Rationally defining the appropriate duration of intravenous (IV) antibiotics for children with bacterial infections is challenging. For example, how long should a 2‐week‐old infant with a urinary tract infection (UTI) caused by Escherichia coli (E coli) be treated intravenously if the infant has responded to treatment and is back to baseline within 1 to 2 days? What if the blood culture was also positive for E coli? What are the risks and benefits of continuing IV antibiotics?

Such questions are common for pediatric hospitalists. Bacterial infections remain a relatively frequent cause of pediatric hospitalization, especially in neonates where 5 of the top 10 causes of hospitalizations are related to bacterial infections.[1] For some conditions, children remain hospitalized after clinical improvement simply for ongoing provision of IV antibiotics. Alternatively, some children are discharged home with a peripherally inserted central catheter (PICC) to complete an IV course.

The decision regarding the duration of IV antibiotics varies according to the condition for which the antibiotic is prescribed and often by practitioner or hospital. Many recommendations are numerically based (eg, 10 days for group B Streptococcus [GBS] bacteremia, 21 days for E coli meningitis), without taking into account patient‐level factors such as initial severity or response to therapy. These concrete recommendations may in fact be preferred by some practitioners, as suggested by a former chairman of the Committee on Infectious Disease for the American Academy of Pediatrics (AAP): The Red Book is designed for people who make decisions. It cannot waffle on an issue. It has to make a positive recommendation even if the data are incomplete.[2] A potential downside of this mentality, however, is that some practitioners may then feel obligated to follow these recommendations despite the lack of supportive evidence.

EXTENDING IV ANTIBIOTICS BEYOND CLINICAL RECOVERY

What is the rationale for continuing IV antibiotics in infants whose symptoms have completely resolved? Several factors likely drive these decisions: prevention of recurrences, concerns about bioavailability of enteral antibiotics and patient compliance, adherence to expert recommendations/guidelines, and perhaps a general sense that more is betterthat serious infections and/or their sequelae require more aggressive treatments.

Recurrence of a potentially life‐threatening infection is an understandable concern. Even when symptoms have resolved and there is documented clearance of the infection, such clearance does not necessarily signify that the body has rid itself of the pathogen completely. Some infections are deep seated and may warrant continuing treatment despite apparent recovery. To some, the risks of prolonging IV therapy may seem inconsequential when juxtaposed to a potentially devastating recurrence. However, in many conditions, recurrences may be related to host issues or ongoing exposures rather than inadequate treatment of the original infection. Recurrent UTIs, for example, are more likely in infants with urologic abnormalities,[3] and recurrent GBS bacteremia has been associated with GBS colonization of maternal breast milk and/or maternal mastitis.[4, 5, 6, 7] Although it is tempting to extend IV courses to prevent recurrences, it is not clear that the benefits of such an approach outweigh the risks.

Concerns over enteral absorption and bioavailability are also understandable, especially in young infants. The superior efficacy of IV over oral antibiotics in general is well accepted for many pediatric conditions, and in some cases (eg, septic shock) it would be unethical to perform a head‐to‐head trial. However, the lack of any published trials (to our knowledge) in pediatrics confirming the superiority of IV antibiotics suggests that oral antibiotic absorption is sufficient for many infections. Even in neonates, several studies have demonstrated that therapeutic serum levels are easily reached with oral dosing of amoxicillin in term and preterm neonates.[8, 9]

For the remainder of this review, the published recommendations and available evidence behind the duration of IV therapy are summarized for 4 bacterial infections in children in which IV antibiotic therapy often continues after clinical recovery: meningitis, bacteremia, UTI, and acute osteomyelitis. We conclude by proposing additional considerations for IV antibiotic durations, especially in situations where guidelines and/or evidence are either nonexistent, dated, conflicting, or contrary to evidence from published studies.

BACTERIAL MENINGITIS

The Infectious Disease Society of America and the British National Institute for Clinical Evidence have both published guidelines with pediatric recommendations for duration of therapy in bacterial meningitis,[10, 11] though the recommendations differ somewhat for 3 of the 4 most common pathogens, and are not always concordant with evidence from randomized controlled trials (Table 1).[12, 13, 14]

A recent meta‐analysis on duration of therapy in meningitis included 5 open‐label trials of ceftriaxone for bacterial meningitis in children.[12] These trials included the 3 most common pathogens and were categorized as short‐course (47 days, n=196 patients) and long‐course (714 days, n=187 patients) therapy. There was no significant difference in clinical success or long‐term neurological complications between groups. Subsequently, a multicountry trial enrolled over 1000 children 2 months to 12 years of age with meningitis caused by Haemophilus influenzae type b, Streptococcus pneumonia, or Neisseria meningititis who were stable after 5 days of IV ceftriaxone therapy and randomized them to receive placebo or an additional 5 days of ceftriaxone.[13] Patients with persistence of seizures, bacteremia, abscess or distant infections, or who were judged to be deteriorating or still severely ill at the 5‐day point were excluded (4.7% of the children who were recruited on day 0). There were no significant differences in bacteriologic failures, clinical failures, or clinical sequelae in survivors. The authors concluded that ceftriaxone can be discontinued in children with bacterial meningitis who are clinically stable after 5 days of IV therapy. Further trials in developed countries are needed.

BACTEREMIA

Because of routine vaccination against H influenzae type b and S pneumoniae, bacteremia beyond the first few months of life in otherwise healthy children is now rare.[15] Even in infants too young to benefit directly from vaccination, the epidemiology of bacteremia has changed considerably over the last few decades, with E coli and GBS constituting the majority (65%77%) of cases.[16, 17] We will limit this review on bacteremia to these 2 organisms in young infants.

Most cases of E coli bacteremia are associated with UTI (91%98%),[16, 17] and most bacteremic UTIs (88%92%) are caused by E coli.[18, 19, 20, 21] There are no official recommendations for the duration of treatment of bacteremic UTI, and only a limited amount of evidence can be gleaned from existing studies. In a trial of oral cefixime for infants aged 1 to 24 months with UTI, all 13 infants with bacteremia fared well whether they received oral cefixime only or IV cefotaxime for 3 days followed by oral cefixime.[18] In a study on length of IV antibiotic therapy in over 12,000 infants <6 months old with UTI, the presence of bacteremia predicted longer IV treatment length (bacteremia was present in 0.5% of the short IV group vs 0.8% of the long IV group, P=0.02) but did not predict treatment failure, defined as readmission within 30 days.[3] In a multicenter investigation of 229 infants <3 months old with bacteremic UTI, the duration of parenteral antibiotics was extremely variable (range, 117 days) and was not associated with treatment failure, defined as recurrent UTI caused by the same organism within 30 days (mean duration 7.8 days in the treatment‐failure group vs 7.7 days in the no‐failure group, P=0.99).[21] In summary, there is no evidence to support a prolonged course (ie, >35 days) of IV antibiotics for bacteremic UTI.

For bacteremia caused by GBS, although the Red Book Committee on Infectious Disease recommends 10 days of IV antibiotics,[22] to our knowledge there are no experimental or observational investigations to support this recommendation. Although available studies suggest that IV courses of at least 10 days are generally provided,[7, 23] no studies have compared outcomes of infants treated with short versus long courses. However, in a study that included 29 full‐term neonates with GBS bacteremia, all 29 had responded initially to 48 hours of intravenous antibiotics (defined as being asymptomatic and fed enterally), and were then treated successfully with high‐dose oral amoxicillin for the remainder of the course, with no recurrences.[8] Although recurrences are estimated to occur in 0.5% to 3% of babies treated for GBS infections, many recurrences are associated with exposure factors such as GBS colonization of the breast milk.[4, 5, 6, 7] In summary, although 10 or more days of IV antibiotic therapy remains a common published recommendation, there is no supportive evidence. More research is needed to assess whether shorter IV courses are safe.

UTI

Most UTIs can be treated with oral antibiotics.[24] In its practice parameter on febrile UTIs in infants 2 months to 2 years of age, the AAP recommends oral antibiotics for well‐appearing children.[25] This recommendation is supported by a recent Cochrane review on the topic,[26] and at least 3 additional trials that have demonstrated that long IV courses do not yield better outcomes than shorter IV courses or oral only courses.[27, 28, 29]

However, all of these trials exclude infants <1 month old, and there are no published recommendations for the <2‐month‐old age group. The study by Brady et al. on >12,000 infants <6 months old with UTI demonstrated no significant differences in UTI readmission rates between infants who were given 4 days of IV antibiotics versus those who were given <4 days.[3] There were 3,383 infants <30 days old in this study, and about one‐third of these babies received a short IV course. Failure rates were nearly identical in each group (2.3% in short course vs 2.4% in long course) even after risk adjustment (personal communication with Patrick Brady, MD, on February 7, 2014). Magin et al. describe 172 neonates (median age 19 days) with UTI who were treated intravenously for a median duration of 4 days (interquartile range, 36 days) and did not experience treatment failures or relapses.[30]

In summary, most cases of UTI can be managed with oral antibiotics. Uncertainty remains over the optimal approach for infants <1 to 2 months old, an age range not considered in current published guidelines. Current evidence suggests that IV treatment for 3 to 4 days followed by oral therapy may be sufficient treatment in this age group.

