Can neck manipulation (by chiropractors or by other practitioners) cause strokes? Many of us think it can, but definitive proof is lacking. A recently published study looked at the available evidence. A systematic review found a small association between stroke and chiropractic care but concluded that the association was spurious and that there was no evidence for causation. My colleagues and I have written about this subject several times; some of the links are listed here. I certainly agree that there is no definitive evidence for causation, but I think there is evidence to support a strong enough probability of causation to constitute a good reason to avoid neck manipulation.
The systematic review
They combed databases and two study authors independently reviewed all articles that were in English and that involved patients with carotid or vertebrobasilar artery (VBA) dissection and recent chiropractic manipulation. The studies were graded using the classification of evidence scheme adopted by the American Academy of Neurology. They found a total of six case control studies, two of which were judged to be class II (of moderate quality and risk of bias), and four as class III (a high risk of bias; class I would indicate high quality and low risk of bias, and class IV would have a very high risk of bias).
Class II studies
Smith et al. compared 51 VBA stroke cases to 100 controls, assessing exposure to spinal manipulative therapy (SMT) by mail survey. VBA stroke patients were 6.6 times more likely than controls to have had chiropractic care in the 30 days before the stroke, even after adjusting for neck pain or headache. There were limitations to the study: possible recall or survivor bias, reason for seeing chiropractor was not assessed, and significant variability among diagnostic procedures that might have resulted in interviewer bias.
Dittrich et al. compared 47 patients with cervical artery dissection to a control group of stroke patients with etiologies other than dissection, using face-to-face interviews with blinding. They found no association between any individual risk factor and dissection but cumulative analysis of all mechanical risk factors in the previous 24 hours did show an association (p=.01). The study was presumably too small to have found an association with individual mechanical risk factors like SMT. This study was also subject to recall bias.
Class III studies
Cassidy et al. identified 818 patients hospitalized for VBA strokes in Ontario during the years 1993-2002, matched them to controls, and relied on health billing records to determine whether patients had seen a chiropractor or a primary care provider (PCP) in the previous year. They found that visits to a chiropractor were not more likely than visits to a PCP, but that visits to either for headache or neck pain were more likely in patients subsequently diagnosed with a VBA stroke. This study had limitations, most significantly that the records did not reveal whether a patient had had neck manipulation.
Rothwell et al. examined some of the same cases as the Cassidy study. They found that patients under the age of 45 were 5 times more likely to have seen a chiropractor in the previous week; they found no association with older patients.
Thomas et al. compared the records of 47 patients with neck artery dissections to 43 controls. They found a signification association with neck manipulation (odds ratio 1.67, confidence interval 1.43-112.0). There were flaws in this study including a lack of blinding.
Engelter et al. identified 966 patients with cervical artery dissections, 651 with strokes from other causes, and 280 healthy controls. Cervical manipulation was 12.1 times as likely for strokes due to dissections as for strokes due to other causes. Data were collected by questionnaires, and there may have been recall bias.
Since the patients in the Rothwell study were a subset of those in the Cassidy study, they did not include it in their analysis, but analyzed only the remaining 5 studies. They gave the Cassidy study a weight of 73.9% allowing them to calculate an overall odds ratio of 1.74. I don’t understand why they gave that study so much weight. If any statisticians are reading this, perhaps they can help me out. Did they adjust it to reflect variance? The Cassidy study had a much narrower 95% confidence interval than the other studies, but it overlapped the confidence interval of only one other study, and its data didn’t even include information about manipulation. Look at the Forest plot they published:
Detail of the Forest plot; click to embiggen and reveal associated table.
The size of the squares indicates the weight given to the data, the lines represent the 95% confidence intervals, and the diamond represents the overall odds ratio for an association between chiropractic treatment and stroke.
Notice that they list “manipulation” and “no manipulation” events for each study including the Cassidy study, which is deceptive because that study did not collect data on manipulation, only on visits to providers. You can see at a glance that if they had not given the Cassidy study so much weight, the overall odds ratio would have been significantly higher.
Mark Crislip’s critique of the Cassidy study
Mark Crislip wrote an analysis of the Cassidy study. He pointed out that the records listed the diagnosis of VBA stroke but not the reason for the stroke, which could have been a thrombus, an embolus, or a tear in the lining of the artery (dissection). In the young, where stroke is rare, the cause is more likely to be a dissection, which can be precipitated by trauma. The worry in chiropractic is a stroke caused by a dissection. The young are more likely to get a thorough workup and a precise diagnosis, whereas the diagnosis might be over-represented in older patients with lots of risk factors for stroke. In the over-45 group, a rare dissection due to chiropractic manipulation might well be lost in the sea of strokes due to other causes.
For patients under 45 who visited a chiropractor, there was a strong association with stroke in the next 24 hours; the odds ratio was 12. They diluted this finding by looking at the entire month following the visit.
