NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
National Guideline Centre (UK). Cirrhosis in Over 16s: Assessment and Management. London: National Institute for Health and Care Excellence (NICE); 2016 Jul. (NICE Guideline, No. 50.)
13.1. Introduction
Bacterial infections are common in people with cirrhosis and are more frequent in people with decompensated disease. Once infection develops, renal dysfunction, hepatic encephalopathy and a deterioration of liver function may follow, which adversely affect survival. In people with cirrhosis, a bacterial infection is present either at the time of hospital admission or during hospitalisation in 25–35% of people63 and bacterial infection increases the probability of death, reaching 30% at 1 month and 63% by 1 year.12
The most common infections in people with cirrhosis are spontaneous bacterial peritonitis (SBP) (25%), followed by urinary tract infection (20%), pneumonia (15%), bacteraemia (either spontaneous or following a therapeutic procedure), and cellulitis.104 Clinical factors associated with an increased risk of infection are severity of liver disease, low ascitic fluid protein levels, prior SBP and hospitalisation.63 Enterobacteriaceae and non-enterococcal streptococci cause the majority of spontaneous infections in cirrhosis12,63 and translocation of enteric bacteria from the gut appears to be the main cause of Gram-negative infections in cirrhosis.104
Since most episodes of SBP in cirrhosis result from the translocation of enteric bacteria, the prophylactic administration of an oral antibiotic that effectively reduces the concentration of these bacteria in the gut whilst preserving the protective anaerobic flora could reduce the incidence of SBP. However, the use of prophylactic antibiotics is associated with a risk of developing multi-resistant organisms and infections caused by antibiotic-resistant bacteria have a higher hospital mortality than those caused by susceptible bacteria.63 Since these factors lead to uncertainty around the use of antibiotics for the primary prevention of SBP in people with cirrhosis and ascites, the GDG wanted to examine the clinical and cost-effectiveness of oral antibiotics compared to placebo in patients with cirrhosis at risk of SBP associated with the presence of ascites.
13.2. Review question: What is the clinical and cost-effectiveness of antibiotics compared with placebo for the primary prevention of spontaneous bacterial peritonitis (SBP) in people with cirrhosis and ascites?
For full details see the review protocol in Appendix C.
Table 89
PICO characteristics of review question.
13.3. Clinical evidence
One Cochrane review41 (comprised of adults with cirrhosis and ascites regardless of aetiology of cirrhosis or severity of the disease) included 9 studies overall, however only 5 relevant studies were included in this review;66,92,175,208,224. One additional RCT published since the Cochrane review was identified and included in this review.223 The included studies are summarised in Table 90 below. Evidence from the included studies is summarised in the clinical evidence summary below (Table 91). See also the study selection flow chart in Appendix E, study evidence tables in Appendix H, GRADE tables in Appendix J and excluded studies list in Appendix L.
Table 90
Summary of studies included in the review (all also included in Cochrane review).
Table 91
Clinical evidence summary: prophylactic oral antibiotics versus placebo.
Two studies included in the Cochrane review were excluded from this review because they looked at a head-to-head comparison of different antibiotics, not a placebo controlled comparison. The other 2 studies included in the Cochrane review86,204 were excluded from this systematic review because they included a large proportion (>15%) of patients who previously had SBP. Two papers,175,208 with small proportions of people with prior SBP, were included in this review. One study, Rolachon et al. (1995),175 reported how many patients developed SBP who also had SBP previously, so it was possible to amend the analysis to include primary prophylaxis only. The numerator and denominator were amended in the analysis for this outcome. For other outcomes it was not possible to determine how many cases in the numerator had previously had SBP and hence this study was considered as having an indirect population. For Soriano et al. (1991),208 it was unclear whether or not those who developed SBP had primary or secondary antibiotic prophylaxis, but as there were only 7 patients who developed SBP in the study (in the placebo group) and only 1 patient in this group had SBP previously, it was unlikely to alter the overall result substantially. Sensitivity analysis was performed by removing this participant and there was minimal effect.