ACUTE OSTEOMYELITIS

Given the excellent blood supply to rapidly growing tissues in children, shorter durations of IV therapy have been studied with increasing frequency. A 2002 systematic review included 12 prospective cohort studies with at least 6 months of follow‐up.[31] Studies were stratified into 7 days or >7 days IV therapy, and there were no differences in cure rates. Subsequently, a large Finnish trial reported on 131 children who received an initial short IV course (24 days) followed by 20 versus 30 total days of therapy with very low treatment failure rates.[32]

The largest study from the United States to date analyzed nearly 2000 cases of osteomyelitis from 29 hospitals.[33] This study defined a prolonged IV course by placement of a central venous catheter. The rates of prolonged IV therapy varied significantly across hospitals, ranging from 10% to 95% of patients, without detectable differences in outcomes. Furthermore, the readmission rate for catheter related complications (3%) was nearly as high as the overall treatment failure rate (4%5%). Recently, Arnold et al. reported 8 years' experience with a management algorithm to guide the transition to oral antibiotics in pediatric osteoarticular infections in a patient specific manner.[34] This study included 194 patients (154 uncomplicated and 40 complicated cases), all with culture‐proven disease. Transition to oral antibiotics occurred based on resolution of fever and pain, improved function of the affected region, and a C‐reactive protein level of <3 mg/dL, and occurred at an average of 10 days into the treatment course. These authors also provided extensive information about complications to demonstrate that the proposed strategy can be used with a wide range of patients and pathogens. There was a single microbiologic treatment failure after oral step‐down therapy in a complicated osteoarticular infection, with a retained bony fragment. This study represents a successful example of a patient‐centered approach to IV antibiotic duration.

A PATIENT‐CENTERED APPROACH

Returning to the example above of the 2‐week‐old with UTI (with or without bacteremia), there are no published guidelines and only limited available evidence to help guide the duration of IV antibiotics in this case. When standards of care (eg, from published guidelines, review articles, textbooks, or local expert guidance) are nonexistent, conflicting, dated, or contrary to existing evidence, patient‐level factors can be incorporated into the decision‐making process (Table 2). In these cases, tailoring the IV antibiotic course to the individual's response (referred to in 1 review as the ultimate bioassay of the therapy[2]), while also weighing risks and benefits of ongoing therapy, is a logical approach.

SEVERITY OF INITIAL INFECTION AND RESPONSE TO THERAPY

The severity of the initial infection, whether in terms of presentation or clinical recovery, can factor into the duration of therapy. Provision of a longer IV course to prevent (albeit theoretically) a recurrence makes more logical sense in an infant with GBS bacteremia who was ill enough to warrant intensive care unit admission than in an infant whose only symptom was a fever. Similarly, most practitioners would be reluctant to stop IV antibiotics and discharge a patient with a bacterial infection who is persistently febrile or vomiting. Although the use of inflammatory markers and other clinical symptoms to guide therapy has been limited to osteomyelitis, this approach might be useful and should be studied in other conditions.

SHARED DECISION MAKING

Shared decision making can also be employed. Parents of sick, hospitalized children generally prefer to be involved in the decision‐making process.[35] For a parent who has concerns about their child's well‐being in the hospital, or has multiple other children at home, competing career obligations, and/or limited family support, the burden of ongoing hospitalization can be significant, and should be factored into decision making. Involving parents in medical decisions may lead to a reduction in utilization for some conditions.[36]

ASSESSMENT OF RISKS/COSTS

The risks and costs of pediatric hospitalization and prolonged IV antibiotics are well described in the literature and are summarized in Table 3. Although the benefits of prolonging IV antibiotics in a child who has recovered from an acute bacterial infection are largely theoretical, many of the risks are concrete and quantifiable. For example, a young infant being treated for a bacteremic UTI may run out of potential IV sites and need a PICC line to continue IV therapy, which according to a recent review of 2574 PICC lines has a 21% complication rate. This rate is even higher in children for whom the PICC line indication was provision of antibiotics (27%) and for infants <1 year of age (44%).[37] Moreover, this procedure often requires sedation or anesthesia for placement, which has both known and unknown risks, including concerns about subsequent adverse effects on development in young children.[38] Nosocomial exposure to seasonal viruses poses an additional risk to hospitalized children.[39]

These additional considerations for the duration of IV antibiotics are not evidence based and should not be used to justify an IV duration that differs dramatically from an accepted standard of care. These are merely considerations that incorporate clinical judgment and a comprehensive analysis of risks and benefits in situations where the available evidence is suboptimal. This approach can be adopted both as a framework for future research and directly in clinical practice.

CONCLUSION

In an era of increasing focus on overtreatment/waste,[40] patient safety,[41] and patient‐centered care,[42] the duration of IV antibiotics for common bacterial infections is a prime target for improving pediatric healthcare value. As emphasized by Michael Porter recently in The New England Journal of Medicine, value should always be defined around the customer.[43] A high‐value approach to IV antibiotic duration incorporates a rigorous assessment of risks and benefits that focuses on best evidence and patient‐level factors.

In discussing published guidelines in a review on bacterial meningitis therapy, Michael Radetsky noted that [R]ecommended criteria, even if provisional, may inadvertently become invested with an independent power to force submission and prohibit deviation. The danger is that sensitivity to individual responsiveness and variability will be lost.[2] Guidelines are useful tools in pediatrics and should continue to be used to direct IV antibiotic durations for bacterial infections in children. However, the emphasis on fixed durations of IV antibiotics might not always serve the best interest of the patient. When guidelines are lacking or contradictory, patient factors should also be considered.

Rationally defining the appropriate duration of intravenous (IV) antibiotics for children with bacterial infections is challenging. For example, how long should a 2‐week‐old infant with a urinary tract infection (UTI) caused by Escherichia coli (E coli) be treated intravenously if the infant has responded to treatment and is back to baseline within 1 to 2 days? What if the blood culture was also positive for E coli? What are the risks and benefits of continuing IV antibiotics?

Such questions are common for pediatric hospitalists. Bacterial infections remain a relatively frequent cause of pediatric hospitalization, especially in neonates where 5 of the top 10 causes of hospitalizations are related to bacterial infections.[1] For some conditions, children remain hospitalized after clinical improvement simply for ongoing provision of IV antibiotics. Alternatively, some children are discharged home with a peripherally inserted central catheter (PICC) to complete an IV course.

The decision regarding the duration of IV antibiotics varies according to the condition for which the antibiotic is prescribed and often by practitioner or hospital. Many recommendations are numerically based (eg, 10 days for group B Streptococcus [GBS] bacteremia, 21 days for E coli meningitis), without taking into account patient‐level factors such as initial severity or response to therapy. These concrete recommendations may in fact be preferred by some practitioners, as suggested by a former chairman of the Committee on Infectious Disease for the American Academy of Pediatrics (AAP): The Red Book is designed for people who make decisions. It cannot waffle on an issue. It has to make a positive recommendation even if the data are incomplete.[2] A potential downside of this mentality, however, is that some practitioners may then feel obligated to follow these recommendations despite the lack of supportive evidence.

EXTENDING IV ANTIBIOTICS BEYOND CLINICAL RECOVERY

What is the rationale for continuing IV antibiotics in infants whose symptoms have completely resolved? Several factors likely drive these decisions: prevention of recurrences, concerns about bioavailability of enteral antibiotics and patient compliance, adherence to expert recommendations/guidelines, and perhaps a general sense that more is betterthat serious infections and/or their sequelae require more aggressive treatments.

Recurrence of a potentially life‐threatening infection is an understandable concern. Even when symptoms have resolved and there is documented clearance of the infection, such clearance does not necessarily signify that the body has rid itself of the pathogen completely. Some infections are deep seated and may warrant continuing treatment despite apparent recovery. To some, the risks of prolonging IV therapy may seem inconsequential when juxtaposed to a potentially devastating recurrence. However, in many conditions, recurrences may be related to host issues or ongoing exposures rather than inadequate treatment of the original infection. Recurrent UTIs, for example, are more likely in infants with urologic abnormalities,[3] and recurrent GBS bacteremia has been associated with GBS colonization of maternal breast milk and/or maternal mastitis.[4, 5, 6, 7] Although it is tempting to extend IV courses to prevent recurrences, it is not clear that the benefits of such an approach outweigh the risks.

Concerns over enteral absorption and bioavailability are also understandable, especially in young infants. The superior efficacy of IV over oral antibiotics in general is well accepted for many pediatric conditions, and in some cases (eg, septic shock) it would be unethical to perform a head‐to‐head trial. However, the lack of any published trials (to our knowledge) in pediatrics confirming the superiority of IV antibiotics suggests that oral antibiotic absorption is sufficient for many infections. Even in neonates, several studies have demonstrated that therapeutic serum levels are easily reached with oral dosing of amoxicillin in term and preterm neonates.[8, 9]

For the remainder of this review, the published recommendations and available evidence behind the duration of IV therapy are summarized for 4 bacterial infections in children in which IV antibiotic therapy often continues after clinical recovery: meningitis, bacteremia, UTI, and acute osteomyelitis. We conclude by proposing additional considerations for IV antibiotic durations, especially in situations where guidelines and/or evidence are either nonexistent, dated, conflicting, or contrary to evidence from published studies.

