ICU and You · Procedures · Number 6
The sequence first, then everything that decides whether it goes well — and a one-page checklist for the trolley
This is the line your trainees judge themselves by, and the one most likely to hurt somebody while you are still learning it. The good news is that the technique is short and teachable. The bad news is that almost all of the serious harm comes from two or three steps that take seconds, and from one decision made after things have already gone wrong.
So this piece opens with the sequence. Indications, consent, contraindications and complications follow, because you will look those up before a list and the sequence before a line. There is also a one-page checklist meant to be printed and put where the trolley lives.
What follows assumes a right internal jugular in an adult, with real-time ultrasound, which is the default and should be.
Scan first, undraped. Look at both sides before you commit. You are checking the vein is there, that it is patent and compressible, where the carotid actually sits in this patient, and whether the other side is better. This takes twenty seconds and changes the plan often enough to be worth it every time.
Then set up as though it will go badly. Monitoring on and visible. An assistant who is staying. Everything opened before you scrub, because the thing you discover you need at the wrong moment is the thing that breaks asepsis. Flush and check every lumen, and know which one the wire exits.
Blood colour and the absence of pulsatility are not confirmation. The American Society of Anesthesiologists says so in terms: they should not be relied upon. In a pooled series of eighty arterial injuries, pulsatile flow identified twelve of them, a blood gas six, and a transduced waveform four; in thirty-seven the method was not even recorded. A hypoxic, hypotensive patient has dark arterial blood, and a flat-looking column can still be an artery.
Use two confirmations, at least one of them objective, before the dilator:
See the wire in the vein. Ultrasound, in both axes, before you dilate. It is easier to see
than the needle, and it is the single most useful habit on this page.
Or transduce it, or run a manometry column — a length of sterile tubing held up and
allowed to fill. Venous blood sits; arterial blood climbs and keeps climbing. If you transduce,
check the monitor is on an arterial scale, or you will convince yourself with a flat trace.
And know the limit of your tools: ultrasound does not always exclude an arterial placement. There is a reported case of a needle passing through the back wall of the vein into the carotid under real-time vision. That is an argument for two checks, not for abandoning the first.
The effect is large and it is not in dispute. Against landmarks, for internal jugular lines, real-time ultrasound reduces inadvertent arterial puncture by around seventy per cent, reduces haematoma similarly, raises first-attempt success by about half, and takes roughly one attempt off the average. Those are the Cochrane figures, from thirty-five studies and five thousand patients.
Two honesties to carry with them. The quality of that evidence is graded low to very low for most outcomes, because the trials are old, unblindable and heterogeneous. And the comparison is against landmarks, not against a careful operator using ultrasound properly: nothing in that body of work tells you how much of the benefit survives a probe held badly.
Marking the skin beforehand is not ultrasound guidance. Static pre-procedural marking is explicitly advised against as a substitute — the neck is not the same shape once you have draped it, turned the head and tilted the bed. Scan beforehand to plan; scan in real time to puncture.
No Australian college mandates it. You will hear that one does. The strong wording is British — NICE calls two-dimensional ultrasound the preferred method and the Association of Anaesthetists says it should be used routinely. Safer Care Victoria says “whenever possible” and labels that expert opinion. What binds you is your own service's credentialling, and the fact that the evidence is one-directional.
Oblique and long axis are not proven better. The pooled adult analysis found no difference between axes for success or complications. One good trial found less posterior wall puncture in long and oblique axis — four and seven per cent against fifteen — which is a reason to know them, not a reason to claim superiority.
There is no graded list of indications anywhere, which is itself worth knowing. The defensible ones are drugs that cannot go peripherally for long — vasoactives, concentrated potassium, central parenteral nutrition — renal replacement, transvenous pacing, plasmapheresis, central venous oxygen saturation sampling, and a patient in whom no other access can be obtained.
To measure the central venous pressure. The pooled correlation between CVP and the change in stroke volume after a fluid bolus is 0.18, and the area under the curve for predicting fluid responsiveness is 0.56 — which is a coin toss with a decimal point. CVP has narrow uses as a trend, and as the downstream term in a perfusion pressure, but it is not a reason to put a line in somebody.
“Needs intravenous antibiotics.” No. The guidance runs the other way: assess daily whether the access is still needed, and take it out when it is not.
“Difficult peripheral access”, on its own, and before anyone has tried an ultrasound-guided peripheral cannula or a midline. Often the honest translation is that three people have failed with a landmark technique.
“Needs noradrenaline” — which is a reason, but not an emergency one. Short peripheral infusions at or above the antecubital fossa are unlikely to cause tissue injury; the great majority of reported extravasation injury was from cannulae distal to the elbow or knee. Peripheral noradrenaline is a legitimate bridge to a line placed properly, in daylight, by someone who is not rushing.