All the studies used a combination of clinical, laboratory, and ultrasonographic data or histology to confirm cirrhosis. The participants were typically Child-Pugh B or C (7 or greater) but some studies included small numbers of participants who were Child-Pugh A (1 study92 did not report severity of cirrhosis). Five of the studies included people at high risk of SBP, defined by having an ascitic protein concentration of <15 g/litre. One study223 included people at low risk of SBP, defined by having an ascitic protein concentration of ≥15g/litre.
All studies compared a quinolone (either norfloxacin or ciprofloxacin) to placebo, usually administered for between 6 and 12 months, but 1 study208 treated patients for the hospitalised period only (<4 weeks) and another study only treated people for 1 month.223
Some outcomes prioritised for this review were not reported in the Cochrane review so these were extracted from the individual papers. These include time-to-event data for all-cause mortality, incidence of antibiotic-resistant organisms (the Cochrane review reported this for 2 relevant studies175,208), liver failure, renal failure, and length of hospital stay. There was no evidence found on health-related quality of life or on readmission rates.
For the 3 studies that reported all-cause mortality as time-to-event data it was possible to calculate the hazard ratio and these were combined in a meta-analysis. Since 3 of the studies did not report time-to-event all-cause mortality data, dichotomous data were presented in a separate forest plot. As each study reported at varying follow-up times, the results from these studies were not combined into a meta-analysis but were instead presented separately as subgroups by follow-up duration.
Four studies looked at antibiotic resistance but did not report in a format that could be combined in a meta-analysis. One study66 reported no cases of quinolone-resistant SBP in the 12-month study. However, it was unclear what investigations were performed or if investigations were performed for all patients. Two studies92,175 took stool samples from a proportion of patients to detect resistant microorganisms, and it was unclear why not all patients were tested. One of these studies 175 tested 10 of the 28 patients who were treated with antibiotics and reported that none had acquired resistance to ciprofloxacin after 6 months. The other study92 was a multi-centre study that tested 46 patients at 5 centres (24/53 antibiotic, 22/54 placebo) and found that 42% (10/24) of those tested who were treated with antibiotics had norfloxacin-resistant organisms at some point during the 6-month treatment period compared with 14% (3/22) of those in the placebo group. The final study223 reported that E.Coli resistant to ciprofloxacin was found in 6 out of the 7 people who developed a urinary tract infection during the trial.
All studies that reported liver failure only did so where this outcome lead to death, so this was extracted. While the I2 statistic was at an acceptable level for the forest plot on liver failure (see forest plot K.5 in Appendix K), there appeared to be heterogeneity in point estimates in opposite directions (that is, favouring opposing treatment). It was not possible to explore heterogeneity in the pre-specified subgroups (see the review protocol in Appendix C) since all treatments were in the same antibiotic class and trial participants had a similar level of risk of ascites (defined as <15 g/litre ascitic protein). Any subgrouping by severity would result in only 1 study in each subgroup, so this subgroup analysis was not performed. As it was not possible to explain the heterogeneity, a random effects meta-analysis was used and the outcome was downgraded for inconsistency.
Heterogeneity was also found for the meta-analysis on length of hospital stay, where each study reporting the outcome had point estimates favouring opposing treatments. It was not possible to explore heterogeneity for this outcome as there were only 2 studies, and any possible subgroup analysis would leave only 1 study in each subgroup. As it was not possible to explain the heterogeneity, a random effects meta-analysis was used and the outcome was downgraded for inconsistency.
13.4. Economic evidence
13.4.1. Published literature
No relevant economic evaluations were identified.
See also the economic article selection flow chart in Appendix F.
13.4.2. Unit costs
See Tables 96 and 97 in Appendix O.
13.5. Evidence statements
13.5.1. Clinical
- Antibiotics were considered more clinically effective than placebo for the following outcomes:
- Incidence of spontaneous bacterial peritonitis (6 studies, n=482, Moderate quality)
- All-cause mortality at:
- –
all time points (time-to-event data from 3 studies, n=263, High quality)
- –
∼1 month (n=59), ∼4 months (n=107), 6 months (n=53), all one study, dichotomous data and Low to Very Low quality.