BACTERIAL MENINGITIS

The Infectious Disease Society of America and the British National Institute for Clinical Evidence have both published guidelines with pediatric recommendations for duration of therapy in bacterial meningitis,[10, 11] though the recommendations differ somewhat for 3 of the 4 most common pathogens, and are not always concordant with evidence from randomized controlled trials (Table 1).[12, 13, 14]

A recent meta‐analysis on duration of therapy in meningitis included 5 open‐label trials of ceftriaxone for bacterial meningitis in children.[12] These trials included the 3 most common pathogens and were categorized as short‐course (47 days, n=196 patients) and long‐course (714 days, n=187 patients) therapy. There was no significant difference in clinical success or long‐term neurological complications between groups. Subsequently, a multicountry trial enrolled over 1000 children 2 months to 12 years of age with meningitis caused by Haemophilus influenzae type b, Streptococcus pneumonia, or Neisseria meningititis who were stable after 5 days of IV ceftriaxone therapy and randomized them to receive placebo or an additional 5 days of ceftriaxone.[13] Patients with persistence of seizures, bacteremia, abscess or distant infections, or who were judged to be deteriorating or still severely ill at the 5‐day point were excluded (4.7% of the children who were recruited on day 0). There were no significant differences in bacteriologic failures, clinical failures, or clinical sequelae in survivors. The authors concluded that ceftriaxone can be discontinued in children with bacterial meningitis who are clinically stable after 5 days of IV therapy. Further trials in developed countries are needed.

BACTEREMIA

Because of routine vaccination against H influenzae type b and S pneumoniae, bacteremia beyond the first few months of life in otherwise healthy children is now rare.[15] Even in infants too young to benefit directly from vaccination, the epidemiology of bacteremia has changed considerably over the last few decades, with E coli and GBS constituting the majority (65%77%) of cases.[16, 17] We will limit this review on bacteremia to these 2 organisms in young infants.

Most cases of E coli bacteremia are associated with UTI (91%98%),[16, 17] and most bacteremic UTIs (88%92%) are caused by E coli.[18, 19, 20, 21] There are no official recommendations for the duration of treatment of bacteremic UTI, and only a limited amount of evidence can be gleaned from existing studies. In a trial of oral cefixime for infants aged 1 to 24 months with UTI, all 13 infants with bacteremia fared well whether they received oral cefixime only or IV cefotaxime for 3 days followed by oral cefixime.[18] In a study on length of IV antibiotic therapy in over 12,000 infants <6 months old with UTI, the presence of bacteremia predicted longer IV treatment length (bacteremia was present in 0.5% of the short IV group vs 0.8% of the long IV group, P=0.02) but did not predict treatment failure, defined as readmission within 30 days.[3] In a multicenter investigation of 229 infants <3 months old with bacteremic UTI, the duration of parenteral antibiotics was extremely variable (range, 117 days) and was not associated with treatment failure, defined as recurrent UTI caused by the same organism within 30 days (mean duration 7.8 days in the treatment‐failure group vs 7.7 days in the no‐failure group, P=0.99).[21] In summary, there is no evidence to support a prolonged course (ie, >35 days) of IV antibiotics for bacteremic UTI.

For bacteremia caused by GBS, although the Red Book Committee on Infectious Disease recommends 10 days of IV antibiotics,[22] to our knowledge there are no experimental or observational investigations to support this recommendation. Although available studies suggest that IV courses of at least 10 days are generally provided,[7, 23] no studies have compared outcomes of infants treated with short versus long courses. However, in a study that included 29 full‐term neonates with GBS bacteremia, all 29 had responded initially to 48 hours of intravenous antibiotics (defined as being asymptomatic and fed enterally), and were then treated successfully with high‐dose oral amoxicillin for the remainder of the course, with no recurrences.[8] Although recurrences are estimated to occur in 0.5% to 3% of babies treated for GBS infections, many recurrences are associated with exposure factors such as GBS colonization of the breast milk.[4, 5, 6, 7] In summary, although 10 or more days of IV antibiotic therapy remains a common published recommendation, there is no supportive evidence. More research is needed to assess whether shorter IV courses are safe.

UTI

Most UTIs can be treated with oral antibiotics.[24] In its practice parameter on febrile UTIs in infants 2 months to 2 years of age, the AAP recommends oral antibiotics for well‐appearing children.[25] This recommendation is supported by a recent Cochrane review on the topic,[26] and at least 3 additional trials that have demonstrated that long IV courses do not yield better outcomes than shorter IV courses or oral only courses.[27, 28, 29]

However, all of these trials exclude infants <1 month old, and there are no published recommendations for the <2‐month‐old age group. The study by Brady et al. on >12,000 infants <6 months old with UTI demonstrated no significant differences in UTI readmission rates between infants who were given 4 days of IV antibiotics versus those who were given <4 days.[3] There were 3,383 infants <30 days old in this study, and about one‐third of these babies received a short IV course. Failure rates were nearly identical in each group (2.3% in short course vs 2.4% in long course) even after risk adjustment (personal communication with Patrick Brady, MD, on February 7, 2014). Magin et al. describe 172 neonates (median age 19 days) with UTI who were treated intravenously for a median duration of 4 days (interquartile range, 36 days) and did not experience treatment failures or relapses.[30]

In summary, most cases of UTI can be managed with oral antibiotics. Uncertainty remains over the optimal approach for infants <1 to 2 months old, an age range not considered in current published guidelines. Current evidence suggests that IV treatment for 3 to 4 days followed by oral therapy may be sufficient treatment in this age group.

ACUTE OSTEOMYELITIS

Given the excellent blood supply to rapidly growing tissues in children, shorter durations of IV therapy have been studied with increasing frequency. A 2002 systematic review included 12 prospective cohort studies with at least 6 months of follow‐up.[31] Studies were stratified into 7 days or >7 days IV therapy, and there were no differences in cure rates. Subsequently, a large Finnish trial reported on 131 children who received an initial short IV course (24 days) followed by 20 versus 30 total days of therapy with very low treatment failure rates.[32]

The largest study from the United States to date analyzed nearly 2000 cases of osteomyelitis from 29 hospitals.[33] This study defined a prolonged IV course by placement of a central venous catheter. The rates of prolonged IV therapy varied significantly across hospitals, ranging from 10% to 95% of patients, without detectable differences in outcomes. Furthermore, the readmission rate for catheter related complications (3%) was nearly as high as the overall treatment failure rate (4%5%). Recently, Arnold et al. reported 8 years' experience with a management algorithm to guide the transition to oral antibiotics in pediatric osteoarticular infections in a patient specific manner.[34] This study included 194 patients (154 uncomplicated and 40 complicated cases), all with culture‐proven disease. Transition to oral antibiotics occurred based on resolution of fever and pain, improved function of the affected region, and a C‐reactive protein level of <3 mg/dL, and occurred at an average of 10 days into the treatment course. These authors also provided extensive information about complications to demonstrate that the proposed strategy can be used with a wide range of patients and pathogens. There was a single microbiologic treatment failure after oral step‐down therapy in a complicated osteoarticular infection, with a retained bony fragment. This study represents a successful example of a patient‐centered approach to IV antibiotic duration.

A PATIENT‐CENTERED APPROACH

Returning to the example above of the 2‐week‐old with UTI (with or without bacteremia), there are no published guidelines and only limited available evidence to help guide the duration of IV antibiotics in this case. When standards of care (eg, from published guidelines, review articles, textbooks, or local expert guidance) are nonexistent, conflicting, dated, or contrary to existing evidence, patient‐level factors can be incorporated into the decision‐making process (Table 2). In these cases, tailoring the IV antibiotic course to the individual's response (referred to in 1 review as the ultimate bioassay of the therapy[2]), while also weighing risks and benefits of ongoing therapy, is a logical approach.

SEVERITY OF INITIAL INFECTION AND RESPONSE TO THERAPY

The severity of the initial infection, whether in terms of presentation or clinical recovery, can factor into the duration of therapy. Provision of a longer IV course to prevent (albeit theoretically) a recurrence makes more logical sense in an infant with GBS bacteremia who was ill enough to warrant intensive care unit admission than in an infant whose only symptom was a fever. Similarly, most practitioners would be reluctant to stop IV antibiotics and discharge a patient with a bacterial infection who is persistently febrile or vomiting. Although the use of inflammatory markers and other clinical symptoms to guide therapy has been limited to osteomyelitis, this approach might be useful and should be studied in other conditions.

