J Manipulative Physiol Ther 2004 (Mar); 27 (3): 197–210 ~ FULL TEXT
Drew Oliphant, DC
OBJECTIVE: To provide a qualitative systematic review of the risk of spinal manipulation in the treatment of lumbar disk herniations (LDH) and to estimate the risk of spinal manipulation causing a severe adverse reaction in a patient presenting with LDH.
DATA SOURCES: Relevant case reports, review articles, surveys, and investigations regarding treatment of lumbar disk herniations with spinal manipulation and adverse effects and associated risks were found with a search of the literature.
DATA SYNTHESIS: Prospective/retrospective studies and review papers were graded according to quality, and results and conclusions were tabulated. From the data published, an estimate of the risk of spinal manipulation causing a clinically worsened disk herniation or cauda equina syndrome (CES) in patients presenting with LDH was calculated. This was compared with estimates of the safety of nonsteroidal anti-inflammatory drugs (NSAIDs) and surgery in the treatment of LDH.
RESULTS: An estimate of the risk of spinal manipulation causing a clinically worsened disk herniation or CES in a patient presenting with LDH is calculated from published data to be less than 1 in 3.7 million.
CONCLUSIONS: The apparent safety of spinal manipulation, especially when compared with other "medically accepted" treatments for LDH, should stimulate its use in the conservative treatment plan of LDH.
From the FULL TEXT Article:
Many authors recommend the use of spinal manipulation in the treatment of LDH, [2, 6, 8, 14, 16, 21–26, 54] while some recommend against it, [5, 31, 49] and disk herniation is the most common claim against chiropractors.  Therefore, it is important to estimate the risk of serious complications of spinal manipulation in the treatment of LDH using the best available evidence. The serious complications that spinal manipulation could cause in a patient with LDH are a significantly worsened LDH or a CES.
By attributing all of the reported cases of LDH and CES following spinal manipulation to patients presenting with LDH, an estimate of the risk of manipulation causing a clinically worsened disk herniation or CES in a patient with LDH can be calculated. The risk is less than 1 in 3.7 million manipulations, according to calculations using published estimates of the number of patients attending chiropractors, the percentage of those with disk herniations, the number of those patients receiving spinal manipulation, and the number of cases of CES or disk herniation following spinal manipulation. This does not, of course, include the kind of temporary or mild symptoms that are commonly reported following manipulation [45, 48] and placebo treatments.  If complications following MUA are included, the risk doubles (Table 8).
The numbers that calculations have been based on can be argued to be rough estimates at best, and therefore with each calculation, the accuracy of this risk estimate may have been reduced. However, there has been an increased emphasis on evidence-based care. This risk was calculated according to the best evidence available, and the numbers used err in favor of overestimating the risk. Specifically, the lowest estimate of 2.2% of patients attending a chiropractor having LDH was used (Table 6) and not the estimate of 40% of low back pain patients having pain due to internal disk derangement,  which may be a more accurate estimate of those at risk and would have lowered the estimate of risk of spinal manipulation causing LDH or CES in a patient presenting with discogenic pain to one in 46 million (6245 M × 40% × 68% × 84%/31 = 1 in 46 M) (Table 6, Table 7, and 8). The estimate of chiropractic visits in the United States over the last 40 years was used to represent worldwide spinal manipulation by all disciplines since 1911 (the date of the first published case of CES). An average of only one manipulation per treatment was also assumed, which is not normally the case. 
Finally, the results were compared with another calculation based on estimates from the literature. An estimate of one case of CES in 286 million office manipulations derived from Haldeman and Rubenstein  was multiplied by 2.2%, the number of patients who likely presented with a disk herniation,  and this suggests a risk of one CES in 6.2 million manipulations. Multiplying this by 13/31 to account for the other disk-related complications, as well as CES (Table 6), gives a risk estimate of one in 2.6 million, which is similar to the estimate presented above.
After the number of patients presenting with disk herniation, the biggest source of error may be the number of LDH or CES which have occurred following spinal manipulation, [30, 33] as detractors state that many cases of disk injury following spinal manipulation may go unreported, [28, 30, 31, 49] but this has not been verified.
The safety of spinal manipulation in the treatment of lumbar disk herniations (LDH) should be compared with other commonly accepted treatments for the same condition. Comparing this, or any treatment, with “no treatment” is not meaningful, because while no treatment may be the safest option, most patients seeking relief will opt for some type of treatment, hoping it will provide more relief than nothing.