- Renal failure (2 studies, n=168, Low quality).
- It was unclear whether or not antibiotics had a clinical benefit on liver failure leading to death or length of hospital stay compared to placebo (liver failure: 4 studies, n=328, Very Low quality; length of hospital stay: 2 studies, n=123, Very Low quality).
- There was no evidence on health-related quality of life.
13.5.2. Economic
- No relevant economic evaluations were identified.
13.6. Recommendations and link to evidence
| Recommendation |
|
|---|---|
| Research recommendation |
|
| Relative values of different outcomes | The main purpose of the use of antibiotics in people with cirrhosis and ascites is to prevent the occurrence of spontaneous bacterial peritonitis (SBP). The GDG was particularly interested in the impact of antibiotics on these patients in terms of their quality of life and whether or not antibiotics have an impact on overall mortality in these very sick patients. The GDG agreed critical outcomes of interest as the incidence of SBP, quality of life and all-cause mortality. Length of hospital stay, readmission rates, incidence of resistant organisms, renal failure, and liver failure were agreed to be important additional outcome measures which may be influenced by antibiotic treatment. The GDG noted that people with cirrhosis and ascites are generally managed in the outpatient setting but the development of SBP necessitates hospital admission. The GDG was interested not only in hospital admission rates but also whether length of hospital stay, associated with the occurrence of SBP or adverse events, was reduced for patients receiving antibiotics. As there is a risk of the development of antibiotic resistance, particularly with the long-term use of antibiotics, the GDG felt that it was important to consider this outcome. People with cirrhosis and ascites are at risk of renal or liver failure, so the GDG wished to consider whether the use of antibiotics influenced the occurrence of these complications. Length of hospital stay and admission rates were agreed to be important outcomes as they can be used as surrogates for quality of life and for the cost of treating SBP and any other complications or adverse effects. |
| Trade-off between clinical benefits and harms | The evidence presented identified a clinical benefit of prophylactic antibiotics on the critical outcomes of SBP and all-cause mortality for people with cirrhosis and ascites who were judged to be at a high risk of SBP (an ascitic fluid protein level of less than 15 g/litre). No evidence was available for the critical outcome of health-related quality of life. A clinical benefit of antibiotic prophylaxis was also seen in this group when assessing the important outcome of renal failure. The clinical effectiveness of antibiotics on the outcomes of liver failure and length of hospital stay was unclear. Overall, the GDG agreed that the clinical benefit observed for mortality and occurrence of SBP warranted the use of prophylactic antibiotics in people with cirrhosis, ascites and a high risk of SBP. The GDG acknowledged that this recommendation represents a change in UK practice, as only secondary prophylaxis is currently recommended. The mean length of treatment in most of the included studies varied between 6 and 12 months. However, the GDG recommended that treatment be continued for the total duration of time an individual has ascites. Antimicrobials should only be withdrawn once ascites has been successfully resolved, either by a re-compensation of the liver disease, sodium restriction and diuretic therapy, TIPS insertion or liver transplantation. The GDG noted 4 trials that reported data on the development of antibiotic resistance. The data were not sufficient to influence the GDG's recommendation as this outcome was only reported from a proportion of people in the trials; however, the GDG did not underestimate the importance of this outcome for healthcare professionals in deciding whether to initiate primary prophylaxis. It was noted that the Grange 1998 paper92 reported a 14% incidence in the development of norfloxacin-resistant organisms during the 6-month follow-up in the placebo group and a higher percentage in the treatment group (42%) among the patients whose stool samples they tested. However, stool samples were tested from fewer than half of patients. The GDG commented that all-cause mortality outcome data should include any instances of fatal antibiotic-resistant infections occurring during the treatment period. Evidence from 1 study was identified in people with ascites and a lower risk of SBP (ascitic fluid protein level of 15 g/litre or more). This study showed a similar benefit of antibiotics on the all-cause mortality to that in studies in people at high risk. However, unlike studies in people at high risk, this study appears to show no benefit of antibiotics