SHARED DECISION MAKING

Shared decision making can also be employed. Parents of sick, hospitalized children generally prefer to be involved in the decision‐making process.[35] For a parent who has concerns about their child's well‐being in the hospital, or has multiple other children at home, competing career obligations, and/or limited family support, the burden of ongoing hospitalization can be significant, and should be factored into decision making. Involving parents in medical decisions may lead to a reduction in utilization for some conditions.[36]

ASSESSMENT OF RISKS/COSTS

The risks and costs of pediatric hospitalization and prolonged IV antibiotics are well described in the literature and are summarized in Table 3. Although the benefits of prolonging IV antibiotics in a child who has recovered from an acute bacterial infection are largely theoretical, many of the risks are concrete and quantifiable. For example, a young infant being treated for a bacteremic UTI may run out of potential IV sites and need a PICC line to continue IV therapy, which according to a recent review of 2574 PICC lines has a 21% complication rate. This rate is even higher in children for whom the PICC line indication was provision of antibiotics (27%) and for infants <1 year of age (44%).[37] Moreover, this procedure often requires sedation or anesthesia for placement, which has both known and unknown risks, including concerns about subsequent adverse effects on development in young children.[38] Nosocomial exposure to seasonal viruses poses an additional risk to hospitalized children.[39]

These additional considerations for the duration of IV antibiotics are not evidence based and should not be used to justify an IV duration that differs dramatically from an accepted standard of care. These are merely considerations that incorporate clinical judgment and a comprehensive analysis of risks and benefits in situations where the available evidence is suboptimal. This approach can be adopted both as a framework for future research and directly in clinical practice.

CONCLUSION

In an era of increasing focus on overtreatment/waste,[40] patient safety,[41] and patient‐centered care,[42] the duration of IV antibiotics for common bacterial infections is a prime target for improving pediatric healthcare value. As emphasized by Michael Porter recently in The New England Journal of Medicine, value should always be defined around the customer.[43] A high‐value approach to IV antibiotic duration incorporates a rigorous assessment of risks and benefits that focuses on best evidence and patient‐level factors.

In discussing published guidelines in a review on bacterial meningitis therapy, Michael Radetsky noted that [R]ecommended criteria, even if provisional, may inadvertently become invested with an independent power to force submission and prohibit deviation. The danger is that sensitivity to individual responsiveness and variability will be lost.[2] Guidelines are useful tools in pediatrics and should continue to be used to direct IV antibiotic durations for bacterial infections in children. However, the emphasis on fixed durations of IV antibiotics might not always serve the best interest of the patient. When guidelines are lacking or contradictory, patient factors should also be considered.

Rationally defining the appropriate duration of intravenous (IV) antibiotics for children with bacterial infections is challenging. For example, how long should a 2‐week‐old infant with a urinary tract infection (UTI) caused by Escherichia coli (E coli) be treated intravenously if the infant has responded to treatment and is back to baseline within 1 to 2 days? What if the blood culture was also positive for E coli? What are the risks and benefits of continuing IV antibiotics?

Such questions are common for pediatric hospitalists. Bacterial infections remain a relatively frequent cause of pediatric hospitalization, especially in neonates where 5 of the top 10 causes of hospitalizations are related to bacterial infections.[1] For some conditions, children remain hospitalized after clinical improvement simply for ongoing provision of IV antibiotics. Alternatively, some children are discharged home with a peripherally inserted central catheter (PICC) to complete an IV course.

The decision regarding the duration of IV antibiotics varies according to the condition for which the antibiotic is prescribed and often by practitioner or hospital. Many recommendations are numerically based (eg, 10 days for group B Streptococcus [GBS] bacteremia, 21 days for E coli meningitis), without taking into account patient‐level factors such as initial severity or response to therapy. These concrete recommendations may in fact be preferred by some practitioners, as suggested by a former chairman of the Committee on Infectious Disease for the American Academy of Pediatrics (AAP): The Red Book is designed for people who make decisions. It cannot waffle on an issue. It has to make a positive recommendation even if the data are incomplete.[2] A potential downside of this mentality, however, is that some practitioners may then feel obligated to follow these recommendations despite the lack of supportive evidence.

EXTENDING IV ANTIBIOTICS BEYOND CLINICAL RECOVERY

What is the rationale for continuing IV antibiotics in infants whose symptoms have completely resolved? Several factors likely drive these decisions: prevention of recurrences, concerns about bioavailability of enteral antibiotics and patient compliance, adherence to expert recommendations/guidelines, and perhaps a general sense that more is betterthat serious infections and/or their sequelae require more aggressive treatments.

Recurrence of a potentially life‐threatening infection is an understandable concern. Even when symptoms have resolved and there is documented clearance of the infection, such clearance does not necessarily signify that the body has rid itself of the pathogen completely. Some infections are deep seated and may warrant continuing treatment despite apparent recovery. To some, the risks of prolonging IV therapy may seem inconsequential when juxtaposed to a potentially devastating recurrence. However, in many conditions, recurrences may be related to host issues or ongoing exposures rather than inadequate treatment of the original infection. Recurrent UTIs, for example, are more likely in infants with urologic abnormalities,[3] and recurrent GBS bacteremia has been associated with GBS colonization of maternal breast milk and/or maternal mastitis.[4, 5, 6, 7] Although it is tempting to extend IV courses to prevent recurrences, it is not clear that the benefits of such an approach outweigh the risks.

Concerns over enteral absorption and bioavailability are also understandable, especially in young infants. The superior efficacy of IV over oral antibiotics in general is well accepted for many pediatric conditions, and in some cases (eg, septic shock) it would be unethical to perform a head‐to‐head trial. However, the lack of any published trials (to our knowledge) in pediatrics confirming the superiority of IV antibiotics suggests that oral antibiotic absorption is sufficient for many infections. Even in neonates, several studies have demonstrated that therapeutic serum levels are easily reached with oral dosing of amoxicillin in term and preterm neonates.[8, 9]

For the remainder of this review, the published recommendations and available evidence behind the duration of IV therapy are summarized for 4 bacterial infections in children in which IV antibiotic therapy often continues after clinical recovery: meningitis, bacteremia, UTI, and acute osteomyelitis. We conclude by proposing additional considerations for IV antibiotic durations, especially in situations where guidelines and/or evidence are either nonexistent, dated, conflicting, or contrary to evidence from published studies.

BACTERIAL MENINGITIS

The Infectious Disease Society of America and the British National Institute for Clinical Evidence have both published guidelines with pediatric recommendations for duration of therapy in bacterial meningitis,[10, 11] though the recommendations differ somewhat for 3 of the 4 most common pathogens, and are not always concordant with evidence from randomized controlled trials (Table 1).[12, 13, 14]

A recent meta‐analysis on duration of therapy in meningitis included 5 open‐label trials of ceftriaxone for bacterial meningitis in children.[12] These trials included the 3 most common pathogens and were categorized as short‐course (47 days, n=196 patients) and long‐course (714 days, n=187 patients) therapy. There was no significant difference in clinical success or long‐term neurological complications between groups. Subsequently, a multicountry trial enrolled over 1000 children 2 months to 12 years of age with meningitis caused by Haemophilus influenzae type b, Streptococcus pneumonia, or Neisseria meningititis who were stable after 5 days of IV ceftriaxone therapy and randomized them to receive placebo or an additional 5 days of ceftriaxone.[13] Patients with persistence of seizures, bacteremia, abscess or distant infections, or who were judged to be deteriorating or still severely ill at the 5‐day point were excluded (4.7% of the children who were recruited on day 0). There were no significant differences in bacteriologic failures, clinical failures, or clinical sequelae in survivors. The authors concluded that ceftriaxone can be discontinued in children with bacterial meningitis who are clinically stable after 5 days of IV therapy. Further trials in developed countries are needed.

BACTEREMIA

Because of routine vaccination against H influenzae type b and S pneumoniae, bacteremia beyond the first few months of life in otherwise healthy children is now rare.[15] Even in infants too young to benefit directly from vaccination, the epidemiology of bacteremia has changed considerably over the last few decades, with E coli and GBS constituting the majority (65%77%) of cases.[16, 17] We will limit this review on bacteremia to these 2 organisms in young infants.

Most cases of E coli bacteremia are associated with UTI (91%98%),[16, 17] and most bacteremic UTIs (88%92%) are caused by E coli.[18, 19, 20, 21] There are no official recommendations for the duration of treatment of bacteremic UTI, and only a limited amount of evidence can be gleaned from existing studies. In a trial of oral cefixime for infants aged 1 to 24 months with UTI, all 13 infants with bacteremia fared well whether they received oral cefixime only or IV cefotaxime for 3 days followed by oral cefixime.[18] In a study on length of IV antibiotic therapy in over 12,000 infants <6 months old with UTI, the presence of bacteremia predicted longer IV treatment length (bacteremia was present in 0.5% of the short IV group vs 0.8% of the long IV group, P=0.02) but did not predict treatment failure, defined as readmission within 30 days.[3] In a multicenter investigation of 229 infants <3 months old with bacteremic UTI, the duration of parenteral antibiotics was extremely variable (range, 117 days) and was not associated with treatment failure, defined as recurrent UTI caused by the same organism within 30 days (mean duration 7.8 days in the treatment‐failure group vs 7.7 days in the no‐failure group, P=0.99).[21] In summary, there is no evidence to support a prolonged course (ie, >35 days) of IV antibiotics for bacteremic UTI.

