Study characteristics

Table 1 and Appendix 3 (online) present the characteristics of included studies. The selected studies included 18 RCTs^{Reference Garland, Alex and Mueller19-Reference Arpa and Aygün36} and 2 quasi-experimental studies,^{Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37,Reference O’Horo, Corson and Baum38} and all of them were performed in hospital settings. Furthermore, 6 studies were multicentric (≥ 2 hospitals),^{Reference Garland, Alex and Mueller19-Reference Timsit, Mimoz and Mourvillier23,Reference Biehl, Huth and Panse27} 14 studies were performed on adults,^{Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27-Reference Pedrolo, Danski and Vayego29,Reference Chambers, Sanders and Patton31,Reference Ruschulte, Franke and Gastmeier33-Reference Yu, Lu, Meng, Zhao and Li35,Reference O’Horo, Corson and Baum38} and 6 studies examined neonates or pediatric populations.^{Reference Garland, Alex and Mueller19,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Gerceker, Yardimci and Aydinok30,Reference Levy, Katz and Solter32,Reference Arpa and Aygün36,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} Of the studies with this information, 13 included ICU patients,^{Reference Garland, Alex and Mueller19,Reference Arvaniti, Lathyris and Clouva-Molyvdas21-Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29,Reference Levy, Katz and Solter32,Reference Roberts and Cheung34-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} 4 included onco-hematological patients,^{Reference Biehl, Huth and Panse27,Reference Gerceker, Yardimci and Aydinok30,Reference Chambers, Sanders and Patton31,Reference Ruschulte, Franke and Gastmeier33} and 1 included non-ICU patients.^{Reference Chan, Northfield and Larsen28} The definition of CRBSI, exit-site and tunnel infections, and adverse events varied across the studies (Appendix 4 online). Also, 5 studies reported CRBSI rates ≥ 2.0 per 1,000 catheter days^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Gerceker, Yardimci and Aydinok30,Reference Yu, Lu, Meng, Zhao and Li35} in at least one study group and 2 studies reported CRBSI rates <1.5 per 1,000 catheter days in both groups.^{Reference Timsit, Schwebel and Bouadma22,Reference Timsit, Mimoz and Mourvillier23}

Types of catheters and insertion practices

Among studies with available data, 11 included short-term CVCs,^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Margatho, Ciol and Hoffman24-Reference Biehl, Huth and Panse27,Reference Pedrolo, Danski and Vayego29,Reference Levy, Katz and Solter32,Reference Ruschulte, Franke and Gastmeier33,Reference Yu, Lu, Meng, Zhao and Li35-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} 3 included long-term CVCs,^{Reference Gerceker, Yardimci and Aydinok30,Reference Chambers, Sanders and Patton31,Reference O’Horo, Corson and Baum38} 3 included CVCs and arterial catheters,^{Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20,Reference Timsit, Schwebel and Bouadma22,Reference Timsit, Mimoz and Mourvillier23} 1 included long and short-term CVCs,^{Reference Garland, Alex and Mueller19} and 1 included PICC lines.^{Reference Chan, Northfield and Larsen28} In addition, 2 studies used antimicrobial-impregnated catheters.^{Reference Biehl, Huth and Panse27,Reference Ruschulte, Franke and Gastmeier33} The duration of catheterization and catheter insertion sites between CHG and control groups were similar in studies that reported these data. Femoral insertions represented ≥44% of catheters in 3 studies.^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Yu, Lu, Meng, Zhao and Li35} One study used different skin antiseptics (alcohol versus povidone-iodine) in each study arm,^{Reference Garland, Alex and Mueller19} and another study applied daily CHG bathing in all patients.^{Reference Margatho, Ciol and Hoffman24} Also, 3 studies used >0.5% of CHG skin preparation before catheter insertion.^{Reference Chan, Northfield and Larsen28,Reference Yu, Lu, Meng, Zhao and Li35,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} Maximal sterile barrier precautions were cited in 14 studies.^{Reference Garland, Alex and Mueller19,Reference Arvaniti, Lathyris and Clouva-Molyvdas21-Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29,Reference Gerceker, Yardimci and Aydinok30,Reference Levy, Katz and Solter32-Reference Arpa and Aygün36}