on the outcome of SBP, albeit with very large confidence intervals. This apparent difference in the effectiveness of antibiotics may reflect the fact that people were at lower risk, or may be due to other factors such as the shorter treatment period of 1 month in this study. As this study carried a low weighting in the meta-analysis, there was no statistical heterogeneity in the outcome of SBP overall. Consequently, as this study carried little weighting and the majority of the evidence was in people at high risk of SBP, the GDG made their recommendation in people at high risk. The GDG chose not to make a recommendation for people with low risk of SBP. No evidence was identified for other antibiotics such as penicillin or other quinolones. The GDG acknowledged that they could not be certain about the clinical effectiveness of other antibiotics and therefore chose to limit the recommendation to the use of either oral ciprofloxacin or oral norfloxacin. |
| Trade-off between net clinical effects and costs | No relevant economic evaluations were identified. The GDG noted the low cost of therapy with ciprofloxacin (£40 for 500 mg per day for a year) and moderate cost of norfloxacin (£313 for 400 mg per day for a year), and that both costs and benefits of treatment could continue indefinitely for as long as the patient had ascites, which would vary considerably between individuals. However, the GDG also noted the high cost of treating an episode of SBP. Using a 5-day course of piperacillin/tazobactam, together with a 7-day hospital stay, paracentesis and an ultrasound would cost £1,925 to treat a single episode of SBP (see unit costs in Appendix O). This would be cost saving if prophylactic treatment with ciprofloxacin prevented just 1 case of SBP per 71 person-years of prophylactic treatment, and for norfloxacin 1 case of SBP per 6 person-years of prophylactic treatment. The clinical evidence reviewed in this chapter found prophylactic treatment to reduce incidence of SBP by an average of 142 cases per 1000 people (1 in every 7 people) within 6 months, and thus it would be cost saving with the average reduction of SBP in published studies. The benefits to all-cause mortality shown above make this intervention even more favourable. The GDG was therefore confident that the reduction in episodes of SBP and the costs associated with treating SBP would make primary prophylaxis using ciprofloxacin or norfloxacin cost saving or at worst highly cost-effective compared to no prophylaxis at a cost-effectiveness threshold of £20,000 per QALY gained. |
| Quality of evidence | Six studies were identified from the evidence search. The GDG noted that the studies spanned a 22-year period and that 3 studies were over 15 years old. The most recent study was from 2013. All but 1 study was included in a Cochrane review. The quality of available evidence for the critical outcome of the incidence of SBP was Moderate. The quality of the evidence for all-cause mortality (time-to-event) was of High quality. The quality of evidence for all other outcome measures was low or very low. For dichotomous outcomes, most studies were at risk of bias due to a comparable or a higher rate of drop-outs. While it was unsurprising that many patients dropped out given that these patients were very sick, there is still a risk that the effect estimate does not reflect the true effect. These patient drop-outs also had a larger effect on the outcome because of the small size of the trials, which are difficult to recruit to given the relatively rare occurrence of SBP. Furthermore, the GDG noted a smaller number of occurrences of liver failure than expected in both groups. They felt that this may be partly due to the small size of the studies which is likely to have influenced the imprecision for this outcome. It was noted that in 2 of the studies participants were included even if they had been diagnosed with SBP previously (secondary prophylaxis). The GDG agreed that these studies could be included, but only if 15% or fewer of the participants had previously had SBP. Two papers which were included in the Cochrane review were excluded from this review on this basis. In Gines 1990,86 all participants had prior SBP. In Singh 1995,204 14% in the treatment group and 17% in the no prophylaxis group had previously had SBP. The GDG noted that Singh 1995 was the only paper that compared trimethoprim-sulfamethoxazole with a placebo, but that it had been excluded, because of the pre-specified conditions in the protocol. The GDG also commented on the fact that a small number of participants were classified as Child-Pugh Grade A in the available evidence. The GDG felt that this may have been a mistake, as people with ascites would normally be classified as a minimum of Child-Pugh B. The GDG believed that the small number of participants who were classified as Child-Pugh