Spinal surgery has a reputation problem, and a good deal of it is deserved. Not because the operations do not work - many of them work very well - but because the results are more variable than in most other areas of surgery, and that variability is not always explained to people before they consent to having their procedure.

It is worth setting out why lumbar spine surgery is harder to predict than a hip or a knee, what the honest numbers look like, and what a careful surgeon should be weighing up before recommending anything at all.

Start with the comparison everyone makes

Patients arrive expecting spinal surgery to be like a hip or knee replacement - operations with a deservedly good reputation, where most people expect a good result and get one.

Spine surgery is less predictable, and the reason is largely structural.

A hip replacement addresses one joint. You are replacing a single articulation with a known mechanical problem, and the pain almost always originates there. I sometimes wish I was a hip surgeon. Sometimes.

Do a fusion at one level of the lumbar spine and you are not operating on one joint. You're actually fusing three: the disc itself, and the two facet joints on either side. Each of those can generate pain independently. Each behaves differently after being fused. Perhaps even more importantly, the consequences of what you do may not necessarily be confined to the level intended.

And - the point that undoes most predictions - the level that looks worst on the scan may not be where the pain is coming from at all.

Why the lumbar spine resists prediction

Three reasons, compounding.

The cohort is entirely heterogeneous. Pretty much everyone who needs a hip replacement needs broadly the same operation for broadly the same reason. People who might benefit from lumbar spine surgery arrive with wildly different symptoms, different anatomy, different numbers of affected levels, and different amounts of prior surgery. They are not one group. Any single figure quoted for "the success rate of lumbar fusion" is averaging across people who have very little in common, which is why I am going to be careful with the numbers below.

There are multiple candidate pain generators. The disc, the facet joints, the nerve root, the muscles, the sacroiliac joint. Establishing which one is responsible is genuinely difficult, and getting it wrong means doing the correct operation on the wrong thing.

The same-sounding operation can be done several different ways. A one-level lumbar fusion can be approached from the back, from the front, from the side, a version of sideways, or a combination of the above. Over my career I have done that fusion five different ways, and each carries its own risk profile, its own recovery, and its own likelihood of success in a given patient. "You need a fusion" is therefore not one recommendation. It is a family of them.

The honest arithmetic

Here is where I have to be more careful than surgeons usually are, because the number I used to quote from memory - about three in four do well - turns out to be defensible only when you say "what exactly do you mean by well".

Take one of the better-studied indications: degenerative spondylolisthesis at L4/5, where one vertebra has slipped on the one below. Five years after a posterior lumbar interbody fusion, in a series of just over a hundred patients, 78% were satisfied with their surgery. Eighty-eight per cent said they had improved. Seventy-four per cent would recommend it. Seventy-one per cent would go through it again.

Those are good numbers, not perfect, but good. But notice they are not the same number - what you ask determines what you get told. The differences are subtle, but who am I to define what's important to you?

Now the finding that I think matters more than any of them.

A study of patients having lumbar fusion sorted them before surgery by how much improvement they expected. Although the objective results across the groups were similar, satisfaction was not:

  • Of those with high expectations, 43% were satisfied
  • Of those with moderate expectations, 71% were satisfied
  • Of those with low expectations, 85% were satisfied

Same operation. Comparable measured improvement. Satisfaction ranging from four in ten, to more than eight in ten - determined almost entirely by what people had been led to expect.

That is not a surgical result. It is a communication result, and it is the single strongest argument I know for spending longer in the consulting room than in the operating theatre.

There is a temptation to read that as "manage expectations downward", and that is not what it says either. Statistically, the high-expectation group started with worse disability and hoped for more improvement, which is entirely understandable. The goal is not to talk people down. It is to make sure that what the operation can realistically deliver and what the patient is hoping for are the same conversation.

One more thing belongs in the same honest reckoning. Fusion surgery can be a morbid procedure - meaning it can cause significant complications of its own. It doesn't have to be, but it can be, and a patient weighing it up is entitled to hear that stated plainly rather than discovered afterwards.

