Retrospective study of outcome in patients treated for Staphylococcus aureus bacteremia

Authors


Corresponding author and reprint requests: Pauline E. Gosden, Department of Microbiology, Level 8, Bristol Royal Infirmary, United Bristol Healthcare Trust, Marlborough Street, Bristol, BS2 8HW, UK Tel: 0117-9282567 Fax: 0117-9299162

Abstract

Objective: To investigate whether a change in current treatment practice for Staphylococcus aureus bacteremia from flucloxacillin and aminoglycoside to flucloxacillin and fusidic acid was associated with any changes in outcome.

Method: A retrospective analysis was carried out of 316 episodes of S. aureus bacteremia diagnosed and treated in a tertiary hospital complex between 1983 and 1993. Outcomes considered were (1) death related to the infection and (2) relapse following cessation of antibiotic therapy.

Results: Mortality related to infection, which occurred in 24% of patients, was unrelated to treatment with the combination of flucloxacillin and fusidic acid; however, increasing age was a significant risk factor (OR per decade = 1.35, 95% CI = 1.18-1.55), and increasing duration of treatment (OR per week of treatment = 0.63, 95% CI = 0.52-0.77), use of flucloxacillin (OR = 0.30, 95% CI = 0.14-0.64), presence of an intravascular device (OR = 0.39,95% CI = 0.20-0.78) and presence of a skin lesion (OR = 0.51, 95% CI = 0.26-0.99) were significant protective factors. The only factor significantly related to relapse, which occurred in 11% of patients, was treatment with the combination of flucloxacillin and fusidic acid (OR = 0.32, 95% CI = 0.12-0.85). There was approximately a 70% reduction in the risk of relapse if this combination was used.

Conclusions: This retrospective analysis suggests a clinically important protective effect of fusidic acid against relapse in patients with S. aureus bacteremia. Although the results were adjusted for potential confounding factors, the possibility of bias remains. There is a need for a prospective randomized trial to evaluate the effectiveness of flucloxacillin and fusidic acid for treating S. aureus bacteremia.

INTRODUCTION

Staphylococcus aureus is the most frequent cause of bacteremia due to Gram-positive bacteria and the commonest cause of wound infection in the UK. The reported mortality associated with S. aureus bacteremia ranges from 21% to 43% [1–5]. Factors that predispose to infection with S. aureus include breaches in the skin and the presence of foreign bodies, although infection may arise without a predisposing factor. Infection with S. aureus is frequently associated with bacteremia and metastatic disease [6–8].

Conflicting observations have resulted in uncertainty concerning the optimum duration of antibiotic therapy for bacteremia caused by S. aureus [8]. Many believe that all patients with S. aureus bacteremia should be treated with a 4–6 week course of intravenous antibiotics [9,10] in order to prevent serious sequelae such as endocarditis. Recently, some authors have suggested that shorter courses of therapy for catheter-related bacteremia and uncomplicated right-sided endocarditis in intravenous drug users would be adequate [11–16], whereas others have maintained that short-course therapy results in unacceptable rates of complication and relapse [17,18].

Fusidic acid is frequently used in the UK but rarely in the USA [19,20]. Use of fusidic acid was not emphasized in the laboratory recommendations for the treatment of S. aureus bacteremia in the United Bristol Hospital Trust (UBHT) complex of hospitals until 1991, when a laboratory recommendation of flucloxacillin and fusidic acid, given as the sodium salt for 4 weeks, was introduced.

At this time it was felt that an antibiotic combination including fusidic acid would, because of its better tissue penetration, offer benefit in terms of outcome for the patient over a combination involving an aminoglycoside. We report a review of the outcome of patients treated for S. aureus bacteremia in the UBHT between 1983 and 1993 with particular emphasis on the impact of the use of fusidic acid.

PATIENTS AND METHODS

Data were collected by a retrospective review of case records for patients diagnosed and treated for S. aureus bacteremia between January 1983 and December 1993 in the UBHT hospital complex. These hospitals form a large teaching hospital unit with regional specialities including cardiac surgery, oncology services and neonatal intensive care. In 1993, 54 000 inpatients were treated in the UBHT. In addition to drawing patients from the central and south Bristol areas, it also acts as a referral centre for the southwest of England. The Bristol catchment population was 361 000 in 1983, rising to 370 000 in 1991 (estimates based on Office of Population and Censuses Surveys 1993 mid-year population).

