The Effect of Preoperative Antimicrobial Prophylaxis on Intraoperative Culture Results in Patients with a Suspected or Confirmed Prosthetic Joint Infection: a Systematic Review

ABSTRACT

Obtaining reliable cultures during revision arthroplasty is important to adequately diagnose and treat a prosthetic joint infection (PJI). The influence of antimicrobial prophylaxis on culture results remains unclear. Since withholding prophylaxis increases the risk for surgical site infections, clarification on this topic is critical. A systematic review was performed with the following research question: in patients who undergo revision surgery of a prosthetic joint, does preoperative antimicrobial prophylaxis affect the culture yield of intraoperative samples in comparison with nonpreoperative antimicrobial prophylaxis? Seven articles were included in the final analysis. In most studies, standard diagnostic culture techniques were used. In patients with a PJI, pooled analysis showed a culture yield of 88% (145/165) in the prophylaxis group versus 95% (344/362) in the nonprophylaxis group (P = 0.004). Subanalysis of patients with chronic PJIs showed positive cultures in 88% (78/89) versus 91% (52/57), respectively (P = 0.59). In patients with a suspected chronic infection, a maximum difference of 4% in culture yield between the prophylaxis and nonprophylaxis groups was observed. With the use of standard culture techniques, antimicrobial prophylaxis seems to affect cultures in a minority of patients. Along with the known risk of surgical site infections due to inadequate timing of antimicrobial prophylaxis, we discourage the postponement of prophylaxis until tissue samples are obtained in revision surgery. Future studies are necessary to conclude whether the small percentage of false-negative cultures after prophylaxis can be further reduced with the use of more-sensitive culture techniques, like sonication.

INTRODUCTION

In order to adequately diagnose and treat infections, it is of utmost importance that reliable cultures be obtained without prior use of antimicrobial therapy, because it may interfere with bacterial growth (1). This principle is even more crucial in patients with a prosthetic implant in situ that needs to be revised (2, 3). Misdiagnosis in this subgroup of patients may lead to unnecessary and invasive re-revision surgeries. These surgeries are accompanied not only by high health care costs due to longer hospital admissions and rehabilitation periods but also by a high patient mortality (4–6). In approximately 10 to 20% of patients with an aseptic loosening, an infection that was not known preoperatively is diagnosed during revision arthroplasty (7–9). This is partly explained by false-negative cultures due to prior antibiotic therapy (10–12). Subsequently, it has been advocated that preoperative antimicrobial prophylaxis should be withheld in this patient category until intraoperative tissue samples for culture are obtained. However, the importance of timely administration of preoperative antimicrobial prophylaxis to prevent surgical site infections is well known (13–17). This is particularly important in those patients undergoing revision surgery in whom infection is not suspected. Although extensive studies have shown that prolonged use of antimicrobial therapy before surgery results in false-negative intraoperative cultures, the influence of a single dose of antibiotics prior to surgery on culture yield is less clear (3). Since the protection of the new prosthetic implant for infection is critical, clarification on this topic is essential. We therefore performed a systematic review to evaluate if culture yield is affected by preoperative antimicrobial prophylaxis in revision arthroplasties.

RESULTS

Study inclusion.

Figure 1 shows the results of the search strategy. The search strategy included studies from 1979 until 2017. Five hundred twenty articles from PubMed and EMBASE were screened by title and abstract. Five hundred eleven articles were excluded based on patient group (patients without a clinical suspicion of a prosthetic joint infection [PJI]) and/or intervention (not studying the effect of antimicrobial prophylaxis on culture yield). Of the nine remaining articles, two were excluded after the full text was read: one study analyzed the culture results after preoperative prophylaxis in primary shoulder arthroplasty (18), and another study analyzed the culture results of prior antimicrobial therapy instead of preoperative antimicrobial prophylaxis (2). Finally, seven studies were considered eligible and included in the final analysis (3, 19–24). There was 100% agreement between the two reviewers (M. Wouthuyzen-Bakker and N. Benito) for the inclusion of the studies.

FIG 1.

Flow chart of included studies. Key terms that were used during the search in the electronic databases are described in Materials and Methods. P/I, patients or intervention.

