How safe is thoracentesis? Giant case series sheds light

Thoracentesis for pleural effusion — that is, inserting a long needle between someone's ribs to drain a fluid collection from the chest — has always come with a scary menu of potential risks, including pneumothorax, hemothorax and pulmonary edema. A new study reports a low complication rate from thousands of thoracenteses. But debate will remain whether the safety results -- achieved by a high-volume, highly specialized hospital-based team -- can be duplicated by less expert practitioners. Thoracentesis is performed over 170,000 times each year in the U.S. -- so it's surprising that there is little data on its outcomes or risks. The best previous estimate of pneumothorax risk is 6% (<2% with ultrasound guidance), based on a 2010 meta-analysis including 6,600 thoracenteses. Absent robust evidence, British Thoracic Society 2010 guidelines for thoracentesis advise:
Don’t withdraw more than 1,500 mL at once (you might precipitate re-expansion pulmonary edema or REPE);
Give fresh frozen plasma or platelets to patients with elevated INR or thrombocytopenia, to prevent bleeding.
Extreme caution with thoracentesis on patients receiving mechanical ventilation has also been advised (you could cause a fatal tension pneumothorax). But all this is largely based on consensus and tradition, as there have been almost no large-scale studies on the risks of thoracentesis. There is one, now: between 2001 and 2013, Dr. Mark Ault and his team performed 9,320 thoracenteses on thousands of patients at Cedars-Sinai Medical Center in Los Angeles. (Yes, that’s a lot of thoracenteses -- an average of about 2 a day, every day for 13 years.) In the early 2000s, Cedars-Sinai invested in dedicated proceduralist services for various procedures including thoracentesis and central lines, and these teams handle most of the load for the medical center. Dr. Ault et al reported their collected results in Thorax. Some of their methods contradicted usual practice and the BTS guidelines:
No blood products for coagulopathy or platelets for thrombocytopenia were provided to reduce thoracentesis bleeding risk.
No routine chest films after thoracentesis.
Bilateral thoracenteses were performed sequentially (without delay) about 40% of the time.
Patients on ventilators were not routinely excluded.
Ultrasound and hand suction (not vacuum bottles) were used routinely; pleural manometry was not.
Authors tracked patient factors (e.g., on a ventilator), procedure factors (needle passes, volume removed), and complication rates, which were 0.98% overall, including:
Iatrogenic pneumothoraces (0.6%)
Re-expansion pulmonary edema or REPE (0.01%), with no deaths
Bleeding episodes (0.18%), of which 5 were considered hemothorax (0.05%)
Vasovagal reactions (0.06%)
Removing >1,500 mL was associated with more complications (3.1% overall), including 2.2% pneumothorax risk and 0.75% risk for re-expansion pulmonary edema. Severe thrombocytopenia did not increase bleeding risk: among 53 patients with platelets under 20,000, there were no cases of hemorrhage. There was no association between INR and bleeding risk, among 301 patients undergoing thoracentesis with INR >3.1. Interestingly, elevated PTT was associated with complications (authors suspect as a marker for underlying disease or frailty). Invasive or non-invasive mechanical ventilation did not increase the likelihood of a complication, among 1,377 patients. Unfortunately, the analysis was not stratified between invasive and non-invasive ventilation. Two or more needle passes were associated with increased risk for pneumothorax. In an indication of their high proficiency, authors completed thoracentesis with one needle pass >99% of the time. They reported “dry taps” (zero fluid) in only 35 patients (0.4%). Authors believe they probably over-counted, rather than under-counted pneumothoraces (i.e. counting some cases of trapped lung as pneumothorax).
So How Safe Is Thoracentesis, Really?
Ironically, the sheer scale of this study may limit some observers' confidence in its applicability outside Dr. Ault and team's hyper-expert hands. Most thoracenteses are done by physicians or midlevels who do one a week or fewer, not 2 a day. A study showing the real-world complication rates among a large pooled group of these slightly more error-prone practitioners would be more informative. The 1,500 mL rule-of-thumb remains a bit of an enigma, as well. Dr. Ault's team's complication rate tripled with larger effusions drained, which they hypothesized may be due to a larger disease burden. But the overall risk was still low -- and should be balanced against the risk/harm from leaving an excess amount of fluid in the thorax, or of performing a second procedure. In all, it seems reasonable to counsel patients on a ~2% risk for pneumothorax by experienced practitioners using ultrasound for effusions <1,500 mL, based on this study and the meta-analysis by Gordon et al. (This may be on the high side, but it shouldn't be low.) Risks are likely higher when draining larger effusions. Bleeding risks and pulmonary edema seem low (<1%), based on Dr. Ault's data in patients at high bleeding risk. Other authors have already advised not using fresh frozen plasma to reverse INR for thoracentesis. But the most applicable results are your own, of course. Why don't we all start recording our thoracentesis complication rates and that of our practice groups, and reporting them to our patients? (Saying "I've done thousands of these" doesn't count.)

Clinical Takeaway: How safe is thoracentesis? As with any procedure or surgery, the answer mostly depends on how many you've done.
Read more:
Ault MJ et al. Thoracentesis outcomes: a 12-year experience. Thorax 2015;70:127-132. Gordon CE et al. Pneumothorax following thoracentesis: a systematic review and meta-analysis. Arch Intern Med. 2010 Feb 22;170(4):332-9.
Pleural procedures and thoracic ultrasound: British Thoracic Society pleural disease guideline 2010. Thorax 2010;65:i61-i76.