The UBHT cases were identified by searching laboratory records of blood cultures, collected between January 1983 and December 1993, from which S. aureus had been grown. Patients with no clinical diagnosis of infection and who were not treated for infection, and those with polymicrobial bacteremia without corroborating clinical or laboratory evidence of S. aureus infection, were excluded.

The method of obtaining blood cultures changed in July 1988. Prior to 1988, blood cultures were performed by inoculation of 5–10 mL of aseptically drawn blood into brain–heart infusion (Oxoid CM 225, Unipath Ltd, Basingstoke) and thioglycollate broths. Broth cultures were subcultured onto solid media after 2 and 7 days. After July 1988 a commercial automated blood culture system was adopted (Bactec 660, Becton Dickinson UK Ltd, Oxford). Identification of S. aureus was carried out by standard methods [21]. Antibiotic susceptibility was determined by Stokes’ method for the following antibiotics: penicillin, gentamicin, fusidic acid and erythromycin [22]. Susceptibility to methicillin was determined with a paper strip containing 25 μg methicillin, on blood agar and mannitol salt plates incubated at 30°C and 37°C respectively.

Demographic data recorded from the notes included date of birth, sex, age at time of the positive blood culture, and whether infection was acquired in the community or the hospital. Associated conditions which might influence outcome, such as malignancy, diabetes mellitus, rheumatoid arthritis, cardiac disease, intravenous drug use and immunosuppression (asplenia, steroid therapy, neutropenia, cytotoxic therapy), were documented. Possible risk factors for infection, such as bone fracture, skin lesions, transcutaneous foreign bodies, operation within the previous year, prosthetic joints, bone screws and nails and prosthetic heart valves, were also recorded. Microbiological data included the date of the positive blood culture, and whether S. aureus was isolated from the respiratory tract, urinary tract, skin/wound lesions or intravascular line tips. Treatment information recorded included antibiotics and their duration of use.

Information about the outcome of treatment included the date of death, whether death was related to infection and whether a relapse occurred. The infection was classified as community acquired if the blood cultures had been collected within 48 h of hospital admission from the community and as hospital acquired in all other cases. A relapse was said to have occurred if a patient developed a new suppurative complication or blood culture with S. aureus within 5 years following cessation of antibiotic therapy. In 92% of cases, relapse occurred within 10 months. Phage typing was documented in only three cases and showed identical phage types for both isolates. Antibiotic susceptibility profiles were identical in all cases with the exception of one. Death was considered to be related to infection (DRI) if the death resulted from acute sepsis syndrome or a complication of S. aureus bacteremia, such as endocarditis. Post-mortem data were used when available.

Information recorded from the notes was transferred to a database. Descriptive statistics for demographic and relevant clinical data were calculated, including the proportions of patients receiving fusidic acid and the proportion experiencing the outcomes of interest, in each year. The effects of various clinical factors and treatment combinations on each outcome were then investigated in more detail by statistical modeling using multiple logistic regression, allowing estimates of odds ratios (ORs) for different factors to be obtained with and without adjustment for possible confounding variables.

RESULTS

The total number of patients for whom blood cultures were requested from 1985 to 1993 was 66 447 and requests showed an upwards trend from 6523 in 1985 to 9193 in 1993. Data for total requests were unobtainable for 1983 and 1984, as laboratory computerization was not implemented until 1985. The annual total of all patients with positive blood cultures rose from 437 in 1985 to 739 in 1993, but the proportion positive for S. aureus remained constant at approximately 11% of the total. Between January 1983 and December 1993, 522 patients with blood cultures positive for S. aureus were identified from manual and computer records. From this initial total, 206 (39.5%) were excluded or not evaluated; case records were unavailable for 100 (19.2%), 26 (5%) had polymicrobial sepsis and 80 (15.3%) showed no clinical evidence of infection (Table 1).

Table 1.  The distribution of positive blood cultures over the period of the study, the number of S. aureus cases in each of the years studied, the number of missing records, those not clinically significant and the number of patients who experienced the two outcomes of interest
YearBlood cultures positive for S. aureusNot clinically significant n(%)aMissing records n(%)aS. aureus cases included in study n(%)aDying from infection n(%)bRelapsen(%)b
  1. aPercentages of the total number of cases with blood cultures positive for S. aureus in the year.

  2. bPercentages of the number of cases included in the study from the year.