Characteristics of included studies.

Table 1 shows the study characteristics of the seven selected studies representing a total of 739 patients. The majority of studies were performed on hip and/or knee arthroplasties. With the use of several study designs and the inclusion of different patient populations (e.g., suspected or confirmed PJI prior to surgery), there was a high clinical heterogeneity between the studies. In the majority of studies, sonication was not performed and only standard culture techniques were used.

TABLE 1.

Study characteristics^a

Reference	Authors	Yr	Country	Patients (n)	Study design	Joint	Antibiotic prophylaxis	Timing of antibiotic prophylaxis	Patient population	Use of blood culture bottles	Intraoperative biopsy samples (n)	Use of sonication	Incubation of cultures (days)
24	Wouthuyzen-Bakker et al.	2017	Spain	425	Retrospective	Knee	Ceftazidime and teicoplanin	During anesthesia	Suspected PJI	Yes	≥5	No	5
20	Pérez-Prieto et al.	2016	Spain	28	RCT	Hip, knee, shoulder	Cefazolin or glycopeptide	30 to 60 min before surgery	Suspected PJI	Yes	5	Yes	14
22	Bedenčič et al.	2015	Slovenia	40	Prospective	Hip, knee	Cefazolin	No specific data	Chronic PJI	No	3	No	14
3	Al-Mayahi et al.	2015	USA	241	Retrospective	Hip, knee	No specific data	60 min before surgery	PJI (type not defined)	No	ND	No	5
21	Tetreault et al.	2014	USA	65	RCT	Hip, knee	Cefazolin and glycopeptide	60 min before surgery	Acute and chronic PJI	No	≥3, including swabs	No	ND
19	Burnett et al.	2010	USA	26	Prospective	Knee	Cefazolin and glycopeptide	18 to 33 min before surgery	Acute and chronic PJI	No	3 swabs, 1 biopsy	No	ND
23	Ghanem et al.	2007	USA	171	Retrospective	Hip	Cefazolin, vancomycin, and/or ciprofloxacin	ND	Acute and chronic PJI	No	≥2	No	5

^aPatient characteristics, study design, antibiotic prophylaxis, and diagnostic culture methods of the included studies (3, 19–24). RCT, randomized controlled trial; PJI, prosthetic joint infection; ND, no data.

Results of quality assessment.

The results of the quality assessment are given in Table 2 and were judged as described in Materials and Methods. There was 75% agreement between the two reviewers (M. Wouthuyzen-Bakker and N. Benito). On the points on which they obtained no consensus, the third reviewer (A. Soriano) made the final decision.

TABLE 2.

Quality assessment^a

Study	Risk of bias based on internal validation				Applicability concerns based on external validation
	Patient selection	Index test	Reference test	Flow and timing	Patient selection	Index test	Reference test
Wouthuyzen-Bakker et al. (24)	LR	LR	HR	LR	LR	LR	LR
Pérez-Prieto et al. (20)	LR	LR	HR	LR	LR	LR	LR
Bedenčič et al. (22)	LR	LR	LR	LR	LR	LR	LR
Al-Mayahi et al. (3)	LR	?	HR	LR	HR	LR	LR
Tetreault et al. (21)	LR	HR	LR	LR	?	LR	LR
Burnett et al. (19)	LR	HR	LR	LR	?	LR	LR
Ghanem et al. (23)	HR	HR	LR	LR	?	LR	LR

^aQuality assessment of the included studies (3, 19–24) using the QUADAS-2 tool for diagnostic studies (29). Specific details on internal and external validation are described in Materials and Methods. LR, low risk; HR, high risk; ?, unclear risk.

(i) Internal validation.

We did not find major concerns in patient selection and/or flow and timing but did notice some limitations for the index and reference test. In most studies (five out of seven), the number of intraoperative tissue cultures was less than four and/or included swab cultures, which probably underestimated the diagnostic yield for the index test (25). Regarding the reference test, in three out of seven studies, patients did not serve as their own control and culture yield was compared only with that of a control group.

(ii) External validation.

Concerning the applicability to our review question, four out of seven studies comprised patients who already had a confirmed PJI and for which the causative microorganism was already known before the administration of prophylaxis in most of them. For most studies, data solely on the chronic infections were available in the article or were provided by the corresponding authors of the article.