Then they tried to explain it all away by saying patients with headache and neck pain may have already been having a stroke in progress when they sought care from a chiropractor or a PCP. But the recorded diagnostic codes didn’t bear that out. There was very little overlap between the recorded reasons for the visit and the typical symptoms of a VBA stroke. Mark concluded that given the number of chiropractor visits and the rarity of vertebral artery tear, chiropractic manipulation is probably a rare cause of a rare event. He compared the suggestion that the VBA occurred before the manipulation to suggesting that a hanging victim already had a broken neck when he mounted the scaffold.
Flaws in the study’s Discussion section
Here’s where it really becomes problematic. They report that they found a small association between chiropractic care and cervical artery dissection, but then they discount their own finding and try to rationalize it away. They say the evidence is low quality; I agree. They say they found no evidence of causation; I agree. But then they try to say there is no convincing evidence of even the association that they themselves found.
They found no evidence for causation. On the other hand, they found no evidence against it.
They say they are concerned that a false belief in a causal connection “may have significant adverse effects such as numerous episodes of litigation.” On the other hand, a lack of belief might prevent justified litigation where patients were harmed or killed.
They say neck pain and headache are confounders. They say patients with headache and neck pain more frequently visit chiropractors, and patients with cervical artery dissection more frequently have neck pain and headache, so the appearance of more cervical artery dissections after visits to chiropractors is spurious. They say the Cassidy study showed that visits to a primary care provider or a chiropractor were likely to be due to pain from an existing dissection. It did not. There is absolutely no evidence to support that speculation.
In fact, there are numerous “smoking gun” cases where patients consulted a chiropractor not for neck pain or headache, but for pain in parts of the body below the neck, such as shoulder pain or tailbone pain, and developed stroke symptoms on the chiropractor’s table at the time of neck manipulation. Sandra Nette had no pain at all; she felt fine and let the chiropractor manipulate her neck only because she falsely believed that regular maintenance adjustments were an effective means of keeping her healthy.
They fail to even mention the smoking gun cases or the evidence that the incidence of stroke rises with the proximity to the time of manipulation.
They make a big deal of Hill’s criteria for assigning causation to association. With a rare condition like VBA dissection, and with the characteristics of this condition, it would be very difficult to fulfill all of Hill’s criteria. We may never see that kind of proof, so we must rely on lesser quality evidence.
They cite cadaver studies to claim that SMT doesn’t place significant strain on the vertebral artery. And yet we know that very small strains can cause strokes in susceptible live patients. “Shampoo strokes” have been reported after hyperextension of the neck at beauty parlors.
Other considerations
The indications for neck manipulation are questionable. Upper cervical (NUCCA) chiropractors do neck manipulations on every patient, no matter what the complaint. Some chiropractors do neck manipulations for somatovisceral conditions rather than just for musculoskeletal conditions. Whatever the indications, chiropractors certainly have no business doing neck manipulations on a patient with an existing dissection, and they have not shown that they can reliably diagnose a stroke in progress. There are no tests to reliably identify patients at risk of dissection.
Underreporting is a problem. If a dissection is temporarily sealed by a clot that breaks loose several hours or days later, the connection with manipulation may be missed. Patients may never return to the chiropractor. I heard of at least one case where a patient developed stroke symptoms immediately following manipulation, was hospitalized for a disastrous stroke, and never let the chiropractor know what had happened.
Edzard Ernst weighs in
Edzard Ernst reported on a case of a man who had a stroke following chiropractic manipulation for chronic neck pain. Chronic, not a new symptom suggesting a stroke in progress. He also reported on a case of phrenic nerve injury from neck manipulation.
Ernst commented:
- There is no effective monitoring scheme to adequately record serious side-effects of chiropractic care.
- Therefore the incidence figures of such catastrophic events are currently still anyone’s guess.
- Publications by chiropractic interest groups seemingly denying this point are all fatally flawed.
- It is not far-fetched to fear that under-reporting of serious complications is huge.
- The reliable evidence fails to demonstrate that neck manipulations generate more good than harm.
- Until sound evidence is available, the precautionary principle leads most critical thinkers to conclude that neck manipulations have no place in routine health care.
The American Heart Association and American Stroke Association agree. They were concerned enough about the apparent association to have issued a joint scientific statement warning about it and recommending that patients be informed of the possible risk prior to manipulation.
Bottom line: A double standard
Imagine a conventional therapy about which the current Cochrane review says that it has no proven effect for the condition in question. Imagine further that this therapy causes mild to moderate adverse effects in about 50% of all patients in addition to very dramatic complications which are probably rare but, as no monitoring system exists, of unknown frequency…
Now I ask you to imagine that there is a pharmaceutical drug that fits this description. Imagine that there are the same numbers of studies showing an association of that drug with a deadly side effect like stroke or death. The FDA would pull it off the market; they wouldn’t wait for definite evidence of causation that fulfilled all of Hill’s criteria. And I think the people who are making excuses for neck manipulation would want them to take that drug off the market. I don’t think they would want to take such a drug.
I wouldn’t risk taking a drug like that, and I wouldn’t risk neck manipulation.
This article was originally published in the Science-Based Medicine Blog.