1983270 (0)4 (15)23 (85)5 (22)4 (17)
1984220 (0)7 (32)15 (68)3 (20)2 (13)
19855215 (29)5 (10)32 (62)9 (28)5 (16)
19865323 (43)9 (17)21 (40)5 (24)4 (19)
19874817 (35)10 (21)21 (44)5(24)2 (10)
19885619 (34)6 (11)31 (55)4 (13)2 (6)
1989455 (11)9 (20)31 (69)9 (29)3 (10)
1990456 (13)8 (18)31 (69)8 (26)1 (3)
1991575 (9)17 (30)35 (61)8 (23)7 (20)
1992509 (18)10 (20)31 (62)8 (26)3 (10)
1993677 (10)15 (22)45 (67)11 (24)2 (4)
Total522106 (20)100 (19)316 (61)75 (24)35 (11)

The following analyses are therefore based on the data from 316 patients with new episodes of S. aureus bacteremia, 94.7% of whom were admitted from the Bristol catchment area. The distribution of patients for whom medical records were not available was examined by calendar year, age, gender, admitting hospital and year of admission. Some records were unobtainable for each year throughout the study and ranged from four patients in 1983 to a maximum of 17 patients in 1991. There was no trend over time (see Table 1), or any indication from consideration of other variables that patients for whom records were missing were different from those for whom data were available.

The number of cases in each year ranged from 15 to 45. There appeared to be an increase in the number of cases occurring over time, paralleling the increase in total requests and positive blood cultures, despite an essentially constant catchment population (X2=23.65, degrees of freedom 10, p<0.01; linear regression of number of cases against time gave a β coefficient for each year of 1.95 cases, p=0.004). These trends may have been due to better detection of cases, better use of the laboratory or a real increase in numbers of cases.

The median age was 51.6 years with an inter-quartile range (Q1–Q3) of 20.8–68.5 years. However, the distribution of age was bimodal, with peaks in the distribution amongst both young and elderly. The >65 age group accounted for almost one third of cases. Staphylococcal bacteremia affected both sexes but there was a slight preponderance of males: 178 (56%) of the cases were male and 138 (44%) of the cases were female. Infection was classified as hospital acquired in 166 (53%) of the cases and as community acquired in the remaining 150 (47%). There was no evidence of a trend with time during the study period for either of these factors.

Antibiotic susceptibility

Most isolates were resistant to penicillin only (87%). Resistance to methicillin, gentamicin or fusidic acid was rare (0.9% for each of the antibiotics). Erythromycin resistance occurred in 4% of isolates. Multiple resistance was also unusual, with only 0.9% of isolates resistant to three or more of the antibiotics tested; these were all methicillin-resistant strains of S. aureus.

Antibiotic therapy

The number and type of antibiotics used in each patient for the definitive treatment of S. aureus bacteremia once blood culture results were known were analyzed. One hundred and forty-nine (47%) of the patients received two different antibiotics. Twenty (6%) of the patients received no antibiotic treatment, 57 (18%) received one antibiotic, 63 (20%) received three antibiotics, 22 (7%) received four antibiotics and five (1%) of the patients received five antibiotics.

Flucloxacillin was the preferred antibiotic and was used in 251 (79%) of the patients. Other antibiotics used included fusidic acid (38%), aminoglycosides (32%), cephalosporins (17%), erythromycin (7%), vancomycin (6%) and clindamycin (0.9%). In patients who received flucloxacillin it was used alone in 42/251 (17%) of these patients. Combination therapy was used in the remainder, with the addition of fusidic acid in 108/251 (43%), of aminoglycosides in 86/251 (34%), or of another antibiotic in 15/251 (6%). Twenty-two (9%) of the patients treated with flucloxacillin received both aminoglycosides and fusidic acid during therapy, but not simultaneously.

Figure 1 shows the pattern of use of flucloxacillin with fusidic acid or an aminoglycoside from 1983 to 1993. The numbers of patients who received the two commonest treatment combinations, flucloxacillin plus aminoglycoside and flucloxacillin plus fusidic acid, with or without additional antibiotics, are shown in Table 2. Also shown are the numbers of patients who received flucloxacillin alone and of those who did not receive any flucloxacillin. Figure 1 shows that the use of flucloxacillin remained high during the study period. Although the total number of patients receiving either fusidic acid or an aminoglycoside in combination with flucloxacillin remained fairly constant during the study period, the frequencies with which each combination was used changed over time. Use of fusidic acid decreased from 1984 until 1987 but increased thereafter, with a more dramatic rise in 1991 following the introduction of guidelines. Use of aminoglycosides remained fairly constant from 1983 to 1987, and then declined in parallel with the increase in use of fusidic acid.