Interventions

In the intervention group, 10 studies used CHG discs (Biopatch)^{Reference Garland, Alex and Mueller19-Reference Timsit, Schwebel and Bouadma22,Reference Chan, Northfield and Larsen28,Reference Chambers, Sanders and Patton31-Reference Roberts and Cheung34,Reference O’Horo, Corson and Baum38} , 8 used CHG transparent dressings (Tegaderm or unspecified),^{Reference Timsit, Mimoz and Mourvillier23-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Gerceker, Yardimci and Aydinok30,Reference Yu, Lu, Meng, Zhao and Li35-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} and 2 used “CHG-impregnated dressings” without further detail.^{Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29} In the control group, 16 studies used transparent dressings,^{Reference Garland, Alex and Mueller19-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Chan, Northfield and Larsen28,Reference Gerceker, Yardimci and Aydinok30,Reference Levy, Katz and Solter32-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} 2 used gauzes,^{Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29} 1 used standard-of-care dressings,^{Reference O’Horo, Corson and Baum38} and 1 did not use dressings.^{Reference Chambers, Sanders and Patton31}

Most studies planned a 7-day dressing change for the CHG group (n = 13).^{Reference Garland, Alex and Mueller19,Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20,Reference Margatho, Ciol and Hoffman24-Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Chan, Northfield and Larsen28-Reference Chambers, Sanders and Patton31,Reference Ruschulte, Franke and Gastmeier33,Reference Yu, Lu, Meng, Zhao and Li35-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} When the scheduled dressing change was different between the CHG and control groups,^{Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29,Reference Yu, Lu, Meng, Zhao and Li35,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} dressings in the control group were expected to be changed more often (ie, every 24 hours to 3 days).

Risk of bias assessment

The risk of bias for each study is reported in Figure 2. None of the studies blinded participants or healthcare personnel to the type of dressing used, and 6 RCTs blinded the outcome assessors.^{Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20-Reference Margatho, Ciol and Hoffman24,Reference Chan, Northfield and Larsen28} Furthermore, 6 RCTs were stopped before reaching the planned sample size or had insufficient power to detect differences between study groups.^{Reference Garland, Alex and Mueller19,Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Gerceker, Yardimci and Aydinok30,Reference Roberts and Cheung34}

Fig. 2. Risk of bias for the CRBSI outcome assessment in the included studies. (A) Randomized controlled trials (n = 18). (B) Quasi-experimental studies (n = 2).

Catheter-related bloodstream infections

After pooling the 20 studies, the CRBSI rate was 2.0% (160 of 7,826) in the CHG group and 3.2% (252 of 7,764) in the control group. The risk of CRBSI was significantly lower in the CHG group than in the control group (pRR, 0.71; 95% CI, 0.58–0.87). The test for heterogeneity was not statistically significant (P = .70; I ² = 0%) (Fig. 3). The funnel plot was symmetric, and there was no evidence of publication bias (Appendix 5 online). CHG dressings remained associated with a significant reduction of CRBSIs regardless of the type of dressing used: CHG-impregnated disc (10 studies; pRR, 0.74; 95% CI, 0.57–0.95)^{Reference Garland, Alex and Mueller19-Reference Timsit, Schwebel and Bouadma22,Reference Chan, Northfield and Larsen28,Reference Chambers, Sanders and Patton31-Reference Roberts and Cheung34,Reference O’Horo, Corson and Baum38} or CHG transparent dressing (8 studies; pRR, 0.64; 95% CI, 0.44–0.94).^{Reference Timsit, Mimoz and Mourvillier23-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Gerceker, Yardimci and Aydinok30,Reference Yu, Lu, Meng, Zhao and Li35-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37}

Fig. 3. Forest plots of the comparison of CHG dressings versus standard of care for the main study outcomes. (A) Catheter-related bloodstream infections. (B) exit-site infection/tunnel infection. (C) Adverse events.

Sensitivity analyses accounting for study differences (eg, CRBSI definition) and subgroup analyses are shown in Appendix 6 (online). The risk of CRBSI in RCTs versus nonrandomized trials showed differing results: pRR, 0.65 (95% CI, 0.49–0.85) versus pRR, 0.80 (95% CI, 0.59–1.09), respectively.