Results of quantitative assessment. (i) Culture yield.

Tables 3 and 4 show the details of the intraoperative culture yield of the included studies. For the total group of patients with a confirmed PJI prior to surgery, the pooled culture yield was 88% (145/165) in the prophylaxis group versus 95% (344/362) in the control group without prophylaxis (P = 0.004), which represents a difference of 7% (95% confidence interval [CI], 4 to 10%). The statistical heterogeneity of this group of patients was negligible (heterogeneity [I²], 0%). Based on a subanalysis of patients with solely a chronic PJI (n = 146), the pooled culture yields were not different between the prophylaxis and the nonprophylaxis groups (88% [78/89] versus 91% [52/57], respectively; P = 0.59). This group of patients with solely a chronic PJI showed a moderate statistical heterogeneity (I², 33%). As depicted in Table 4, in patients with a suspected chronic PJI (two studies, n = 439), culture yield greatly depended on the pretest probability of having an infection in the studied population (e.g., 26 to 27% in patients with presumed aseptic loosening and/or chronic pain versus 56 to 60% in patients with at least one minor or major diagnostic criterion for infection as defined by the Musculoskeletal Infection Society [26]). In both studies, a difference of ≤4% in culture yield was observed between the prophylaxis group and the control group without prophylaxis (P ≥ 0.78).

TABLE 3.

Culture yield of patients with a confirmed PJI prior to surgery^a

Study authors	Reference test	Prophylaxis	Patients with acute and chronic infections			Patients with chronic infections only
			No. of patients with culture yield/total no. (%)	95% CI	P value	No. of patients with culture yield/total no. (%)	95% CI	P value
Bedenčič et al. (22)b	Intraoperative samples before prophylaxis	Yes	19/24 (79)	58–93		19/24 (79)	58–93
		No
Tetreault et al. (21)	Control group without prophylaxis	Yes	29/34 (85)	73–98	0.43	13/16 (81)	60–100	1.00
		No	29/31 (94)	84–100		14/16 (88)	70–100
Burnett et al. (19)	Preoperative synovial fluid	Yes	26/26 (100)	NA		19/19 (100)	NA
		No
Ghanem et al. (23)	Control group without prophylaxis	Yes	63/72 (88)	80–95	0.44	27/30 (90)	79–100	1.00
		No	91/99 (92)	86–98		38/41 (93)	84–100
Al-Mayahi et al. (3)	Control group without prophylaxis	Yes	8/9 (89)	63–100	0.29	ND
		No	224/232 (97)	95–99		ND
Pooled (all studies)		Yes	145/165 (88)	83–93	0.004	78/89 (88)	80–99	0.59
		No	344/362 (95)	93–97		52/57 (91)	83–99

^aSummary of the culture yield of patients with a diagnosed prosthetic joint infection (PJI) prior to surgery who underwent revision surgery with or without preoperative antimicrobial prophylaxis (3, 19, 21–23). Data are given for the total patients in the group as well as for the subanalysis of the patients with solely chronic infections. NA, not applicable; ND, no data.

^bFrom the data for this cohort, only 24 out of 40 patients with a positive culture before the administration of antimicrobial prophylaxis were analyzed in order to calculate the sensitivity of cultures after antimicrobial prophylaxis.

TABLE 4.

Culture yield of patients with a suspected chronic PJI prior to surgery^a

Study authors	Reference test	Studied population	Prophylaxis	No of patients with culture yield/total no. (%)	95% CI	P value
Wouthuyzen-Bakker et al. (24)	Control group without prophylaxis	Low probability of infectionb	Yes	73/284 (26)	21–31	0.78
			No	38/141 (27)	20–35
Pérez -Prieto et al. (20)	Control group without prophylaxis	High probability of infectionc	Yes	5/9 (56)	16–97	1.00
			No	3/5 (60)	0–72

^aSummary of the culture yield of patients with a suspected prosthetic joint infection (PJI) prior to surgery who underwent revision surgery with or without preoperative antimicrobial prophylaxis (20, 24).