What the guidelines actually say, which surprises people

For low back pain itself - back pain without a clear structural cause - the UK's national guidance is blunt: do not offer spinal fusion, except as part of a randomised controlled trial.

That reads as a startling sentence coming from a spinal surgeon's website, so let me be precise about its scope.

That recommendation is about non-specific low back pain - pain without an identified structural explanation. It is not a recommendation against fusion for degenerative spondylolisthesis, for instability, for deformity, or for a nerve compression that fusion is needed to address. Those are different indications with different evidence, and the 78% figure above comes from exactly such an indication.

But it does mean this: if the reason offered for fusing your spine is "your back hurts and the scan shows wear and tear changes", that is the indication the guidance is warning about. The operation is not the problem. Using it for a problem that is almost certainly never going to be solved by a fusion is.

For sciatica, the same guidance takes a different view, and it is worth knowing the sequence: consider decompression when non-surgical treatment has not improved pain or function, and the imaging findings match the symptoms. It's not a guarantee, for the reasons I outlined earlier, but it is honest and the best we can offer at this point in time. In the end no one needs a fusion, but some people may benefit greatly from one.

Kicking the can down the road

There is one consequence of fusion that deserves more airtime than it gets.

Fusing a segment stops it moving. The load it used to carry has to go somewhere, and it goes to the segments immediately above and below (and may have more distant consequences also). Those neighbouring levels are, almost by definition, already worn - people needing a fusion for degenerative reasons rarely have one bad level in an otherwise pristine spine.

So, you fix one level, which deals with one pain generator, but in doing so you put additional stress on a level that was already on its way out.

The numbers here need splitting in two, and this is exactly the same lesson as reading any spinal scan.

Changes visible on imaging at the neighbouring level are common. Pooled across studies, they accumulate at roughly 6% per year, and around a third of fusion patients show them. But they correlate poorly with how people actually feel - which is the recurring theme of this whole subject.

Changes that actually cause trouble are considerably less common. Symptomatic adjacent segment disease accumulates at roughly 2% per year. Over a decade that is something like one person in six; over longer horizons it climbs further, which is where the figure of one in five to one in four over the long term comes from. Around 7 to 11% of fusion patients develop it, depending on the series and the follow-up, and roughly 7% end up having further surgery at the adjacent level.

So: a real risk, meaningfully lower than the imaging might suggest, and it accumulates rather than striking early (though for an unfortunate small number it may). That means a fusion decision is a decision about the next twenty years, not the next twelve months, and it should be weighed openly on that basis. It is not an argument against fusion. For some people the long-term arithmetic still clearly favours doing it. But for some it's a bumpy road.

What should be weighed before recommending anything

Whether there is a clear structural problem that plausibly explains the symptoms the person actually describes - not merely an abnormality on imaging.

Whether the symptoms are the kind surgery reliably helps. Leg pain from a compressed nerve responds better than diffuse back pain without a clear source. That distinction is fundamental in understanding.

Whether everything reasonable has been tried first - and here I want to correct something I would have said less carefully a few years ago.

In terms of isolated back pain - time, staying active and physiotherapy are genuinely first-line. Prolonged bed rest is not: Australia's national standard for low back pain states plainly that it is harmful and should be discouraged. The evidence that staying active beats resting is real but modest, noting however that for sciatica specifically, the difference between the two is small to negligible, and should be thought of in a different context, i.e. sciatica - caused by irritation of a nerve root - is a different pathological process and warrants different management from non-specific back pain.

Injections deserve more scepticism than they usually get, including from me. For back pain without leg pain, the national guidance is not to offer spinal injections at all, and the evidence for injecting facet joints therapeutically (as distinct from it being done as a diagnostic endeavour) does not support the practice. For sciatica, the picture is better but still sobering: an epidural injection of local anaesthetic and steroid does produce a measurable reduction in leg pain and disability, but the effect is small - smaller than the threshold the reviewers themselves set for a difference patients would notice - it is confined to the first weeks to three months, and pooled across multiple randomised trials it does not change whether you end up having surgery. Guidance says "consider" it in acute, severe sciatica. This fits with my practice philosophy. Think of an injection as a "bridge" to get you where you need to go - for patients whose leg pain was destined to resolve spontaneously (as most do), injections provide the short-term relief to help them through the worst period of pain. For patients with intractable sciatica, who ultimately do require surgical decompression, targeted nerve root injections based off clinical findings combined with contemporaneous imaging can be a useful diagnostic intervention that may help surgical decision making, in addition to also providing a degree of pain relief until something definitive can be done.