A did not allow any recommendation to be made for low risk people with cirrhosis. The GDG noted the lack of evidence to assess the efficacy of prophylaxis in low risk groups (that is, with an ascitic fluid albumin level of greater than 15 g/litre). The GDG highlighted that the Soriano 1991 paper208 described a short period of prophylaxis (mean 26 days), whilst the participants were inpatients for another decompensating event. One study223 also reported a very short intervention period of 4 weeks which may explain the lack of effect of the antibiotic to prevent SBP. All other studies reported people receiving treatment for 6 to 12 months. The GDG agreed that Soriano 1991 should remain in the analyses, but considered only as indirect evidence as it may represent a slightly different population. The GDG felt that this was likely to be the reason for the heterogeneity in the outcome length of hospital stay. The Soriano 1991 study208 reported that, on average, those in the placebo group had a shorter hospital stay, while the other study which reported this outcome reported the opposite. The indirectness of this study did not affect the overall quality of the critical outcomes on which the GDG based their decision. The GDG commented on the lack of availability of good quality evidence on the development of antibacterial resistance. It was also noted that microbial resistance patterns vary between countries and may have changed significantly since these trials were conducted. |
| Other considerations | The GDG highlighted that the prime management goal is the treatment of ascites. Antimicrobial prophylaxis is not an alternative to the effective treatment of ascites but is seen as an additional measure. As SBP is an infection of the ascitic fluid, the GDG recommended that antibiotics should be stopped once ascites has resolved. The GDG noted that existing American Liver Association guidelines suggest that antibiotics for primary prophylaxis of SBP should be considered for people at high risk of developing this complication, which was defined as an ascitic fluid protein of less than 1.5 g/dl (15 g/litre) along with impaired renal function or liver failure. The GDG discussed the limitations of the certain antibiotics. Quinolones have interactions which may need to be taken into consideration for some people with cirrhosis and ascites (for example, with concurrent sulphonylurea treatment for diabetes). The GDG were uncertain of the wide availability of norfloxacin in the UK. The GDG also noted that other antibiotics are widely used for secondary prophylaxis, including co-trimoxazole, co-amoxiclav and colistin. The popularity of these antibiotics among healthcare professionals is due to a perceived lower incidence of antibiotic resistance and development of Clostridium difficile diarrhoea. There was no available evidence on these drugs to make recommendations for primary prophylaxis. Research recommendation Spontaneous bacterial peritonitis is the most common serious infection in patients with cirrhosis, occurring in 25% of people who develop ascites. It is associated with significant morbidity and mortality rates of 20–40%. It occurs most commonly in people with advancing liver disease; approximately 70% of cases occur in people with Child-Pugh class C cirrhosis. Several oral antibiotics that have been investigated for the prophylaxis of SBP have shown benefits and a significant reduction in the incidence of SBP in people at high risk of having, or developing, cirrhosis. They are, however, associated with antibiotic resistance, adverse reactions and drug interactions. There is a lack of good quality, recent evidence regarding the prevalence and consequences of antibacterial resistance that may occur during long-term oral antibiotic therapy when used for the prevention of spontaneous bacterial peritonitis. |
- b
At the time of publication (July 2016), neither ciprofloxacin nor norfloxacin had a UK marketing authorisation for this indication. The prescriber should follow relevant professional guidance, taking full responsibility for the decision. Informed consent should be obtained and documented. See the General Medical Council's Good practice in prescribing medicines – guidance for doctors for further information.
- Introduction
- Review question: What is the clinical and cost-effectiveness of antibiotics compared with placebo for the primary prevention of spontaneous bacterial peritonitis (SBP) in people with cirrhosis and ascites?
- Clinical evidence
- Economic evidence
- Evidence statements
- Recommendations and link to evidence
- Primary prevention of spontaneous bacterial peritonitis (SBP) in people with cir...Primary prevention of spontaneous bacterial peritonitis (SBP) in people with cirrhosis and ascites - Cirrhosis in Over 16s
Your browsing activity is empty.
Activity recording is turned off.
See more...