^bPatients with a (presumed) aseptic loosening or chronic pain who underwent a total, partial, or minor revision surgery.

^cPatients who met one or more of the minor or major diagnostic criteria as defined by the Musculoskeletal Infection Society (26). From the data for this cohort, patients that were treated with debridement or antibiotics and with retention of the prosthesis in this study (14 out of 28) were excluded from the analysis.

(ii) Cultured microorganisms.

Limited data were available regarding specific details of the culture results. The available data are provided in Table 5. No differences in the cultured microorganisms were found between the prophylaxis and nonprophylaxis groups.

TABLE 5.

Intraoperative culture yield per microorganism^a

Microorganism(s)	No. of patients with culture yield/total no. (%) in indicated study
	Wouthuyzen-Bakker et al. (24)		Pérez-Prieto et al. (20)b		Tetreault et al. (21)		Bedenčič et al. (22), postprophylaxis	Burnett et al. (19), postprophylaxis
	Control group	Antibiotic prophylaxis	Control group	Antibiotic prophylaxis	Control group	Antibiotic prophylaxis
Staphylococcus aureus	2/284 (0.7)	1/141 (0.7)	0/5 (0)	1/9 (11)	13/13 (100)	14/15 (93)	8/8 (100)	14/14 (100)
Coagulase-negative staphylococci	60/284 (21)	30/141 (21)	0/5 (0)	3/9 (33)	9/10 (90)	6/7 (86)	8/10 (80)	4/4 (100)
Streptococcus species	2/284 (0.7)	0/141 (0)	0/5 (0)	0/9 (0)	4/4 (100)	2/3 (67)	1/1 (100)	3/3 (100)
Enterococcus species	3/284 (1)	0/141 (0)	1/5 (20)	0/9 (0)	0/0	2/3 (67)	1/1 (100)
Propionibacterium species	0/284 (0)	0/141 (0)	1/5 (20)	1/9 (11)	1/1 (100)	0/0	0/2 (0)
Corynebacterium species	6/284 (2)	2/141 (1.4)	0/5 (0)	0/9 (0)	0/1 (0)	0/1 (0)	2/3 (67)
Gram negatives	6/284 (2)	2/141 (1.4)	0/5 (0)	1/9 (11)			2/2 (100)	2/2 (100)

^aAvailable data of included studies (19–22, 24) on the most frequently cultured microorganisms in patients with or without antimicrobial prophylaxis. See Table 1 for the specific details on the study design. In this table, the group of patients without prophylaxis is labeled as either (i) “control group” or (ii) “postprophylaxis,” when the patient served as his own control, i.e., with positive cultures prior to surgery. In none of the studies was a statistically significant difference found between the cultured microorganisms in patients with or without prophylaxis.

^bOnly the data on the cultured microorganisms in chronic infections were available in this study.

In the study of Bedenčič et al. (22), in which positive intraoperative tissue cultures were used as a reference test before the administration of antimicrobial prophylaxis, negative cultures were found after prophylaxis in 5 out of 24 patients (21%) for coagulase-negative staphylococci (2 patients), Propionibacterium acnes (2 patients), and Corynebacterium species (1 patient). Unfortunately, in this study, no control group without the administration of antimicrobial prophylaxis was available to determine whether the rate of culture negativity of the second obtained samples at the same surgical area would be different for these microorganisms.

DISCUSSION

With the scarce clinical heterogeneity of studies available in literature, the results of our systematic review indicate that culture yield is affected in a minority (7%) of patients by the administration of preoperative antimicrobial prophylaxis. Although our analysis does show a significant influence of one dose of antimicrobial prophylaxis on culture results, the clinical relevance of this difference is less clear. In addition, we would like to stress that the results should be interpreted carefully.