Figure 1.

Treatment of S. aureus bacteremia with flucloxacillin and with fusidic acid or aminoglycoside, from 1983 to 1993.

Table 2.  Numbers of patients who received the commonest treatment regimens
YearCasesaFLU plus FUS with or without OTHFUS with or without OTHFLU plus AMINO with or without OTHAMINO with or without OTHFLU aloneFLU plus OTHOTH or none
  1. FLU=flucloxacillin; FUS=fusidic acid; AMINO=gentamicin or netilmicin; OTH=other antibiotic.

  2. aSome cases are included in more than one column, e.g. when a patient received flucloxacillin and fusidic acid and flucloxacillin and aminoglycoside at different times during treatment.

19832373101105
1984155270210
19853241115535
19862130100423
19872170100222
19883180102562
19893110181742
19903171112365
1991358253585
19923120110522
19934529231318
Total316108138615423539

The types of aminoglycoside used were gentamicin in adult patients and netilmicin in pediatric patients. If aminoglycosides were used, serum levels were considered adequate if the post-dose was >6 mg/L on at least one occasion when a multiple dosing regimen was used. Levels were adequate in 57/101 (56%) of the patients and inadequate in 26/101 (26%) of the patients. In the remaining 18 cases aminoglycoside levels were either not documented or not taken. Antibiotics were discontinued because of drug side effects in only 24 patients. Antibiotics were discontinued in seven patients on aminoglycosides (renal toxicity), nine patients on fusidic acid (rash 1, jaundice 6, vomiting 1, fluid overload 1), six patients on flucloxacillin (rash) and one patient on cephalosporins (rash); and in one patient both flucloxacillin and fusidic acid were discontinued (rash).

The median duration of antibiotic treatment for 316 bacteremic patients was 15 days (Q1-Q3=8–29) and almost 90% of patients were treated for less than 6 weeks.

Factors related to treatment outcomes

The numbers of patients who experienced the two outcomes of interest in each year during the study period are shown in Table 3. Two hundred and forty-eight (78%) of the patients survived the infective episode and showed no residual clinical signs of infection. Death related to infection occurred in 75 (24%) of the patients. There were relatively few instances of relapse, which occurred in only 35 (11%) of the patients. Twelve patients who relapsed also died from the staphylococcal infection.

Table 3.  Numbers of patients who died or relapsed
YearCasesRelapseDeath related to infection
19832345
19841523
19853259
19862145
19872125
19883124
19893139
19903118
19913578
19923138
199345211
Total3163575

Death related to infection (DRI)

Risk factors for DRI were investigated in more detail using multiple logistic regression. Age and duration of treatment were fitted as continuous variables, but divided by 10 and 7 respectively to give odds ratios (ORs) per decade and per week. The odds of DRI increased significantly with increasing age, but decreased with increasing duration of treatment (Table 4). The possibility of an increased risk of DRI amongst both the young and elderly was investigated by fitting a quadratic term, but this did not increase the likelihood ratio significantly. Year of treatment was also considered, but had no significant effect on DRI after controlling for age and duration of treatment (OR=1.03; confidence intervals (CI)=0.93–1.13), or in any subsequent models that were fitted.

Table 4.  Estimates of odds ratios for the factors associated with the risk of death related to infection
 Odds ratio (OR)95% confidence interval (CI)z-valuep-value
Age (per decade)1.351.18–1.554.420.001
Duration of treatment (per week)0.630.52–0.77−4.58<0.001
Treatment with flucloxacillin0.30.14–0.64−3.150.002
Presence of IV device0.390.20–0.78−2.680.007
Presence of skin lesions0.510.26–0.99−1.980.048

After adjusting for age and duration of treatment, four other factors appeared to be independently and significantly associated with a reduction in the odds of DRI: treatment with flucloxacillin (with or without any other antibiotic), infection acquired in the community, presence of an intravenous device and presence of a skin lesion. When all of these factors were modeled together, the factor of community-acquired infection failed to reach significance but, from its effect on the odds ratios for presence of an intravenous device and presence of a skin lesion, appeared to be positively associated with these latter two variables. Therefore, the final model included age, duration of treatment, treatment with flucloxacillin, the presence of an intravenous device and the presence of a skin lesion. ORs and CIs for these factors, based on this model, are shown in Table 4.