I raise this not to talk anyone out of an injection, but because "we will try an injection first" is often offered as the conservative, cautious step. It is reasonable as short-term symptom relief in the right patient. It is not a treatment that changes the trajectory, and it should not be presented as one. The only caveat I would add here is that for patients who are poor surgical candidates due to other factors (e.g. advanced age), periodic injections can sometimes be very helpful.

Whether the smallest sufficient operation would do. Around two-thirds of the degenerative spinal work I do is decompression alone - taking the pressure off the nerve without fusing anything. Only about a third involved stabilisation or fusion. Decompression is the smaller operation with the shorter recovery, and where it is enough, it should be what happens. Unfortunately, it isn't what always happens, but that is a different conversation.

Whether the patient's expectations and the achievable outcome are in the same territory. Given the satisfaction figures above, this is not a soft skill. It is the intervention with the largest measured effect on whether someone is glad they had surgery.

When the answer is no

Sometimes the correct recommendation is not to operate. Not "not yet", not "let us try something else first" - simply no.

That is harder to say than it sounds. A patient in significant pain has come specifically to see a surgeon, often after a long wait, frequently having been told an operation is the answer. Saying that surgery is unlikely to help them is unwelcome, it takes longer than agreeing would, and it can look like a refusal to help.

But an operation performed for the wrong indication carries every one of the risks and very little of the benefit. Declining to do it is not withholding treatment. It is the treatment decision. This requires qualification, however. Just because I say no - based on what I think is best for the patient before me - doesn't mean another, similarly good-intentioned surgeon, will agree. And vice-versa. I have operated on patients that others had said no to, with genuinely excellent results. And vice-versa.

I have seen patients who have undergone fusions performed by other surgeons for reasons that I could not in good conscience advocate for myself, yet I am also certain that other surgeons will have seen some of my patients and felt the same way. A detailed and honest examination of why that may be the case requires an appreciation of the nuances of this area of practice that I would struggle to articulate in a succinct and informative way for a lay audience. What I can say is that with experience and self-reflection, one can at least come to understand that this area of medicine requires a good deal more perspicacity and humility than perhaps a good majority of others.

It is worth knowing that this is a recognised problem at a population level, rather than just a personal opinion. The 2018 Lancet series on low back pain concluded that globally there is limited use of recommended first-line treatments alongside inappropriately high use of imaging, rest, opioids, spinal injections and surgery. Closer to home, the Australian Atlas of Healthcare Variation has found marked variation in rates of lumbar spinal surgery across the country. Variation of that kind is not explained by disease. In spine surgery it seems that the old adage "there's more than one way to skin a cat" would be a resounding understatement.

As strongly as I feel that a patient should get the operation they need, not just the one the surgeon can perform (a separate topic altogether...) - the measure of a spinal surgeon is not the range of operations they can perform, but whether the operation they recommend is the one the patient in front of them actually needs.

This article is general information about surgical decision-making. It is not advice about your own care, which needs your history, your imaging and a conversation with your own surgeon.

Principal sources: Okuda et al., Spine 2016 (satisfaction after fusion for L4 degenerative spondylolisthesis); Zhang et al., The Spine Journal 2024 (expectations and satisfaction); Zhang et al., Clinical Spine Surgery 2016 and Donnally et al., The Spine Journal 2020 (adjacent segment degeneration and disease); NICE guideline NG59, Low back pain and sciatica in over 16s; Australian Commission on Safety and Quality in Health Care, Low Back Pain Clinical Care Standard 2022; Chou et al., Annals of Internal Medicine 2015 and Oliveira et al., Cochrane 2020 (epidural steroid injections); Dahm et al., Cochrane 2010 (activity versus bed rest); Foster et al., Lancet 2018.