NSAIDs are the most commonly used medications in the world, and adverse events occur in 25% of patients, with significant complications occurring in 1% to 4% per year.
The major side effects include gastrointestinal (GI) ulceration and bleeding, hepatorenal dysfunction, organ failure, and skin reactions, and they may accelerate cartilage destruction. 
GI complications due to NSAIDs cause more than 100,000 hospitalizations and an estimated 16,500 deaths each year in the United States, 
and NSAID-related congestive heart failure may exceed the mortality resulting from gastrointestinal tract damage. 
Cauda equina syndrome (CES) is reported as a sequela of surgeries for LDH in 0.2% to 1%, [59, 60] while Kardaun et al  found the frequency of CES in 3289 surgically treated LDH patients was about 0.5%. The “any-complication rate” has been estimated to be 3.7% or more, including 1.5% mortality. [61, 62]. The most recent study comparing surgical with nonsurgical treatment of chronic LBP patients found 24% of the surgical group had complications, almost half being major complications, and almost 8% required reoperations. 
If “significant complications” occur in 1% to 4% of NSAID users,  in 1.5% to 12% of LDH surgeries, [12, 61, 62] and in one in 3.7 million patients receiving spinal manipulation for LDH,
then spinal manipulation is at least 37,000 to 148,000 times safer than NSAIDs and 55,500 to 444,000 times safer than surgery for the treatment of LDH.
If CES occurs in one in 3.7 million spinal manipulations for LDH and in 0.2% to 1% of surgeries, [59, 60] then CES is at least 7,400 to 37,000 times more likely to occur as a complication of surgery than of spinal manipulation.
Meanwhile, neither NSAIDs nor surgery has been proven to be more effective in the treatment of LDH than spinal manipulation. There are no data to support the premise that operative intervention will restore neurologic function more rapidly than natural history or nonoperative intervention. 
Estimates for the effectiveness of surgical procedures for disk herniation range from 30% to 96%,  but over 4 years or more, many feel the effectiveness of surgery is the same as natural history. [4, 13, 63].
Rhyne et al,  in a study of discogram-positive low back pain patients who were offered surgery but rejected it for various reasons, found their outcome over a mean follow-up of 4.9 years was comparable with, or better than, those reported for surgical treatment of this condition.
Weber  also found that between 4 and 10 years, there was no difference between those who had surgery for LDH and those who did not, although after the first year, the surgical group reported 66% “good” and 25% “fair” results, whereas the conservative group had 32% “good” and 49% “fair” results. His decade-old paper was recently criticized for not reaching current research standards, but the critics still generally support his conclusions.  A 2–year follow-up of surgical and nonsurgical treatment of chronic LBP patients found that 63% of the surgical group rated themselves as “much better” or “better” compared with 29% of the nonsurgical group and showed significantly greater improvements on the Oswestry Low Back Pain Questionnaire, Million Visual Analogue Score, General Function Score, and Zung Depression Scale.  These authors were criticized, however, for not specifying or standardizing the nonsurgical treatment and because this group apparently received more of the same care they had all “failed” to be accepted into this study, making this more of a study of surgery versus no treatment.  Even so, repeating care the nonsurgical group had all failed once still provided significant relief for 29%. This supports the belief that many operations for LDH could be avoided if energetic conservative management was continued for longer periods before surgery. 
Or, since one third of patients who found significant relief with the second course of conservative treatment failed the first, perhaps 2 exhaustive courses of conservative care by different providers, including spinal manipulation, needs to be the minimum standard of care before proceeding to surgical treatment. Patients may respond differently to treatment provided by one provider than they do to the same treatment given by another provider.
Furthermore, there is controversy as to whether or not spinal manipulation can actually cause a disk herniation. The reported cases of office manipulation causing CES are poorly documented, and it appears the cause-effect relationship was assumed when there was a temporal association between the manipulation and the symptoms  and this may be the subject of bias, as in other reported cases of manipulation iatrogenesis. [2, 37]. Terrett  said that some cases in the literature attributed to chiropractors and spinal manipulation were, in fact, caused by lay people or were reported more than once. Interestingly, only 7 of the 13 cases of CES occurred in the United States, where most of the world's chiropractors are located.  Some of the older reports may have been misdiagnosed.  It is also interesting that more new cases have not been reported in the literature recently, considering the increased use of chiropractic services. 