Because of the particular interest in the treatment combination of flucloxacillin and fusidic acid (with or without other antibiotics), a further analysis was carried out to compare the risks of DRI in patients who received this treatment combination and in those who were treated with flucloxacillin alone or in combination with any other antibiotics except fusidic acid. In this analysis patients who did not receive flucloxacillin were excluded. After adjusting for age, duration, intravenous device and skin lesion, no significant effect was demonstrated for the inclusion of fusidic acid in treatment, and the odds ratio estimates for all the other factors in the model were essentially the same as those shown in Table 4.

Relapse

A similar approach was used to investigate risk factors for relapse. Although neither age nor duration of treatment reached significance, estimates of the effects of other factors were always adjusted for these factors because one would intuitively expect them to be associated with the likelihood of relapse. The only factor that appeared to be significantly related to relapse was treatment with the combination of flucloxacillin plus fusidic acid; there was approximately a 70% reduction in the odds of relapse if this combination was used (OR=0.32, 95% CI=0.12–0.85, p=0.02). Estimates from the final model of the ORs and CIs for age, duration of treatment and treatment with the combination of flucloxacillin and fusidic acid are shown in Table 5.

Table 5.  Estimates of odds ratios for the factors associated with the risk of relapse
 Odds ratio (OR)95% confidence interval (CI)z-valuep-value
Age (per decade)1.060.93–1.210.860.392
Duration of treatment (per week)1.080.98–1.181.610.107
Treatment with flucloxacillin and fusidate0.320.12–0.85−2.280.023

DISCUSSION

This retrospective survey of S. aureus bacteremia is the only large study reported from the UK in the last 10 years. The risk factors identified were similar to those reported in previous studies [4,7]. The total number of cases of bacteremia showed an increase during the study period. The number of cases attributed to S. aureus as a proportion of all causes of bacteremia did not change and is similar to national data [23]. Bacteremia with S. aureus occurs at all ages, and in our study almost one third of cases occurred in the over-65-year age group. Infection acquired in hospital occurred in slightly more than half of cases, as reported in other studies [4,24].

In this study we determined the outcome following treatment of S. aureus bacteremia. Unresolved issues relating to the treatment of this condition include whether a single antibiotic or a combination of antibiotics should be used, which antibiotic(s) should be used and what should be the optimum duration of therapy. Whether a single antibiotic or a combination of antibiotics is associated with a better outcome for the patient is not known. It has been suggested that any antibiotic to which the isolate is sensitive, as shown by in vitro laboratory testing, would be appropriate [4,25]. Animal models have demonstrated that a combination of a penicillin with an aminoglycoside is synergistic [26,27] and moderately improves the outcome in experimental infective endocarditis, but there are no data showing a benefit from this type of combination in patients with endocarditis [26]. Some authors advocate the use of 2 weeks of an aminoglycoside in addition to a β-lactam agent for left-sided S. aureus endocarditis but not right-sided endocarditis [28,29]. In the UK, combination therapy for S. aureus endocarditis is usually employed. In this study 47% of patients received two different antibiotics, 18% received one antibiotic and 20% received three antibiotics. In practice, the latter patients initially received a combination of two antibiotics, and another antibiotic was used subsequently for completion of the course of therapy. Most current treatment regimens for S. aureus suggest the parenteral administration of a penicillinase-resistant penicillin as the mainstay of treatment. In the UK, both the British National Formulary and the BSAC Working Party's recommendations for the treatment of staphylococcal endocarditis include fusidic acid [20,30]. In Denmark, S. aureus bacteremia would usually be treated with fusidic acid in combination with other antibiotics such as methicillin or dicloxacillin [31]. American literature does not cite the use of fusidic acid, because it is not licensed for use in the USA.

In this study flucloxacillin was used in 79% of cases and in combination with either an aminoglycoside or fusidic acid in 27% and 34% of cases, respectively. The results suggest that the rate of death due to S. aureus infection remained essentially unchanged over a 10-year period, and was unrelated to a change in the preferred treatment from flucloxacillin and aminoglycoside to flucloxacillin and fusidic acid. This finding agrees with other studies which have shown similar mortality rates whether or not fusidic acid was used for therapy [32]. Increasing age was a significant risk factor for death due to infection, and longer duration of treatment and treatment with flucloxacillin were significant protective factors. These findings were not unexpected; older patients are more likely to be immunocompromised, flucloxacillin is a potent antistaphylococcal agent and treatment regimens are often given for prolonged periods.

