influenzae, S. aureus, Klebsiella pneumoniae, Mycobacterium tuberculosis||Mild|
|An advanced macrolided or amoxycillin 1 g PO 8 hourly plus doxycycline 200 mg PO for first dose and then 100 mg PO daily||A respiratory fluoroquinolonee|
|A respiratory fluoroquinolonee|
|An advanced macrolidedPlus a β-lactam (benzylpenicillin 1.2 g IV 6 hourly or ampicillin 1 g IV 6 hourly or cefotaxime/ceftriaxone 1 g IV daily or ampicillin-sulbactam)|
|Severeb||An antipseudomonal β-lactam plus either an advanced macrolided or a respiratory fluoroquinolonee||Requirement for antipseudomonal cover will depend on risk factorsfLocal prevalence rates of community-acquired MRSA should guide empiric treatment and may include vancomycin 1 g IV 12 hourly or linezolid 600 mg IV 12 hourly|
|Ceftriaxone 1 g IV daily or cefotaxime 1 g IV 8 hourly plus either an advanced macrolided or a respiratory fluoroquinolonee|
|Asymptomatic bacteriuria 53||Urine microscopy and culture ≥105 cfu/mL||Various, most frequently Enterobacteriaceae||No treatment|
|Cystitis 54||Urine microscopy and culture||Enterobacteriaceae, Staphylococcus saprophyticus, Enterococcus sp., rarely Candida spp.||3 days oral antibiotic treatment guided by local antibiotic sensitivity pattern e.g. trimethoprim- sulfamethoxazole orally daily||Trimethoprim 300 mg or norfloxacin 400 mg orally 12 hourly or ciprofloxacin 500 mg orally 12 hourly|
|Pyelonephritis 54||Urine microscopy and culture/renal tract imaging||14 days oral/IV antibiotic treatment guided by local antibiotic sensitivity pattern.||Ceftriaxone 1 g IV daily or cefotaxime 1 g IV 8 hourly or ticarcillin-clavulanate 3.0–0.1 g IV 6 hourly or piperacillin-tazobactam 4.0–0.5 g IV 8 hourly||Emphysematous pyelonephritis is a rare but serious complication requiring early surgical intervention.|
|Post-treatment urine microscopy and culture recommended in all.|
|Ciprofloxacin 500/400 mg PO/IV 12 hourly or ampicillin 2 g IV 6 hourly plus gentamicin (see local dosing and monitoring guidelines) or ampicillin/sulbactam 1.5 g IV 6 hourly|
|Necrotizing otitis externa||Clinical examination/ear swab culture/magnetic resonance imaging||P. aeruginosa||Ciprofloxacin 400 mg IV 12 hourly or ticarcillin-clavulanate 3.0–0.1 g IV 6 hourly or cefepime 2 g IV 12 hourly or ceftazidime 2 g IV 8 hourly plus gentamicin 4–6 mg/kg IV daily||Imipenem-cilistatin/meropenem 1 g IV 6/8 hourly||Otolaryngology review.|
|Rhinocerebral mucormycosis||Clinical examination/magnetic resonance imaging/biopsy of infected tissue||Rhizopus (>90%), mucor and absidia species||Amphotericin B 0.8–1.5 mg/kg IV daily||Lipid-based amphotericin B for those with renal impairment or posaconazole 800 mg PO daily in 2–4 divided doses||Control diabetic ketoacidosis if present. Surgical debridement required.|
Skin, nail, mucous membrane and soft tissue infections
Skin infections are common in diabetes. In a large study of unselected diabetic outpatients, skin infections (mainly fungal) were present in 20% of the sample and were the commonest dermatological complaint 55. In another study analysing the epidemiology of inpatient care resulting from peripheral neuropathy, peripheral vascular disease and skin infections in a UK population of >400 000, diabetic patients had a 6–7 times greater risk of hospitalisation owing to skin and soft tissue infections and related conditions compared to age-matched controls 56. A recent Dutch primary care study that captured medically attended episodes of skin and mucous membrane infections showed odds ratios that were >30% greater for patients with either type 1 or 2 diabetes than for controls after adjustment for potential confounders 57.