For bacteremia caused by GBS, although the Red Book Committee on Infectious Disease recommends 10 days of IV antibiotics,[22] to our knowledge there are no experimental or observational investigations to support this recommendation. Although available studies suggest that IV courses of at least 10 days are generally provided,[7, 23] no studies have compared outcomes of infants treated with short versus long courses. However, in a study that included 29 full‐term neonates with GBS bacteremia, all 29 had responded initially to 48 hours of intravenous antibiotics (defined as being asymptomatic and fed enterally), and were then treated successfully with high‐dose oral amoxicillin for the remainder of the course, with no recurrences.[8] Although recurrences are estimated to occur in 0.5% to 3% of babies treated for GBS infections, many recurrences are associated with exposure factors such as GBS colonization of the breast milk.[4, 5, 6, 7] In summary, although 10 or more days of IV antibiotic therapy remains a common published recommendation, there is no supportive evidence. More research is needed to assess whether shorter IV courses are safe.

UTI

Most UTIs can be treated with oral antibiotics.[24] In its practice parameter on febrile UTIs in infants 2 months to 2 years of age, the AAP recommends oral antibiotics for well‐appearing children.[25] This recommendation is supported by a recent Cochrane review on the topic,[26] and at least 3 additional trials that have demonstrated that long IV courses do not yield better outcomes than shorter IV courses or oral only courses.[27, 28, 29]

However, all of these trials exclude infants <1 month old, and there are no published recommendations for the <2‐month‐old age group. The study by Brady et al. on >12,000 infants <6 months old with UTI demonstrated no significant differences in UTI readmission rates between infants who were given 4 days of IV antibiotics versus those who were given <4 days.[3] There were 3,383 infants <30 days old in this study, and about one‐third of these babies received a short IV course. Failure rates were nearly identical in each group (2.3% in short course vs 2.4% in long course) even after risk adjustment (personal communication with Patrick Brady, MD, on February 7, 2014). Magin et al. describe 172 neonates (median age 19 days) with UTI who were treated intravenously for a median duration of 4 days (interquartile range, 36 days) and did not experience treatment failures or relapses.[30]

In summary, most cases of UTI can be managed with oral antibiotics. Uncertainty remains over the optimal approach for infants <1 to 2 months old, an age range not considered in current published guidelines. Current evidence suggests that IV treatment for 3 to 4 days followed by oral therapy may be sufficient treatment in this age group.

ACUTE OSTEOMYELITIS

Given the excellent blood supply to rapidly growing tissues in children, shorter durations of IV therapy have been studied with increasing frequency. A 2002 systematic review included 12 prospective cohort studies with at least 6 months of follow‐up.[31] Studies were stratified into 7 days or >7 days IV therapy, and there were no differences in cure rates. Subsequently, a large Finnish trial reported on 131 children who received an initial short IV course (24 days) followed by 20 versus 30 total days of therapy with very low treatment failure rates.[32]

The largest study from the United States to date analyzed nearly 2000 cases of osteomyelitis from 29 hospitals.[33] This study defined a prolonged IV course by placement of a central venous catheter. The rates of prolonged IV therapy varied significantly across hospitals, ranging from 10% to 95% of patients, without detectable differences in outcomes. Furthermore, the readmission rate for catheter related complications (3%) was nearly as high as the overall treatment failure rate (4%5%). Recently, Arnold et al. reported 8 years' experience with a management algorithm to guide the transition to oral antibiotics in pediatric osteoarticular infections in a patient specific manner.[34] This study included 194 patients (154 uncomplicated and 40 complicated cases), all with culture‐proven disease. Transition to oral antibiotics occurred based on resolution of fever and pain, improved function of the affected region, and a C‐reactive protein level of <3 mg/dL, and occurred at an average of 10 days into the treatment course. These authors also provided extensive information about complications to demonstrate that the proposed strategy can be used with a wide range of patients and pathogens. There was a single microbiologic treatment failure after oral step‐down therapy in a complicated osteoarticular infection, with a retained bony fragment. This study represents a successful example of a patient‐centered approach to IV antibiotic duration.

A PATIENT‐CENTERED APPROACH

Returning to the example above of the 2‐week‐old with UTI (with or without bacteremia), there are no published guidelines and only limited available evidence to help guide the duration of IV antibiotics in this case. When standards of care (eg, from published guidelines, review articles, textbooks, or local expert guidance) are nonexistent, conflicting, dated, or contrary to existing evidence, patient‐level factors can be incorporated into the decision‐making process (Table 2). In these cases, tailoring the IV antibiotic course to the individual's response (referred to in 1 review as the ultimate bioassay of the therapy[2]), while also weighing risks and benefits of ongoing therapy, is a logical approach.

SEVERITY OF INITIAL INFECTION AND RESPONSE TO THERAPY

The severity of the initial infection, whether in terms of presentation or clinical recovery, can factor into the duration of therapy. Provision of a longer IV course to prevent (albeit theoretically) a recurrence makes more logical sense in an infant with GBS bacteremia who was ill enough to warrant intensive care unit admission than in an infant whose only symptom was a fever. Similarly, most practitioners would be reluctant to stop IV antibiotics and discharge a patient with a bacterial infection who is persistently febrile or vomiting. Although the use of inflammatory markers and other clinical symptoms to guide therapy has been limited to osteomyelitis, this approach might be useful and should be studied in other conditions.

SHARED DECISION MAKING

Shared decision making can also be employed. Parents of sick, hospitalized children generally prefer to be involved in the decision‐making process.[35] For a parent who has concerns about their child's well‐being in the hospital, or has multiple other children at home, competing career obligations, and/or limited family support, the burden of ongoing hospitalization can be significant, and should be factored into decision making. Involving parents in medical decisions may lead to a reduction in utilization for some conditions.[36]

ASSESSMENT OF RISKS/COSTS

The risks and costs of pediatric hospitalization and prolonged IV antibiotics are well described in the literature and are summarized in Table 3. Although the benefits of prolonging IV antibiotics in a child who has recovered from an acute bacterial infection are largely theoretical, many of the risks are concrete and quantifiable. For example, a young infant being treated for a bacteremic UTI may run out of potential IV sites and need a PICC line to continue IV therapy, which according to a recent review of 2574 PICC lines has a 21% complication rate. This rate is even higher in children for whom the PICC line indication was provision of antibiotics (27%) and for infants <1 year of age (44%).[37] Moreover, this procedure often requires sedation or anesthesia for placement, which has both known and unknown risks, including concerns about subsequent adverse effects on development in young children.[38] Nosocomial exposure to seasonal viruses poses an additional risk to hospitalized children.[39]

These additional considerations for the duration of IV antibiotics are not evidence based and should not be used to justify an IV duration that differs dramatically from an accepted standard of care. These are merely considerations that incorporate clinical judgment and a comprehensive analysis of risks and benefits in situations where the available evidence is suboptimal. This approach can be adopted both as a framework for future research and directly in clinical practice.

CONCLUSION

In an era of increasing focus on overtreatment/waste,[40] patient safety,[41] and patient‐centered care,[42] the duration of IV antibiotics for common bacterial infections is a prime target for improving pediatric healthcare value. As emphasized by Michael Porter recently in The New England Journal of Medicine, value should always be defined around the customer.[43] A high‐value approach to IV antibiotic duration incorporates a rigorous assessment of risks and benefits that focuses on best evidence and patient‐level factors.

In discussing published guidelines in a review on bacterial meningitis therapy, Michael Radetsky noted that [R]ecommended criteria, even if provisional, may inadvertently become invested with an independent power to force submission and prohibit deviation. The danger is that sensitivity to individual responsiveness and variability will be lost.[2] Guidelines are useful tools in pediatrics and should continue to be used to direct IV antibiotic durations for bacterial infections in children. However, the emphasis on fixed durations of IV antibiotics might not always serve the best interest of the patient. When guidelines are lacking or contradictory, patient factors should also be considered.

Each year in the United States, over 1 million adults undergo hip fracture surgery or elective total knee or hip arthroplasty.[1] Although highly effective for improving functional status and quality of life,[2, 3] each of these procedures is associated with a substantial risk of developing a deep vein thrombosis (DVT) or pulmonary embolism (PE).[4, 5] Collectively referred to as venous thromboembolism (VTE), these clots in the venous system are associated with significant morbidity and mortality for patients, as well as substantial costs to the healthcare system.[6] Although VTE is considered to be a preventable cause of hospital admission and death,[7, 8] the postoperative setting presents a particular challenge, as efforts to reduce clotting must be balanced against the risk of bleeding.

Despite how common this scenario is, there is no consensus regarding the best pharmacologic strategy. National guidelines recommend pharmacologic thromboprophylaxis, leaving the clinician to select the specific agent.[4, 5] Explicitly endorsed options include aspirin, vitamin K antagonists (VKA), unfractionated heparin, fondaparinux, low‐molecular‐weight heparin (LMWH) and IIa/Xa factor inhibitors. Among these, aspirin, the only nonanticoagulant, has been the source of greatest controversy.[4, 9, 10]

Two previous systematic reviews comparing aspirin to anticoagulation for VTE prevention found conflicting results.[11, 12] In addition, both used indirect comparisons, a method in which the intervention and comparison data come from different studies, and susceptibility to confounding is high.[13, 14] We aimed to overcome the limitations of prior efforts to address this commonly encountered clinical question by conducting a systematic review and meta‐analysis of randomized controlled trials that directly compared the efficacy and safety of aspirin to anticoagulants for VTE prevention in adults undergoing common high‐risk major orthopedic surgeries of the lower extremities.