The presence of an intravascular device and the presence of a skin lesion were significant protective factors. We suspect that these findings arose because the variables acted as ‘markers’ for subsets of patients who were less likely to die from infection. For example, the presence of an intravascular device may have been a marker for an infection that arose in the context of a removable focus of infection, and the presence of a skin lesion may have been a marker for bacteremia derived from a primary skin or wound infection which developed in an otherwise healthy individual in whom the prognosis would be expected to be good. This interpretation is supported by a previous study which found the only factor predictive of the outcome was the presence of a source of bacteremia. Patients in whom the source of bacteremia was defined had a better prognosis with regard to mortality rate, incidence of endocarditis and secondary foci [24].

Factors that were associated with death due to infection were not found to be associated with relapse, suggesting that the circumstances which predispose patients to experience these unfavorable outcomes may be different. Moreover, the treatment combination of flucloxacillin and fusidic acid appeared to give rise to about a 70% reduction in the risk of relapse, although this was not reflected in a significant decrease in the rate of relapse during the study period, or in a significant decrease in DRI.

It has not previously been shown that the combination of flucloxacillin plus fusidic acid is superior to other therapy for the treatment of S. aureus bacteremia. Fusidic acid is an effective narrow-spectrum antibiotic with excellent activity against S. aureus, with MICs in the range 0.03–0.25 mg/L. It has been in clinical use in the UK for 30 years and is regarded as a useful antistaphylococcal antibiotic [32]. Fusidic acid is very well absorbed after oral administration and shows high and long-lasting serum levels with therapeutic concentrations in bone tissue, synovial fluid, heart tissue and bronchial secretions. It has a proven ability to penetrate into avascular sites [33], and perhaps it is this latter property that enables it to act against possible secondary foci of infection in patients with S. aureus bacteremia and thus prevent relapse.

Many advocate short-course therapy in specific situations such as catheter-related bacteremia [11–13,15] or in uncomplicated right-sided endocarditis in intravenous drug users [14,16]. Others disagree and feel short-course therapy is inadequate and results in unacceptable rates of complications and relapses [17,18]. Others conclude that available data are not sufficient and the optimum duration of therapy remains unknown [8]. Early studies reported a risk of endocarditis of 60% with S. aureus bacteremia and thus long term parenteral antibiotic therapy became standard practice. More recent studies report an overall incidence of endocarditis of 10% or less[12,24,25]. Our study showed an incidence of 8.5%. These low rates of endocarditis have led to guidelines for short-term treatment (2–3 weeks) or long-term treatment (4–6 weeks), depending on whether endocarditis is likely to be present [25]. Use of the Nolan and Beaty criteria have been found useful in some at-risk populations [1,28] but not in others [18,34]. This study did show that patients were less likely to die of infection when treatment was continued for longer, although this finding is difficult to interpret. From a statistical point of view, the result is valid for treatment durations within the range of the original data, in this case up to 200 days. However, almost 90% of patients were treated for less than 6 weeks and the finding is based mainly on the data from these patients. Therefore we conclude that an increase in the duration of treatment up to 6 weeks confers protection against death from infection, but that the question of whether longer durations of treatment confer an additional benefit remains unanswered. Furthermore, this general conclusion does not rule out the possibility of interactions between duration of treatment and subsets of the population. For example, periods of treatment much shorter than 6 weeks may be adequate for young and otherwise healthy patients with catheter-related bacteremia. Our study did not have sufficient power to investigate such interactions.

The findings of a retrospective cross-sectional study must always be interpreted with caution, since this type of study design is susceptible to confounding and bias. Our analyses tried to take account of these alternative explanations. Confounding was investigated by modeling the effects of a wide range of other variables, characterizing demographic details, co-morbidity and aspects of treatment, none of which significantly altered the protective effect of the treatment combination of flucloxacillin and fusidic acid on relapse. The possibility of bias is more difficult to exclude, since data were unavailable for a substantial proportion of cases, the information obtained was dependent on the documentation available and the outcome for each patient was not ‘blinded’ to the treatment the patient received. However, it was reassuring that the findings were unaltered when patients not treated with flucloxacillin were omitted from the analysis. The results of this study suggest that a prospective randomized control trial based in a number of centers should be carried out to compare the use of flucloxacillin plus fusidic acid with another antibiotic regimen, such as flucloxacillin alone, for the treatment of S. aureus bacteremia.

Ancillary