Because of the variety of study designs and studied patient populations and the differences in diagnostic methods of the included studies, it is difficult to compose a rather homogenous group of patients to pool and analyze the data. To illustrate, in the pooled analysis, we found an important discrepancy between the total group of analyzed patients in which a PJI was diagnosed prior to surgery and the analysis of patients with solely chronic infections. In the latter (more homogenous) group, the difference in culture yield between the prophylaxis and nonprophylaxis groups was much less pronounced, only 3%, and did not reach statistical significance. We cannot fully understand this observation since one expects that culture yield would become less, not more, affected in a group that also included acute infections with more-virulent microorganisms and a higher bacterial inoculum. A possible explanation of this observation is that the group with chronic infections might be statistically underpowered (n = 146). Another possibility is that the study of Al-Mayahi et al. (3) created a bias in the final results, since this study comprised a large number of patients who did not receive prophylaxis (n = 232; 96% of the studied population). The culture yield of these patients in whom prophylaxis was not administered was much higher (97%) than that of the group without prophylaxis in the other included studies (92 to 94%). We do not have information on the type of infections in the study of Al-Mayahi et al., but if acute infections predominated, this might explain the high(er) culture yield in this study and, thus, would create bias in the pooled analysis of the control group. Indeed, without the inclusion of this study, the culture yields were 88% in the nonprophylaxis group (n = 156) and 92% in the prophylaxis group (n = 130) (P = 0.24) (data not shown). Since the confidence intervals of the study of Al-Mayahi et al. and other studies fell within the same range and the calculated statistical heterogeneity was 0%, we concluded that the study of Al-Mayahi et al. should be included in the final analysis, considering the aforementioned concerns.

When taken into account, the 7% difference in culture yield that we observed in the pooled analysis of all patients with PJIs diagnosed prior to surgery, it needs to be pointed out that, thus, 93% of these patients were correctly diagnosed as infected despite the administration of preoperative antimicrobial prophylaxis. This percentage of correctly diagnosed patients was even higher for patients with a chronic infection, where, in our analysis, the difference in culture yield was as low as 3%. This finding is pivotal to mention in regard to patients who undergo revision surgery with a low preoperative probability of infection. To exemplify, when 100 patients undergo revision surgery because of an aseptic loosening and a maximum of 20 of these patients would have a chronic infection (7–9), 19.4 patients are still correctly diagnosed and only 0.6 (3%) of them will be misdiagnosed as not infected. Moreover, when antimicrobial prophylaxis is postponed in these patients, 80 patients will unnecessarily be put at risk for developing surgical site infections (13–17). It has been shown that when antimicrobial prophylaxis is administered 0 to 60 min after surgical incision, the risk of developing surgical site infections is increased 2-fold, compared to the risk with timely administration (15). Even more importantly, we previously demonstrated in patients who undergo total revision surgery that the PJI rate in the early postoperative period decreased from 6.4% to 1.6% when antimicrobial prophylaxis was no longer postponed until intraoperative tissue cultures were obtained (24). These data suggest that withholding prophylaxis does more harm than good, especially in one-stage revision surgery or in the second stage of a two-stage exchange with a newly placed implant. This risk of developing a postoperative infection may be less pronounced when a two-stage approach is chosen and a spacer is placed. The exact risk/benefit ratio of delaying prophylaxis in different subgroups of patients should be further addressed in future studies.

Although we did observe a lower culture yield in the prophylaxis group than in the nonprophylaxis group, it is important to note that in the majority of studies, only standard culture techniques were applied. Tissue cultures are known for their high rate of false negatives, and therefore it is advised that at least five samples be obtained (27). Culture yield is further increased by the use of blood culture bottles as a culture medium (27). However, most studies collected a suboptimal number of samples or even included swab cultures, and blood culture bottles were not used in the majority of them. Moreover, Trampuz et al. showed that despite prior use of antibiotic therapy 2 days before surgery, culture yield is increased by ±18% with the use of sonication (28). Again, only one out of seven studies included the use of sonication (20). This study indeed showed a better culture yield after prophylaxis than the other studies. Therefore, it is reasonable to assume that with the use of more-sensitive culture techniques, culture yield can be further optimized.

To conclude, culture yield seems to be affected by antimicrobial prophylaxis in a small percentage of patients. Considering the importance of preventing surgical site infection during revision arthroplasty, we believe that physicians should be cautious in withholding prophylaxis to increase culture results. We therefore suggest that preoperative surgical antimicrobial prophylaxis should be maintained, especially in patients with a low probability of infection who are undergoing revision arthroplasty. Future studies are necessary to conclude whether culture yield can be further improved with the use of more-sensitive culture techniques, like sonication.