The honest summary of the site argument is that it is closer than the teaching suggests. In the largest randomised comparison — three thousand patients, three and a half thousand catheters — the composite of bloodstream infection and symptomatic thrombosis favoured the subclavian: 1.5 events per thousand catheter-days against 3.6 for jugular and 4.6 for femoral.
But subclavian bought that with mechanical harm. Pneumothorax was three times commoner, 1.5 per cent against 0.5, and the subclavian arm had the most mechanical complications overall. Put the two together — infection and thrombosis plus serious mechanical complications — and the totals were 3.1 per cent subclavian, 3.7 jugular, 3.3 femoral. There is no winner in that row. Ultrasound use was not randomised, which flatters the subclavian further.
So: the right internal jugular is the sensible default for most intensive care patients, because it is the site ultrasound helps most, it has a straight run to the right atrium, and its failure modes are mostly recoverable. The left is a second choice — two near-right-angle turns, the thoracic duct entering at the jugulo-subclavian junction, and a carotid that sits in the high-risk position more often than on the right.
If the patient has advanced kidney disease or may need dialysis, avoid the subclavian. Subclavian stenosis destroys the arm you were going to make a fistula in. This is not a conflict between guidelines, although it is often taught as one — it is written inside the infection guidance itself, and at a higher grade of evidence than the subclavian preference it qualifies. Read the two recommendations together and the answer is already there.
The genuinely absolute ones are short: refusal by a patient with capacity, and infection or burn over the puncture site. Everything else is relative and most of it is modified by looking first — thrombosed or absent vein, previous surgery or radiotherapy, a tracheostomy that makes the field filthy, a contralateral pneumonectomy that means you cannot afford a pneumothorax, raised intracranial pressure that makes head-down unwise, a patient who will not keep still.
INR first, where the teaching is right. The INR does not predict procedural bleeding, plasma does not meaningfully correct a mildly raised one, and venous access is classed as a low-bleeding-risk procedure. The British haematology guidance withdrew its recommendation to routinely check coagulation before moderate-risk procedures in 2023. Observational series of thousands of ultrasound-guided lines show no excess bleeding with abnormal coagulation until the INR is above five.
Platelets, where it is not. Everyone now says you do not need to transfuse before a central line. The only randomised trial in this space says otherwise. In patients with platelets between 10 and 50, withholding a unit before insertion produced catheter-related bleeding in 11.9 per cent against 4.8 per cent — a relative risk of 2.45, and non-inferiority was not met. Most of that bleeding was moderate rather than dangerous, and the trial was open-label, but the result is the result. The British threshold for a central line remains a platelet count of 20.
Which leaves a defensible position rather than a rule: ultrasound-guided insertion is safe in coagulopathy and is not a reason to delay a line that is needed; and if the count is under 50 and the procedure is elective, the transfusion is more justifiable than you were taught.
Most of these lines go into people who cannot consent, which does not make consent irrelevant — it changes who gives it and what you write down.
If the patient has capacity, the risks worth naming are bleeding, arterial puncture, pneumothorax, infection, thrombosis, and failure requiring another site. New South Wales does not list central lines in either its written-consent category or its minor-procedure category, which leaves a judgement: a line carries known significant risks, so written consent is the better reading where the patient has capacity and circumstances allow. Check your own district's policy, which may settle it.
If they do not, the Guardianship Act 1987 sets a hierarchy of person responsible — guardian with the relevant authority, then spouse or de facto partner, then an unpaid carer, then a close unpaid friend or relative. If somebody at one level is absent, unwilling or lacks capacity themselves, you record that and move down. Where there is a guardian with authority to consent to treatment, only the guardian can.
In an emergency, section 37 lets you proceed without consent where treatment is urgently necessary to save life, prevent serious damage to health, or relieve significant pain or distress. It does not override a valid advance care directive. Write in the notes that you relied on it and why — the documentation is the safeguard here, and it is the part most often skipped at three in the morning.
Modern pooled figures, from a 2024 meta-analysis of central lines generally: arterial puncture about 16 per thousand catheters, pneumothorax about 4 per thousand, placement failure about 20 per thousand. Infection runs at roughly 4.8 per thousand catheter-days and deep vein thrombosis 2.7. Ultrasound cuts arterial puncture to a fifth and pneumothorax to a quarter.
This is the most important paragraph on the page. If a dilator or a catheter of six French or larger is in the carotid, leave it exactly where it is and call vascular surgery or interventional radiology now.
The instinct is to remove it and press, and the instinct is wrong. In one review, complications followed in 47 per cent of patients managed by removal and pressure, against none of those repaired surgically. In a pooled series of eighty cases, removal and compression succeeded in 5.6 per cent against 94.6 per cent for endovascular repair and 100 per cent for open surgery; seventeen of eighteen compression cases had complications, including three deaths. A carotid is not compressible in the way the word implies, and a haematoma in that space closes an airway.