Effectiveness by study population and setting

In adults, CHG dressings significantly reduced CRBSIs (14 studies; pRR 0.63; 95% CI, 0.51–0.77), including onco-hematological patients (3 studies; pRR, 0.54; 95% CI, 0.36–0.81).^{Reference Biehl, Huth and Panse27,Reference Chambers, Sanders and Patton31,Reference Ruschulte, Franke and Gastmeier33} Although CHG dressings did not significantly decrease CRBSIs in ICUs (8 studies; pRR, 0.70; 95% CI, 0.41–1.18),^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Pedrolo, Danski and Vayego29,Reference Roberts and Cheung34,Reference Yu, Lu, Meng, Zhao and Li35} 2 high-quality studies suggested that CHG dressings were effective in ICUs with CRBSI rates <1.5 per 1,000 catheter days (pRR, 0.37; 95% CI, 0.15–0.90).^{Reference Timsit, Schwebel and Bouadma22,Reference Timsit, Mimoz and Mourvillier23} One single study was performed in non-ICU patients (RR, 1.24; 95% CI, 0.05–29.76).^{Reference Chan, Northfield and Larsen28} Analyses considering other settings and populations (eg, hemodialysis patients) were not possible.

In neonates and pediatric populations, CHG dressings did not significantly reduce CRBSIs (6 studies; pRR, 0.90; 95% CI, 0.57–1.40).^{Reference Garland, Alex and Mueller19,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Gerceker, Yardimci and Aydinok30,Reference Levy, Katz and Solter32,Reference Arpa and Aygün36,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} These results were consistent when neonates were excluded from the analysis (4 studies; pRR, 0.71; 95% CI, 0.33–1.50)^{Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Gerceker, Yardimci and Aydinok30,Reference Arpa and Aygün36,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} and when only neonates or pediatric patients admitted to the ICU were considered (5 studies; pRR, 0.86; 95% CI, 0.55–1.35).^{Reference Garland, Alex and Mueller19,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Levy, Katz and Solter32,Reference Arpa and Aygün36,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37}

Effectiveness by catheter type and insertion site

For short-term CVCs, CHG dressings decreased CRBSIs (11 studies; pRR, 0.67; 95% CI, 0.50–0.90).^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Margatho, Ciol and Hoffman24-Reference Biehl, Huth and Panse27,Reference Pedrolo, Danski and Vayego29,Reference Levy, Katz and Solter32,Reference Ruschulte, Franke and Gastmeier33,Reference Yu, Lu, Meng, Zhao and Li35-Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} This benefit was evident for studies with a mean length of catheterization >7 days (7 studies; pRR, 0.58; 95% CI, 0.42–0.81),^{Reference Margatho, Ciol and Hoffman24-Reference Biehl, Huth and Panse27,Reference Ruschulte, Franke and Gastmeier33,Reference Arpa and Aygün36,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37} but not for catheterizations ≤7 days (3 studies; pRR, 1.34; 95% CI, 0.54–3.32).^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Pedrolo, Danski and Vayego29,Reference Roberts and Cheung34}

For long-term CVCs, CHG dressings did not significantly decrease CRBSIs (3 studies; pRR, 0.80; 95% CI, 0.22–2.95).^{Reference Gerceker, Yardimci and Aydinok30,Reference Chambers, Sanders and Patton31,Reference O’Horo, Corson and Baum38} However, only 2 studies were RCTs, and they had small sample sizes. Another small RCT evaluated PICC lines and did not find a protective effect of CHG dressing use (RR, 1.24; 95% CI, 0.05–29.76).^{Reference Chan, Northfield and Larsen28}

None of these studies reported the incidence of CRBSI stratified by insertion site. When we pooled studies with jugular/subclavian insertions or those with frequent femoral insertions, the small sample size precluded definitive conclusions (Appendix 6 online). Additionally, Timsit 2012 showed a significant reduction in CRBSI rates among patients with venous catheters but not with arterial catheters.^{Reference Timsit, Mimoz and Mourvillier23}

Effectiveness by the frequency of dressing change

Among studies with a similar frequency of dressing change in both groups, CHG dressings significantly reduced CRBSIs (12 studies; pRR, 0.58; 95% CI, 0.42–0.82).^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27,Reference Chan, Northfield and Larsen28,Reference Gerceker, Yardimci and Aydinok30,Reference Levy, Katz and Solter32-Reference Roberts and Cheung34,Reference Arpa and Aygün36} When the planned frequency was 7 days for the CHG group and ≤3 days for the control group, the benefit of CHG dressings was lost (5 studies; pRR, 0.73; 95% CI, 0.46–1.14).^{Reference Maki, Mermel, Kluger, Narans, Knasinski, Parenteau and Covington20,Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Pedrolo, Danski and Vayego29,Reference Yu, Lu, Meng, Zhao and Li35,Reference Ergul, Gokcek, Ozcan, Cetin, Gultekin and Torun37}