MATERIAL AND METHODS

RESULTS

Results of Search

Figure 1 shows the number of studies that we evaluated during each stage of the study selection process. After full‐text review, 8 randomized trials met all inclusion criteria.[23, 24, 25, 26, 27, 28, 29, 30]

Figure 1

Rate of Proximal DVT

Pooling findings of all 7 studies that reported proximal DVT rates, we observed no statistically significant difference between aspirin and anticoagulants (10.4% vs 9.2%, relative risk [RR]: 1.15 [95% confidence interval {CI}: 0.68‐1.96], I²=41%). Although rates did not statistically differ between aspirin and anticoagulants in either operative subgroup, there appeared to be a nonsignificant trend favoring anticoagulation after hip fracture repair (12.7% vs 7.8%, RR: 1.60 [95% CI: 0.80‐3.20], I²=0%, 2 trials) but not following knee or hip arthroplasty (9.3% vs 9.7%, RR: 1.00 [95% CI: 0.49‐2.05], I²=49%, 5 trials) (Figure 2).

Figure 2

Rate of Pulmonary Embolism

Just 14 participants experienced a PE across all 6 trials reporting this outcome (aspirin n=9/405 versus anticoagulation n=5/415). Although PE was numerically more likely in the aspirin group, this difference was not statistically significant (overall: 1.9% vs 0.9%, RR: 1.83 [95% CI: 0.64, 5.21], I²=0%). The very small number of events rendered extremely wide 95% CIs in operative subgroup analyses (Figure 3).

Figure 3

Bleeding Rates

Pooling all 8 studies, aspirin was associated with a statistically significant 48% decreased risk of bleeding events compared to anticoagulants (3.8% vs 8.0%, RR: 0.52 [95% CI: 0.310.86], I²=8%). When subgrouped according to procedure, bleeding rates remained statistically significantly lower in the aspirin group following hip fracture (3.1% vs 10%, RR: 0.32 [95% CI: 0.130.77], I²=0%, 2 trials); however, the observed trend favoring aspirin was not statistically significant following arthroplasty (3.9% vs 7.8%, RR: 0.63 [95% CI: 0.331.21], I²=14%, 5 trials) (Figure 4).

Figure 4

Five studies reported major bleeding; event rates were low and no statistically significant differences between aspirin and anticoagulants were observed (hip fracture: 3.5% vs 6.3%, RR: 0.46 [95% CI: 0.141.48], I²=0%, 2 trials; knee/hip arthroplasty: 2.1% vs 0.6%, RR: 2.86 [95% CI: 0.6512.60], I²=0%, 3 trials).

DISCUSSION

We found the balance of risk versus benefit of aspirin compared to anticoagulation differed markedly according to type of surgery. After hip fracture repair, we found a 68% reduction in bleeding risk with aspirin compared to anticoagulants. This benefit, however, was associated with a nonsignificant increase in screen‐detected proximal DVT. Conversely, among patients undergoing knee or hip arthroplasty, we found no difference in proximal DVT risk between aspirin and anticoagulants and a possible trend toward less bleeding risk with aspirin. The rarity of pulmonary emboli (and death) made meaningful comparisons between aspirin and anticoagulation impossible for either type of surgery.

Our systematic review has several strengths that differentiate it from previous analyses. First, we only included head‐to‐head randomized trials such that all included data reflect direct comparisons between aspirin and anticoagulation in well‐balanced populations. Conversely, both recent reviews[11, 12] were based on indirect comparisons, a type of analysis in which data for the intervention and control arms are taken from different studies and thus different populations. This methodology is not recommended by the Cochrane Collaboration[13, 14] because of the increased risk of an unbalanced comparison. For example, Brown and colleagues' meta‐analysis, which pooled data from selected arms of 14 randomized controlled trials, found the efficacy of aspirin comparable to that of anticoagulants, but all aspirin subjects came from a single trial of patients at such low risk of VTE that a placebo arm was considered justified.[31] Similarly, in the indirect comparison of Westrich and colleagues,[12] which found anticoagulation superior to aspirin, the likelihood of an unbalanced comparison was further heightened by their inclusion of observational studies, with the attendant risk of confounding by indication.

Our systematic review further differs from previous analyses by examining both beneficial and harmful clinical outcomes, and doing so separately for the 2 most common types of major orthopedic lower extremity surgery. This allowed us to discover important differences in the comparative efficacy (benefit vs harm) of aspirin versus anticoagulants across different procedure types. Finding that aspirin may have lower efficacy for preventing VTE following hip fracture repair than arthroplasty may not be surprising in light of the nature of the 2 procedures, the disparate mean ages typical of patients who undergo each procedure, and the underlying trauma in hip fracture patients.

The limitations of our review largely reflect the quality of the studies we were able to include. First, our pooled sample size remains relatively small, meaning that observed nonsignificant differences between aspirin and anticoagulation groups (eg, a nonsignificant 60% increased risk of DVT for aspirin after hip repair, 95% CI: 0.803.20) could reasonably reflect up to 3‐fold differences in DVT risk and 5‐fold differences in PE rates. Second, screening for DVT, which is neither recommended nor common in clinical practice, was used in all studies. Reported DVT incidence, therefore, is undoubtedly higher than what would be observed in practice; however, the effect on the direction and magnitude of observed relative risks is unpredictable. Third, included studies used a wide range of aspirin doses, as well as a variety of anticoagulant types. Although supratherapeutic aspirin doses are unlikely to confer additional benefit for venous thromboprophylaxis, they may be associated with excess bleeding risk.[32] Finally, several of the studies were conducted more than10 years ago. Given changes in treatment practices, surgical technique, and prophylaxis options, the findings of these studies may not reflect current practice, in which early mobilization and intermittent pneumatic compression devices are standard prophylaxis against postoperative VTE. In fact, only 2 trials used concomitant pneumatic compression devices, and none treated patients longer than 21 days, the current standard being up to 35 days.[4] Although these limitations may affect overall event rates, this bias should be balanced between comparison groups, because we only included randomized controlled trials.

What is a clinician to do? Based on our findings, current guidelines recommending aspirin prophylaxis against VTE as an alternative following major lower extremity surgery may not be universally appropriate. We found that although overall bleeding complications are lower with aspirin, concerns about poor efficacy remain, specifically for patients undergoing hip fracture repair. Although some have suggested that aspirin use be restricted to low risk patients, this strategy has not been experimentally evaluated.[33] On the other hand, switching to aspirin after a brief initial course of LMWH may be an approach warranting further study, in light of a recent randomized controlled trial of 778 patients after elective hip replacement, which found equivalent efficacy using 10 days of LMWH followed by aspirin versus additional LMWH for 28 days.[34]

We are able to be more definitive, based on our study of best available trial data, in making recommendations to investigators embarking on further study of optimal VTE prophylaxis following major orthopedic surgery. First, distinguishing a priori between the 2 major types of lower extremity major orthopedic surgery is a high priority. Second, both bleeding and thromboembolic outcomes must be evaluated. Third, only symptomatic events should be used to measure VTE outcomes; clinical follow‐up must continue well beyond discharge, for at least 3 months to ensure ascertainment of clinically relevant VTE. Fourth, nonpharmacologic cointerventions should be standardized and represent the standard of care, including early immobilization and mechanical compression devices.

In summary, although definitive recommendations for or against the use of aspirin instead of anticoagulation for VTE prevention following major orthopedic surgery are not possible, our findings suggest that, following hip fracture repair, the lower risk of bleeding with aspirin is likely outweighed by a probable trend toward higher risk of VTE. On the other hand, the balance of these opposing risks may favor aspirin after elective knee or hip arthroplasty. A comparative study of aspirin, anticoagulation, and a hybrid strategy (eg, brief anticoagulation followed by aspirin) after elective knee or hip arthroplasty should be a high priority given our aging population and increasing demand for major orthopedic lower extremity surgery.

Each year in the United States, over 1 million adults undergo hip fracture surgery or elective total knee or hip arthroplasty.[1] Although highly effective for improving functional status and quality of life,[2, 3] each of these procedures is associated with a substantial risk of developing a deep vein thrombosis (DVT) or pulmonary embolism (PE).[4, 5] Collectively referred to as venous thromboembolism (VTE), these clots in the venous system are associated with significant morbidity and mortality for patients, as well as substantial costs to the healthcare system.[6] Although VTE is considered to be a preventable cause of hospital admission and death,[7, 8] the postoperative setting presents a particular challenge, as efforts to reduce clotting must be balanced against the risk of bleeding.