A needle puncture is different — remove it, press properly, and then watch, scan and document.
Pneumothorax is uncommon from a jugular and should still be considered when the patient deteriorates afterwards. Air embolism needs 200 to 300 millilitres, or three to five per kilogram, to be lethal in an adult — but an open 14-gauge hub can entrain around a hundred millilitres a second, so the volume is reached faster than the number suggests. Occlude the entry point, give oxygen at one hundred per cent, aspirate from the line if the tip is in the atrium, support the circulation, and start compressions even without arrest, which fragments the air-lock. The left lateral head-down position everybody teaches rests on a dog experiment from 1947; do it if it costs nothing, but never before the oxygen.
Thrombosis is commoner than clinical impression suggests, and the jugular is the worst site for it. On daily screening, a quarter of patients developed a catheter-related thrombosis; the jugular rate was four times that of other sites, left-sided lines nearly three times right-sided — and every single one was asymptomatic. Chylothorax from thoracic duct injury on the left is rare enough that no incidence has been established; it appears as milky fluid at the site, in the line, or as a left effusion that may only declare itself once feeding starts.
Retained guidewire happens about once in three thousand procedures, is a Never Event in the United Kingdom and falls under the Australian sentinel event for a retained foreign object. Reported mortality is four to six per cent.
The useful finding is about its mechanism: it is an omission error, and it happens to careful people who are interrupted at the wrong second. Operator-focused safeguards were found to be largely ineffective, none of them completely so. What works is a forcing function — a system in which the wire must be physically accounted for before the sharps can be disposed of. Locked procedure packs built on that principle prevented retention in ten out of ten simulated forced errors against two out of ten with ordinary practice.
Until your unit has one: say the wire out loud, hold it up, and have the assistant confirm it before anything is thrown away. And if a wire has gone below skin level, suction retrieves it far more reliably than trying to grasp and pull.
The tip belongs in the lower superior vena cava or the upper right atrium, lying parallel to the vessel wall. A chest film is not mandatory — no guideline requires one, and in more than thirteen hundred ultrasound-guided right jugular insertions, malposition needing adjustment ran at one per cent and pneumothorax at one in a thousand. Order one when something was difficult, when the left side was used, or when you are not sure.
What a film cannot do is tell you the line is venous. A British safety body had to say this explicitly after a case in which the film was accepted as satisfactory and the catheter was in the carotid. Radiography is for depth and for pneumothorax. Vessel identity is settled by transduction or by ultrasound, and it is settled before you dilate, not afterwards.
How long can it stay? There is no maximum and no routine change interval. Replacing lines on a schedule is explicitly advised against, as is exchanging one over a wire to treat suspected infection, and as is removing one for fever alone. The intervention that works is the daily question of whether it is still needed — which is the single most effective infection control measure available, and the most neglected.
Chlorhexidine-impregnated dressings cut bloodstream infection by about sixty per cent in a trial of nearly two thousand patients. American guidance now recommends them for all adult short-term central lines; the Australian intensive care guideline, which dates from 2012 and has not been updated, still treats them as something to add if your rates stay high. The divergence is real and your unit has probably picked a side without telling you which. The one harm worth knowing: severe contact dermatitis in roughly one per hundred catheters, and they are not for premature neonates.
Scan both sides before you scrub. Rotate the head as little as possible. Keep the needle tip in view. See the wire in the vein before the dilator touches anything. If you have dilated an artery, leave it in and pick up the phone. And ask tomorrow whether the line is still needed.
Give one tip to somebody learning this.
Not the textbook. The thing you actually do — a way of holding the probe, a trick for the neck that will not give you a vein, a habit you adopted after something went wrong, a sign you have learned to distrust. One is plenty. Three if you are feeling generous.
I am collecting these because they are the part of the procedure that never gets written down. Everybody who has put in a few hundred of these has four or five things they do that are not in any guideline, and most of them have never told anyone, because nobody asked. The ones I would most like:
A good method — how you set up, how you hold things, what order you do it
in.
The difficult one — the short neck, the hypovolaemic collapsed vein, the
patient who cannot lie flat, the one with a tracheostomy in the way.
Something to watch out for — the near miss you still think about.
I will publish the collection. Anonymously by default unless you tell me otherwise, because the most useful tips tend to come attached to the least flattering stories.
How you are credited is entirely your call — by name, by initials, or not at all — whether you write the piece or I do, and whether I quote you or simply say who sent something in; nothing of yours goes up with your name on it unless you have said yes.
The button opens a reply straight to me. If your device blocks it, write to icuandyou@icloud.com instead.