Certain bacterial and fungal pathogens are found commonly on the skin and mucous membranes of diabetic patients, notably S. aureus and Candidia albicans. Evidence suggests that diabetic patients have an increased prevalence of staphylococcal carriage 58, although this is not a consistent finding 59. Poor glycaemic control appears to be a major predisposing factor 58, 60–62. In type 1 diabetic children, nasal carriage of S. aureus is higher (>70% in some studies) than in both age-matched controls and older type 2 patients 58, 59, and similar to that in young intravenous drug users 63. There is an association between colonisation with S. aureus and local and systemic infections, which are responsible for significant morbidity and mortality in diabetes 62, 64. In the case of common skin infections, S. aureus is the most common pathogen isolated in cases of cellulitis and 10% of patients with this condition will also have diabetes that is either known or unmasked by the infection 65.
The prevalence of onychomycosis is increased several fold in patients with diabetes, especially males 66. Treatment may be complicated by the presence of vascular disease, Candida infection and potentially significant interactions between antifungal drugs and concomitant blood glucose-lowering and cardiovascular therapies. Because of its high efficacy and good tolerability, terbinafine is the treatment of choice for this condition in diabetes 67.
Mucocutaneous candidiasis (usually due to C. albicans) is associated with diabetes and other hyperglycaemic states such as hyperalimentation 68, 69. This association may reflect increased microbial virulence since, in the presence of hyperglycaemia, Candida spp. express a protein that enables more avid adherence to surface epithelial cells 69. In clinical studies, however, findings have been inconsistent. In a recent review of 21 studies 70, oral carriage rates in diabetic subjects ranged from 18 to 80%. In the 15 studies that employed controls, only half showed a statistically significant increase in the carriage of C. albicans in diabetes, and there was a variable relationship between salivary glucose and oral carriage 68. Between-study differences are likely to reflect methods of selection of diabetic subjects and controls, sampling techniques, and perhaps also group differences in putative aetiological factors such as denture wearing and xerostomia 70.
Diabetic foot infections often lead to hospitalisation, and serious consequences such as osteomyelitis and amputation can supervene. This is due, at least in part, to their painless nature in patients with peripheral sensory neuropathy 71 and the fact that most non-severe diabetic foot infections do not produce systemic manifestations such as fever or leucocytosis 72. A culture of swabs collected from ulcers and other foot lesions in diabetic patients often yields several pathogens that can be difficult to distinguish from colonising organisms, especially in chronic or previously treated wounds 73. Evidence suggests that the culture results do not reliably identify the pathogens present unless the involved bone is cultured 74. The role of antibiotics for the treatment of foot ulcers in diabetic patients without clinical signs of infection remains unclear 75. However, surgery, especially consideration of revascularisation 76, remains a high priority in the management of the infected diabetic foot as part of multidisciplinary care 77. The roles of adjunctive therapies, such as growth factors and hyperbaric oxygen, are yet to be established 78.
Respiratory tract infections
In a large UK survey, respiratory tract infections (both upper and lower) were responsible for a significantly higher number of general practitioner consultations among diabetic patients compared with matched control subjects 79. However, in one prospective study, the percentage of diabetic patients hospitalized with community-acquired pneumonia (CAP) was < 5% 80, a figure similar to the prevalence of diabetes in many communities. The spectrum of organisms responsible for CAP in diabetic patients appears to differ from that in non-diabetic individuals, with an over-representation of S. aureus and Gram-negative bacteria such as K. pneumoniae81. The higher prevalence of staphylococcal skin and nasal carriage, increased pharyngeal carriage of Gram-negative bacteria and gastropathy-associated aspiration are likely predisposing factors 81. In a meta-analysis of 33 000 patients with CAP 82, a higher mortality was observed among diabetic patients, but diabetes was not among the three most common co-morbidities predicting death (smoking, pre-existing pulmonary disease and cardiac failure). Diabetes is one of the conditions associated with recurrent pneumonia 83.
Bacteraemic pneumococcal pneumonia occurs relatively commonly in diabetic individuals 84, 85, but a recent population-based study found a death rate that was similar to that in non-diabetic patients with this condition 86. However, the adverse impact of diabetes on morbidity and mortality during influenza epidemics has long been recognized 87. In fact, the first patient to receive insulin died following an influenza-like illness 88. Diabetic patients have a sixfold higher relative risk of hospitalisation than non-diabetic individuals during influenza epidemics 45. This observation might reflect an increased incidence of secondary staphylococcal pneumonia 45, but metabolic compromise and the high prevalence of co-morbid conditions, such as cardiac failure, are also important 89.
