Will you trade me a CLABSI for a pneumothorax?

There was an interesting new study published today in the NEJM by Jean-Jacques Parienti and colleagues and funded by the French Ministry of Health. The study aimed to compare the rates of catheter-related bloodstream infection and symptomatic deep-vein thrombosis (DVT) in 2532 adult ICU patients randomized to subclavian (N=843), jugular (N=845) or femoral vein (844) sites for nontunneled central venous catheter insertion. Baseline characteristics were very similar across the three groups.

The results are not very surprising with higher rates of infection and symptomatic DVT when catheters were inserted using the jugular and femoral veins compared to the subclavian veins but these were almost cancelled out by a higher risk of pneumothorax when using the subclavian approach. (see figure below).

Here are my thoughts on the paper:

1. This is a very nice study that confirms the results of prior smaller studies in the field. 
2. The overall complication rate is low with approximately 97% of catheters in each group inserted safely without a pneumothorax and without infection or DVT. 
3. When all complications are included, it doesn’t appear there is a preferred site for catheter insertion. However, since hospitals are penalized for infections and not other complications**, there has been a preference for chosing the subclavian site, since it is associated with fewer infections. Unfortunately, the subclavian site had a much higher rate of pneumothorax. 
4. One important comment that the authors make is that how long the catheter remains in place can greatly increase the rate of infections and blood clots. So, if the doctor expects the catheter to remain in for a short period of time, she might chose a femoral or jugular vein approach to limit the pneumothorax risk, with very little infection or clotting risk since the catheter will be removed before the complication can occur. However, if the catheter is to remain in place for many days, it is probably worth the higher risk of pneumothorax associated with the subclavian site, which only occurs when the catheter is inserted, in order to reduce the long term infection and DVT risks. 
5. The study sites did not use daily chlorhexidine bathing and did not place chlorhexidine-impregnated dressings at the catheter insertion site. Both of these interventions have been shown to reduce catheter-related bloodstream infections. Thus, these results might not be generalizable to hospitals that use chlorhexidine bathing and/or chlorhexidine dressings. It's possible that either of these interventions or both could mitigate the infection risk rendering femoral or jugular vein approaches safer than they appear in this study.
6. Overall, it looks like fears of using the femoral vein are exaggerated, especially when you consider the rates of other complications like pneumothorax. It may be that in our efforts to get to zero CLABSI, we're putting patients at higher risk for other complications. Perhaps a more nuanced target of "any complications per catheter inserted" could replace CLABSI as a quality metric?

**Edit: Mike kindly informed me that hospitals are also penalized for iatrogenic pneumothorax through an AHRQ patient safety indicator, which is included in the CMS Hospital-Acquired Condition Reduction Program. Thus, selecting a subclavian site could lower your CLABSI penalty but raise your pneumothorax penalty. Of course!