Despite how common this scenario is, there is no consensus regarding the best pharmacologic strategy. National guidelines recommend pharmacologic thromboprophylaxis, leaving the clinician to select the specific agent.[4, 5] Explicitly endorsed options include aspirin, vitamin K antagonists (VKA), unfractionated heparin, fondaparinux, low‐molecular‐weight heparin (LMWH) and IIa/Xa factor inhibitors. Among these, aspirin, the only nonanticoagulant, has been the source of greatest controversy.[4, 9, 10]

Two previous systematic reviews comparing aspirin to anticoagulation for VTE prevention found conflicting results.[11, 12] In addition, both used indirect comparisons, a method in which the intervention and comparison data come from different studies, and susceptibility to confounding is high.[13, 14] We aimed to overcome the limitations of prior efforts to address this commonly encountered clinical question by conducting a systematic review and meta‐analysis of randomized controlled trials that directly compared the efficacy and safety of aspirin to anticoagulants for VTE prevention in adults undergoing common high‐risk major orthopedic surgeries of the lower extremities.

MATERIAL AND METHODS

RESULTS

Results of Search

Figure 1 shows the number of studies that we evaluated during each stage of the study selection process. After full‐text review, 8 randomized trials met all inclusion criteria.[23, 24, 25, 26, 27, 28, 29, 30]

Figure 1

Rate of Proximal DVT

Pooling findings of all 7 studies that reported proximal DVT rates, we observed no statistically significant difference between aspirin and anticoagulants (10.4% vs 9.2%, relative risk [RR]: 1.15 [95% confidence interval {CI}: 0.68‐1.96], I²=41%). Although rates did not statistically differ between aspirin and anticoagulants in either operative subgroup, there appeared to be a nonsignificant trend favoring anticoagulation after hip fracture repair (12.7% vs 7.8%, RR: 1.60 [95% CI: 0.80‐3.20], I²=0%, 2 trials) but not following knee or hip arthroplasty (9.3% vs 9.7%, RR: 1.00 [95% CI: 0.49‐2.05], I²=49%, 5 trials) (Figure 2).

Figure 2

Rate of Pulmonary Embolism

Just 14 participants experienced a PE across all 6 trials reporting this outcome (aspirin n=9/405 versus anticoagulation n=5/415). Although PE was numerically more likely in the aspirin group, this difference was not statistically significant (overall: 1.9% vs 0.9%, RR: 1.83 [95% CI: 0.64, 5.21], I²=0%). The very small number of events rendered extremely wide 95% CIs in operative subgroup analyses (Figure 3).

Figure 3

Bleeding Rates

Pooling all 8 studies, aspirin was associated with a statistically significant 48% decreased risk of bleeding events compared to anticoagulants (3.8% vs 8.0%, RR: 0.52 [95% CI: 0.310.86], I²=8%). When subgrouped according to procedure, bleeding rates remained statistically significantly lower in the aspirin group following hip fracture (3.1% vs 10%, RR: 0.32 [95% CI: 0.130.77], I²=0%, 2 trials); however, the observed trend favoring aspirin was not statistically significant following arthroplasty (3.9% vs 7.8%, RR: 0.63 [95% CI: 0.331.21], I²=14%, 5 trials) (Figure 4).

Figure 4

Five studies reported major bleeding; event rates were low and no statistically significant differences between aspirin and anticoagulants were observed (hip fracture: 3.5% vs 6.3%, RR: 0.46 [95% CI: 0.141.48], I²=0%, 2 trials; knee/hip arthroplasty: 2.1% vs 0.6%, RR: 2.86 [95% CI: 0.6512.60], I²=0%, 3 trials).

DISCUSSION

We found the balance of risk versus benefit of aspirin compared to anticoagulation differed markedly according to type of surgery. After hip fracture repair, we found a 68% reduction in bleeding risk with aspirin compared to anticoagulants. This benefit, however, was associated with a nonsignificant increase in screen‐detected proximal DVT. Conversely, among patients undergoing knee or hip arthroplasty, we found no difference in proximal DVT risk between aspirin and anticoagulants and a possible trend toward less bleeding risk with aspirin. The rarity of pulmonary emboli (and death) made meaningful comparisons between aspirin and anticoagulation impossible for either type of surgery.

Our systematic review has several strengths that differentiate it from previous analyses. First, we only included head‐to‐head randomized trials such that all included data reflect direct comparisons between aspirin and anticoagulation in well‐balanced populations. Conversely, both recent reviews[11, 12] were based on indirect comparisons, a type of analysis in which data for the intervention and control arms are taken from different studies and thus different populations. This methodology is not recommended by the Cochrane Collaboration[13, 14] because of the increased risk of an unbalanced comparison. For example, Brown and colleagues' meta‐analysis, which pooled data from selected arms of 14 randomized controlled trials, found the efficacy of aspirin comparable to that of anticoagulants, but all aspirin subjects came from a single trial of patients at such low risk of VTE that a placebo arm was considered justified.[31] Similarly, in the indirect comparison of Westrich and colleagues,[12] which found anticoagulation superior to aspirin, the likelihood of an unbalanced comparison was further heightened by their inclusion of observational studies, with the attendant risk of confounding by indication.

Our systematic review further differs from previous analyses by examining both beneficial and harmful clinical outcomes, and doing so separately for the 2 most common types of major orthopedic lower extremity surgery. This allowed us to discover important differences in the comparative efficacy (benefit vs harm) of aspirin versus anticoagulants across different procedure types. Finding that aspirin may have lower efficacy for preventing VTE following hip fracture repair than arthroplasty may not be surprising in light of the nature of the 2 procedures, the disparate mean ages typical of patients who undergo each procedure, and the underlying trauma in hip fracture patients.

The limitations of our review largely reflect the quality of the studies we were able to include. First, our pooled sample size remains relatively small, meaning that observed nonsignificant differences between aspirin and anticoagulation groups (eg, a nonsignificant 60% increased risk of DVT for aspirin after hip repair, 95% CI: 0.803.20) could reasonably reflect up to 3‐fold differences in DVT risk and 5‐fold differences in PE rates. Second, screening for DVT, which is neither recommended nor common in clinical practice, was used in all studies. Reported DVT incidence, therefore, is undoubtedly higher than what would be observed in practice; however, the effect on the direction and magnitude of observed relative risks is unpredictable. Third, included studies used a wide range of aspirin doses, as well as a variety of anticoagulant types. Although supratherapeutic aspirin doses are unlikely to confer additional benefit for venous thromboprophylaxis, they may be associated with excess bleeding risk.[32] Finally, several of the studies were conducted more than10 years ago. Given changes in treatment practices, surgical technique, and prophylaxis options, the findings of these studies may not reflect current practice, in which early mobilization and intermittent pneumatic compression devices are standard prophylaxis against postoperative VTE. In fact, only 2 trials used concomitant pneumatic compression devices, and none treated patients longer than 21 days, the current standard being up to 35 days.[4] Although these limitations may affect overall event rates, this bias should be balanced between comparison groups, because we only included randomized controlled trials.

What is a clinician to do? Based on our findings, current guidelines recommending aspirin prophylaxis against VTE as an alternative following major lower extremity surgery may not be universally appropriate. We found that although overall bleeding complications are lower with aspirin, concerns about poor efficacy remain, specifically for patients undergoing hip fracture repair. Although some have suggested that aspirin use be restricted to low risk patients, this strategy has not been experimentally evaluated.[33] On the other hand, switching to aspirin after a brief initial course of LMWH may be an approach warranting further study, in light of a recent randomized controlled trial of 778 patients after elective hip replacement, which found equivalent efficacy using 10 days of LMWH followed by aspirin versus additional LMWH for 28 days.[34]

We are able to be more definitive, based on our study of best available trial data, in making recommendations to investigators embarking on further study of optimal VTE prophylaxis following major orthopedic surgery. First, distinguishing a priori between the 2 major types of lower extremity major orthopedic surgery is a high priority. Second, both bleeding and thromboembolic outcomes must be evaluated. Third, only symptomatic events should be used to measure VTE outcomes; clinical follow‐up must continue well beyond discharge, for at least 3 months to ensure ascertainment of clinically relevant VTE. Fourth, nonpharmacologic cointerventions should be standardized and represent the standard of care, including early immobilization and mechanical compression devices.

In summary, although definitive recommendations for or against the use of aspirin instead of anticoagulation for VTE prevention following major orthopedic surgery are not possible, our findings suggest that, following hip fracture repair, the lower risk of bleeding with aspirin is likely outweighed by a probable trend toward higher risk of VTE. On the other hand, the balance of these opposing risks may favor aspirin after elective knee or hip arthroplasty. A comparative study of aspirin, anticoagulation, and a hybrid strategy (eg, brief anticoagulation followed by aspirin) after elective knee or hip arthroplasty should be a high priority given our aging population and increasing demand for major orthopedic lower extremity surgery.

Each year in the United States, over 1 million adults undergo hip fracture surgery or elective total knee or hip arthroplasty.[1] Although highly effective for improving functional status and quality of life,[2, 3] each of these procedures is associated with a substantial risk of developing a deep vein thrombosis (DVT) or pulmonary embolism (PE).[4, 5] Collectively referred to as venous thromboembolism (VTE), these clots in the venous system are associated with significant morbidity and mortality for patients, as well as substantial costs to the healthcare system.[6] Although VTE is considered to be a preventable cause of hospital admission and death,[7, 8] the postoperative setting presents a particular challenge, as efforts to reduce clotting must be balanced against the risk of bleeding.