Pulmonary tuberculosis has a recognized association with diabetes. Studies carried out during the early part of the twentieth century revealed that diabetic patients with pulmonary tuberculosis had a mortality approaching 50% 90. Recent data from Asia and Africa demonstrate that pulmonary tuberculosis still poses a significant threat to diabetic subjects. A Japanese study that investigated patients hospitalized with tuberculosis over a 6-year period found that 13.2% had diabetes, a percentage that increased steadily with time 91. In a smaller but similar Pakistani study, nearly half of the sample had abnormal glucose tolerance 92. A study of bacterial infections in hospitalized diabetic patients in Papua New Guinea showed that the annual incidence of tuberculosis was 11 times higher than in the general population 93. In addition to an increased risk of tuberculosis, diabetic patients are prone to unusual forms, including predominant lower lobe involvement, multilobar disease and a higher incidence of pleural effusion 94. Diabetic subjects respond appropriately to anti-tuberculous treatment 95. An important treatment-related issue is the possibility that peripheral neuropathy may worsen because of isoniasid therapy unless pyridoxine supplementation is given 96. Interactions between anti-tuberculous and oral hypoglycaemic drugs can occur, including reduced efficacy of sulfonylurea treatment with rifampicin.
Urinary tract infections
Several factors are thought to predispose diabetic subjects to urinary tract infections (UTIs) 21, 97. Reduced sensitivity and altered distensibility of the urinary bladder due to autonomic neuropathy can result in stagnation of urine and higher rates of instrumentation. Glycosuria can enhance bacterial growth and impair phagocytosis. Vaginitis and renal microangiopathy are also associated with recurrent UTIs.
Asymptomatic bacteriuria (ASB), the presence of single bacterial species at > 105 colony forming units/mL in culture of at least two mid stream urine specimens from an individual without symptoms, is the most frequently observed category of UTI in diabetic patients. ASB has a reported prevalence of up to 29% in diabetic women, a figure that is approximately three times greater than that in non-diabetic women 98. In diabetic men, however, the overall prevalence of ASB (1 to 11%) is similar to that in non-diabetic men 99, consistent with the female preponderance of UTIs in the general population. A number of predisposing factors for ASB have been reported. In most 99–102 but not all 103 studies, there is no relationship between glycaemic control and ASB. Chronic complications such as nephropathy and neuropathy have been associated with ASB in type 1 but not type 2 diabetes 99, 104, a pattern that holds for longer diabetes duration 104. ASB is more frequent in patients with cardiac and peripheral vascular disease 105. Localisation studies in diabetic patients using techniques such as bladder wash-out have shown that the upper urinary tract is involved in more than half of the patients 106. Despite this, the decline in renal function in diabetic patients with ASB followed for up to 2 years did not differ from a group with diabetes alone 107. Diabetic women with ASB are at greater risk of UTI 100 and subsequent hospitalisation with urosepsis 103. However, there is no evidence that presumptive antibiotic therapy for ASB is beneficial 101, 108.
Acute pyelonephritis also occurs more commonly in diabetic than non-diabetic individuals. Post mortem studies during the pre-antibiotic era found a fivefold increase 97. A Canadian study of inpatients with pyelonephritis 109 revealed that 63% of women and 21% of men had diabetes, indicating that diabetes is a predisposing factor regardless of gender. Complications of acute pyelonephritis, both systemic and renal, also occur more often and with greater severity in diabetes. These include bilateral renal involvement, intrarenal abscesses, renal carbuncle and emphysematous pyelonephritis 97, 101. Indeed, some authorities recommend routine imaging of the renal tract in all diabetic patients with pyelonephritis to rule out obstruction and emphysematous and other changes 97, 110. Because of the potential for complications, UTIs should be carefully managed in diabetes. Attempts should be made to culture the organism both to guide antibiotic therapy and to confirm cure after treatment, especially since there is evidence that resistant organisms are found more frequently in diabetic patients with community-acquired UTI 111.