The only loading conditions known to cause posterior disk prolapse involve a combination of compression, lateral bending, and forward bending, [67, 68]. and standard lumbar spinal manipulation in the side posture position does not involve a combination of these movements.  Many authors agree that the axial rotation of the lower lumbar vertebrae is limited to 2° to 3° by impaction of the zygapophyseal (facet) joints, which prevents tearing of collagen fibers of the annulus within the physiological range of torsion, and torsional stresses just great enough to damage the facet joints do not generate enough torque to rupture the disk. [2, 16, 69, 70].
Others, however, found that annular fibers restrict rotation first, 0.8° before the facets act as a second barrier, and this indicates the annulus can be injured with rotation.  Bogduk  suggests that after the facets impact and prevent further motion around the normal axis of rotation, sufficient force would then change the axis of rotation of the vertebrae from somewhere within the vertebral body out toward the impacted facet such as to cause a lateral shearing force through the disk. He suggests that 3° rotation of a vertebrae causes 4% elongation of the collagen fibers of the annulus, and collagen fibers suffer microscopic injury at this point. Any further motion, such as flexion or shearing forces, would exceed the 4% limit of collagen elongation and this could cause an annular tear without facet failure. An extra 3° to 4° of rotation may be available at each lumbar joint when the spine is flexed. 
When flexion, rotation, and compression are combined over an adequate length of time, annular separation and subsequent prolapse of annular material will occur.  Brinckman and Porter  concluded that for a disk prolapse to occur, there needs to be both an annular fissure and a fragment within the disk. these researchers sliced from the anterior through the posterior annulus leaving only 1 mm of annulus intact and produced only a small bulge of 0.8 mm with compression and flexion. But when a small fragment of disk material of the size frequently seen at diskectomy was inserted, it took only a small compression load and flexion of less than 10° to prolapse extruded fragments through a complete annular tear. This type of loading is considered to be well within everyday physiological conditions and could happen with a cough, sneeze, laugh, straining at stool, or a stumble. [37, 38, 73].
It may be that for spinal manipulation to cause increased symptoms of disk herniation or cauda equina syndrome, the disk must already be fragmented and fissured such that any increased strain, like that imposed by normal daily activities, will cause a rupture and prolapse. Considering CES occurs most of the time in the absence of manipulation, at least some of the cases attributed to spinal manipulation could have had the same outcome without manipulation. [3, 37, 38]. The practitioner in most cases apparently does not actually cause the injury but aggravates a preexisting lesion for which the practitioner is consulted.  A clinician who administers treatment during the prodrome of a disk herniation is at risk of being identified as the cause, if leg pain and neurological deficit ensue.  Gentle technique and limitation of lumbar flexion during rotational manipulation may further reduce the risk to patients presenting with LDH.
Evidence-based care, as the term implies, bases the care a patient is given on the best evidence available in the research literature. The risk of spinal manipulation causing a clinically worsened disk herniation or CES in a patient presenting with LDH has been calculated to be less than 1 in 3.7 million manipulations.
Definitive treatment for LDH is currently unknown, but conservative care options should be exhausted prior to surgical treatment. Spinal manipulation is often left out of the conservative care of LDH patients, but many authors suggest a trial of spinal manipulation should be included as part of the conservative treatment plan because there is preliminary data supporting its efficacy. Spinal manipulation for the treatment of LDH appears to be very safe, and there is no sound basis to recommend against a trial of spinal manipulation of patients with LDH, although limited lumbar flexion and gentle technique are suggested to further reduce the risk. Perhaps 2 exhaustive courses of conservative care by different providers, including spinal manipulation, should be the minimum standard of care before proceeding to surgical treatment.
Disk herniation is the number one claim against chiropractors; yet, it appears likely that lumbar disk prolapse could occur only in an already fissured and fragmented disk. Even in patients presenting with LDH, the risk of spinal manipulation appears minimal, especially compared with other common treatments for LDH, such as NSAIDs and surgery, and spinal manipulation may be no more dangerous than activities of daily living, such as a cough or stumble.
More research is needed to determine accurately the incidence of disk injury/increased disk symptoms following spinal manipulation; under what conditions, if any, spinal manipulation can actually cause a disk herniation; the benefit of spinal manipulation in the treatment of LDH compared with natural history, other conservative treatments, and surgery; and which patients will benefit most from which type of treatment.