Despite how common this scenario is, there is no consensus regarding the best pharmacologic strategy. National guidelines recommend pharmacologic thromboprophylaxis, leaving the clinician to select the specific agent.[4, 5] Explicitly endorsed options include aspirin, vitamin K antagonists (VKA), unfractionated heparin, fondaparinux, low‐molecular‐weight heparin (LMWH) and IIa/Xa factor inhibitors. Among these, aspirin, the only nonanticoagulant, has been the source of greatest controversy.[4, 9, 10]

Two previous systematic reviews comparing aspirin to anticoagulation for VTE prevention found conflicting results.[11, 12] In addition, both used indirect comparisons, a method in which the intervention and comparison data come from different studies, and susceptibility to confounding is high.[13, 14] We aimed to overcome the limitations of prior efforts to address this commonly encountered clinical question by conducting a systematic review and meta‐analysis of randomized controlled trials that directly compared the efficacy and safety of aspirin to anticoagulants for VTE prevention in adults undergoing common high‐risk major orthopedic surgeries of the lower extremities.

MATERIAL AND METHODS

RESULTS

Results of Search

Figure 1 shows the number of studies that we evaluated during each stage of the study selection process. After full‐text review, 8 randomized trials met all inclusion criteria.[23, 24, 25, 26, 27, 28, 29, 30]

Figure 1

Rate of Proximal DVT

Pooling findings of all 7 studies that reported proximal DVT rates, we observed no statistically significant difference between aspirin and anticoagulants (10.4% vs 9.2%, relative risk [RR]: 1.15 [95% confidence interval {CI}: 0.68‐1.96], I²=41%). Although rates did not statistically differ between aspirin and anticoagulants in either operative subgroup, there appeared to be a nonsignificant trend favoring anticoagulation after hip fracture repair (12.7% vs 7.8%, RR: 1.60 [95% CI: 0.80‐3.20], I²=0%, 2 trials) but not following knee or hip arthroplasty (9.3% vs 9.7%, RR: 1.00 [95% CI: 0.49‐2.05], I²=49%, 5 trials) (Figure 2).

Figure 2

Rate of Pulmonary Embolism

Just 14 participants experienced a PE across all 6 trials reporting this outcome (aspirin n=9/405 versus anticoagulation n=5/415). Although PE was numerically more likely in the aspirin group, this difference was not statistically significant (overall: 1.9% vs 0.9%, RR: 1.83 [95% CI: 0.64, 5.21], I²=0%). The very small number of events rendered extremely wide 95% CIs in operative subgroup analyses (Figure 3).

Figure 3

Bleeding Rates

Pooling all 8 studies, aspirin was associated with a statistically significant 48% decreased risk of bleeding events compared to anticoagulants (3.8% vs 8.0%, RR: 0.52 [95% CI: 0.310.86], I²=8%). When subgrouped according to procedure, bleeding rates remained statistically significantly lower in the aspirin group following hip fracture (3.1% vs 10%, RR: 0.32 [95% CI: 0.130.77], I²=0%, 2 trials); however, the observed trend favoring aspirin was not statistically significant following arthroplasty (3.9% vs 7.8%, RR: 0.63 [95% CI: 0.331.21], I²=14%, 5 trials) (Figure 4).

Figure 4

Five studies reported major bleeding; event rates were low and no statistically significant differences between aspirin and anticoagulants were observed (hip fracture: 3.5% vs 6.3%, RR: 0.46 [95% CI: 0.141.48], I²=0%, 2 trials; knee/hip arthroplasty: 2.1% vs 0.6%, RR: 2.86 [95% CI: 0.6512.60], I²=0%, 3 trials).

DISCUSSION

We found the balance of risk versus benefit of aspirin compared to anticoagulation differed markedly according to type of surgery. After hip fracture repair, we found a 68% reduction in bleeding risk with aspirin compared to anticoagulants. This benefit, however, was associated with a nonsignificant increase in screen‐detected proximal DVT. Conversely, among patients undergoing knee or hip arthroplasty, we found no difference in proximal DVT risk between aspirin and anticoagulants and a possible trend toward less bleeding risk with aspirin. The rarity of pulmonary emboli (and death) made meaningful comparisons between aspirin and anticoagulation impossible for either type of surgery.

Our systematic review has several strengths that differentiate it from previous analyses. First, we only included head‐to‐head randomized trials such that all included data reflect direct comparisons between aspirin and anticoagulation in well‐balanced populations. Conversely, both recent reviews[11, 12] were based on indirect comparisons, a type of analysis in which data for the intervention and control arms are taken from different studies and thus different populations. This methodology is not recommended by the Cochrane Collaboration[13, 14] because of the increased risk of an unbalanced comparison. For example, Brown and colleagues' meta‐analysis, which pooled data from selected arms of 14 randomized controlled trials, found the efficacy of aspirin comparable to that of anticoagulants, but all aspirin subjects came from a single trial of patients at such low risk of VTE that a placebo arm was considered justified.[31] Similarly, in the indirect comparison of Westrich and colleagues,[12] which found anticoagulation superior to aspirin, the likelihood of an unbalanced comparison was further heightened by their inclusion of observational studies, with the attendant risk of confounding by indication.

Our systematic review further differs from previous analyses by examining both beneficial and harmful clinical outcomes, and doing so separately for the 2 most common types of major orthopedic lower extremity surgery. This allowed us to discover important differences in the comparative efficacy (benefit vs harm) of aspirin versus anticoagulants across different procedure types. Finding that aspirin may have lower efficacy for preventing VTE following hip fracture repair than arthroplasty may not be surprising in light of the nature of the 2 procedures, the disparate mean ages typical of patients who undergo each procedure, and the underlying trauma in hip fracture patients.

The limitations of our review largely reflect the quality of the studies we were able to include. First, our pooled sample size remains relatively small, meaning that observed nonsignificant differences between aspirin and anticoagulation groups (eg, a nonsignificant 60% increased risk of DVT for aspirin after hip repair, 95% CI: 0.803.20) could reasonably reflect up to 3‐fold differences in DVT risk and 5‐fold differences in PE rates. Second, screening for DVT, which is neither recommended nor common in clinical practice, was used in all studies. Reported DVT incidence, therefore, is undoubtedly higher than what would be observed in practice; however, the effect on the direction and magnitude of observed relative risks is unpredictable. Third, included studies used a wide range of aspirin doses, as well as a variety of anticoagulant types. Although supratherapeutic aspirin doses are unlikely to confer additional benefit for venous thromboprophylaxis, they may be associated with excess bleeding risk.[32] Finally, several of the studies were conducted more than10 years ago. Given changes in treatment practices, surgical technique, and prophylaxis options, the findings of these studies may not reflect current practice, in which early mobilization and intermittent pneumatic compression devices are standard prophylaxis against postoperative VTE. In fact, only 2 trials used concomitant pneumatic compression devices, and none treated patients longer than 21 days, the current standard being up to 35 days.[4] Although these limitations may affect overall event rates, this bias should be balanced between comparison groups, because we only included randomized controlled trials.

What is a clinician to do? Based on our findings, current guidelines recommending aspirin prophylaxis against VTE as an alternative following major lower extremity surgery may not be universally appropriate. We found that although overall bleeding complications are lower with aspirin, concerns about poor efficacy remain, specifically for patients undergoing hip fracture repair. Although some have suggested that aspirin use be restricted to low risk patients, this strategy has not been experimentally evaluated.[33] On the other hand, switching to aspirin after a brief initial course of LMWH may be an approach warranting further study, in light of a recent randomized controlled trial of 778 patients after elective hip replacement, which found equivalent efficacy using 10 days of LMWH followed by aspirin versus additional LMWH for 28 days.[34]

We are able to be more definitive, based on our study of best available trial data, in making recommendations to investigators embarking on further study of optimal VTE prophylaxis following major orthopedic surgery. First, distinguishing a priori between the 2 major types of lower extremity major orthopedic surgery is a high priority. Second, both bleeding and thromboembolic outcomes must be evaluated. Third, only symptomatic events should be used to measure VTE outcomes; clinical follow‐up must continue well beyond discharge, for at least 3 months to ensure ascertainment of clinically relevant VTE. Fourth, nonpharmacologic cointerventions should be standardized and represent the standard of care, including early immobilization and mechanical compression devices.

In summary, although definitive recommendations for or against the use of aspirin instead of anticoagulation for VTE prevention following major orthopedic surgery are not possible, our findings suggest that, following hip fracture repair, the lower risk of bleeding with aspirin is likely outweighed by a probable trend toward higher risk of VTE. On the other hand, the balance of these opposing risks may favor aspirin after elective knee or hip arthroplasty. A comparative study of aspirin, anticoagulation, and a hybrid strategy (eg, brief anticoagulation followed by aspirin) after elective knee or hip arthroplasty should be a high priority given our aging population and increasing demand for major orthopedic lower extremity surgery.