Available evidence suggests that diabetes is associated with increased periodontal disease and tooth loss, but data implicating diabetes as a risk factor for dental caries are inconsistent 112. Cross-sectional studies have revealed a high prevalence of periodontal disease in both type 1 and type 2 diabetes 113, 114. Although sample sizes have been small and the criteria used in the assessment of gingival involvement have been variable, a number of studies have found an association between poor glycaemic control and periodontal disease 114–117. Microangiopathy, abnormal collagen metabolism, raised salivary glucose concentrations and low salivary pH may also be contributing factors 113, 114, 117, 118. Poor glycaemic control can increase the incidence and accelerate the progression of periodontal disease 114, 116, 117, while, in the presence of periodontal infection, glycaemic control can worsen 119. Good oral hygiene remains an essential component of diabetes care but the addition of a tetracycline may provide additional benefit in type 1 patients with periodontal defects 120.
Glycaemic control and community-acquired infections
Few clinical studies have looked at the relationship between glycaemic control and the general risk of community-acquired infection and the data do not support a strong link. In the UK Prospective Diabetes Study(UKPDS) involving type 2 patients 3, intensive blood glucose control strategies prevented chronic vascular complications but infection-related data have yet to be reported. In the Diabetes Control and Complications Trial (DCCT) in type 1 patients 4, there was a 46% lower rate of self-reported vaginitis requiring medical treatment in intensively treated versus conventionally managed patients, but there were no other between-group differences in rates of foot, urinary, respiratory or gastrointestinal infections 121. In a community-based, 12-month prospective study of type 2 patients and their non-diabetic partners 122, similar proportions of subjects in each group reported any infection but there was a significantly greater number of infections in the diabetic group. There was no relationship between the incidence of infection and HbA1c122. By contrast, a retrospective study of diabetic outpatients demonstrated a significant correlation between infection and glycaemia assessed from blood glucose levels 21.
The relationship between glycaemic control and predisposition to specific community-acquired infections, as reviewed above, is also generally weak. The strongest associations are for periodontal disease and S. aureus skin carriage, but respiratory infections, UTIs (including ASB) and mucocutaneous carriage of C. albicans show no consistent relationship. This includes a number of studies performed before the DCCT 4 and UKPDS 3 results were reported (in 1993 and 1998, respectively), when diabetic patients may not have had as good a glycaemic control as in later years.
Most of the data relating to hospital-acquired infections in diabetes are from studies of post-operative wound infections, especially after cardiac surgery 123–132. In one large prospective study 126, there was a 2.7-fold increased risk of wound infection among diabetic patients. This increased risk was related to hyperglycaemia during the first 48 h after surgery but not HbA1c or pre-operative glycaemia. Patients with previously undiagnosed diabetes had a similar infection rate to that in those with known diabetes 126. The importance of post-operative metabolic control is further highlighted by the observation that a plasma glucose > 12 mmol/L on the first post-operative day increased the rate of hospital-acquired infection (excluding simple UTI) almost sixfold 129.
Owing to the substantial morbidity, mortality and socioeconomic impact associated with deep sternal wound infection (DSWI) following cardiac surgery, several investigators have targeted modifiable risk factors such as glycaemic control. In a large prospective study 133, diabetic patients undergoing open-heart surgery were allocated non-randomly to conventional intermittent subcutaneous insulin or to an intensive continuous intravenous insulin (CII) infusion. CII achieved better post-operative glycaemic control and was associated with a 2.5-fold lower rate of DSWI, results that were in accord with a previous retrospective review 132. In a study of PMN function among diabetic patients randomized to CII or standard insulin therapy after cardiac surgery 134, there was greater phagocytic activity in PMN taken from the CII group immediately after operation. The investigators hypothesized that insulin-mediated hormonal or cytokine effects improved PMN function 134.
There is increasing evidence that hyperglycaemia may substantially increase the risk of hospital-acquired infection in the severely ill, regardless of diabetes status 135, 136. In the Leuven Study 136, critically ill patients in a surgical intensive care unit were randomized to either CII titrated to maintain normoglycaemia or less intensive treatment (target blood glucose 10–11 mmol/L). Mortality was significantly lower in the intensive compared to the less intensive insulin group (4.6% vs 8% respectively), with the greatest reduction in mortality in patients with multi-organ failure and a proven septic focus. CII also reduced the incidence of septicemia by 46%, with a concomitantly reduced need for prolonged antibiotic therapy. More recently, a similar study in a medical ICU showed no improvement in overall mortality or infection-related outcomes in patients given intensive insulin therapy, except for a lower death rate in patients whose ICU stay was ≥3 days 137. Given these inconsistent results, the question of whether selected groups of critically ill patients will benefit from CII should be addressed in further adequately powered multi-centre prospective studies.