Exit-site and tunnel infections

Ten studies reported exit-site and/or tunnel infections,^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Timsit, Schwebel and Bouadma22,Reference Margatho, Ciol and Hoffman24-Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Chan, Northfield and Larsen28,Reference Gerceker, Yardimci and Aydinok30,Reference Chambers, Sanders and Patton31,Reference Arpa and Aygün36,Reference O’Horo, Corson and Baum38} with an incidence of 1% (38 of 3,688) in the CHG group and 2% (74 of 3,625) in the control group. When these studies were pooled, CHG dressings decreased incident exit-site and tunnel infections (pRR, 0.51; 95% CI, 0.33–0.80) (Fig. 3). The test for heterogeneity was not statistically significant (P = .39, I ² = 5%), but there was evidence of publication bias (Appendix 5 online). When stratified by catheter type, CHG dressings remained protective against exit-site and tunnel infections in long-term CVCs (3 studies; pRR, 0.37; 95% CI, 0.22–0.64)^{Reference Gerceker, Yardimci and Aydinok30,Reference Chambers, Sanders and Patton31,Reference O’Horo, Corson and Baum38} but not in short-term CVCs (5 studies; pRR, 0.85; 95% CI, 0.38–1.91).^{Reference Arvaniti, Lathyris and Clouva-Molyvdas21,Reference Margatho, Ciol and Hoffman24-Reference Duzkaya, Sahiner, Uysal, Yakut and Citak26,Reference Arpa and Aygün36} The single RCT that assessed PICC lines did not show an association between CHG dressings and exit-site and tunnel infections (RR, 0.41; 95% CI, 0.06–2.78).^{Reference Chan, Northfield and Larsen28}

Adverse events

Ten articles evaluated the risk of adverse events,^{Reference Garland, Alex and Mueller19,Reference Timsit, Schwebel and Bouadma22-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Biehl, Huth and Panse27-Reference Pedrolo, Danski and Vayego29,Reference Levy, Katz and Solter32,Reference Arpa and Aygün36} with a pooled frequency of 2.8% (138 of 5,013) in the CHG group and 1.7% (75 of 4,807) in the control group. Premature neonates with weights <1,000 g had an increased incidence of 15% and a risk of severe reactions such as pressure necrosis.^{Reference Garland, Alex and Mueller19} Definitions of adverse events were varied, but all included reactions at the catheter insertion site (Appendix 4 online).

In the pooled analysis, CHG dressings were not significantly associated with adverse events (10 studies; pRR, 1.46; 95% CI, 0.85–2.50) (Fig. 3). However, there was evidence of publication bias (Appendix 5 online), and the heterogeneity was high (P = .03; I ² = 52%). When the 3 studies that combined contact dermatitis with unspecific events (eg, bruise at device removal) were excluded,^{Reference Biehl, Huth and Panse27-Reference Pedrolo, Danski and Vayego29} there was an independent association between CHG dressing and contact dermatitis (pRR, 5.16; 95% CI, 2.09–12.70; heterogeneity: P = .68; I ² = 0%). This finding was evident for both neonates and pediatric populations (3 studies; pRR, 9.20; 95% CI, 2.10–40.26)^{Reference Garland, Alex and Mueller19,Reference Levy, Katz and Solter32,Reference Arpa and Aygün36} and adults (4 studies; pRR, 3.65; 95% CI, 1.17–11.40).^{Reference Timsit, Schwebel and Bouadma22-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25} Contact dermatitis occurred with transparent CHG dressings (4 studies; pRR, 3.70; 95% CI, 1.24–11.04)^{Reference Timsit, Mimoz and Mourvillier23-Reference Pivkina, Gusarov, Blot, Zhivotneva, Pasko and Zamyatin25,Reference Arpa and Aygün36} and CHG-impregnated discs (3 studies; pRR, 10.42; 95% CI, 2.12–51.27).^{Reference Garland, Alex and Mueller19,Reference Timsit, Schwebel and Bouadma22,Reference Levy, Katz and Solter32}

CHG resistance

Two studies evaluated and did not find an association between CHG dressings and CHG resistance.^{Reference Timsit, Schwebel and Bouadma22,Reference Chambers, Sanders and Patton31} However, the median follow-up in these studies was either short (7 days)^{Reference Timsit, Schwebel and Bouadma22} or bacterial isolates causing exit-site infections were not investigated.^{Reference Chambers, Sanders and Patton31}