MATERIALS AND METHODS

Research question.

The research question was described by using the PICOS approach (patients [P], intervention [I], control [C], outcome [O], and study design [S]): in patients who undergo a revision surgery of a prosthetic joint (P), does preoperative antimicrobial prophylaxis (I) affect the culture yield (O) of intraoperative samples in comparison with nonpreoperative antimicrobial prophylaxis (C)? Observational studies and clinical trials were included (S) (29).

Inclusion and exclusion criteria.

We included studies that involved all revision arthroplasties. Only peer-reviewed published studies were included. If studies comprised patients with chronic prosthetic joint infections as well as acute infections, corresponding authors were approached and asked to provide the specific data on chronic infections as well. A chronic infection was defined as symptoms persisting for at least 3 weeks in duration. Studies written in English, Spanish, and Dutch were included in the search strategy. Patients who received antimicrobial treatment in the 2 weeks prior to surgery were excluded from the analysis.

Search strategy.

MEDLINE and EMBASE were used as electronic databases. For PubMed, the following predefined keywords were used during the search: (((culture OR culture yield)) AND (prophylaxis OR antibiotic prophylaxis OR preoperative prophylaxis)) AND (revision OR revision surgery OR revision arthroplasty OR orthopedic surgery). For EMBASE, the same search strategy was used, with the addition of the following Emtree search terms: “bacterium culture”/exp OR “bacterium culture”/de AND “prosthesiology”/exp OR “prosthesiology”/de AND “antibiotic prophylaxis”/exp OR “antibiotic prophylaxis”/de. The search strategy was performed by two independent reviewers (M. Wouthuyzen-Bakker and N. Benito).

Methodological quality assessment.

For the quality assessment of the included studies, we used the QUADAS-2 tool that has been developed for the quality assessment of diagnostic studies (30). We evaluated the diagnostic accuracy of intraoperative cultures after the administration of antimicrobial prophylaxis, i.e., the index test. Intraoperative cultures of the same patient before the administration of prophylaxis, preoperative synovial fluid cultures, or a control group without prophylaxis were used as a reference, i.e., the reference test.

(i) Internal validation.

When preoperative antimicrobial prophylaxis was administered based on the judgment of the treating surgeon, this was considered a high risk of bias in patient selection. The risk of bias was considered high for the index test when preoperative antimicrobial prophylaxis was administered too late (25) and/or when ≤4 intraoperative biopsy samples or superficial wound swabs were obtained for culture during surgery (27). We did consider ≥3 intraoperative biopsy samples sufficient when tissue samples of the same surgical site were compared before and after the administration of antimicrobial prophylaxis, since a false-negative culture due to sampling error is less likely in this case. The risk of bias was considered high for the reference test when patients did not serve as their own control and culture yield was compared only with that of a control group (31). Since variations in bacterial inoculum are not expected in chronic PJI, the risk of bias regarding the interval between the collection of preoperative synovial fluid and intraoperative samples (flow and timing) was neglected.

External validation.

The applicability of the studies to our research question in patient selection was considered a high risk when no exclusive data were available for chronic infections and as an unclear risk when patients with preoperative positive synovial fluid cultures were included. The latter was considered a possible risk factor for applicability to our research question because antimicrobial prophylaxis is not always withheld from patients with a known causative microorganism preoperatively and the bacterial inoculum may be higher in these cases. The index and reference tests were judged as a high risk for applicability when the interpretation of the tests differed from our review question. Two independent reviewers (M. Wouthuyzen-Bakker and N. Benito) judged the included studies. When no consensus on the quality assessment was obtained, a third reviewer performed the final judgment (A. Soriano).

Statistical analysis.

The percentages of negative cultures were analyzed as categorical variables. A chi-square test (or a Fisher exact test when appropriate) was used to analyze the difference between the prophylaxis and nonprophylaxis groups. The results on culture yield were pooled for patients with a PJI diagnosed prior to surgery. An I² test was performed to test for statistical heterogeneity. A P value of <0.05 was considered statistically significant. Statistical analysis was performed using SPSS, version 20.0 (SPSS Inc., Chicago, IL).