Systematic Review of Adverse Events Associated with Pediatric Spinal Manipulation
The publication by Vohra, Johnston, Cramer and Humphreys  remains the most current, comprehensive systematic review of the literature on adverse events associated with pediatric spinal manipulation. Vohra et al  performed a comprehensive search of eight major electronic databases including MEDLINE, PubMed, Embase, AMED and MANTIS from inception to 2004. All languages were included. In particular, MEDLINE and PubMed searches covered a period of 58 years from 1966 to 2004. In order for the reports or studies to be included, they must have been primary investigations of spinal manipulation; the study population included participants 18 years of age or younger and data on adverse events was reported. Reports were not limited to any particular healthcare profession.
Adverse events (AE) were categorized as severe, moderate or minor as described previously. Interestingly and controversially, adverse events rated as severe or moderate as the result of a delay in diagnosis and not directly related to the application of spinal manipulation, were also included in the data.
Thirteen reports of adverse events associated with pediatric spinal manipulation were identified over the 58 year period to June 2004. Of these, two were clinical trials [17, 18]; four were case studies [19–22]; and seven were case reports [23–29]. Ten reports were in English, two in French and one in German.
Of the 13 reports, nine were categorized as serious adverse events (SAE), one as a moderate adverse event (MoAE) and three as minor adverse events (MiAE). Interestingly, of the nine SAE, eight were single case reports and one was a case series involving three children. Therefore all case reports and one of the two case series identified by Vohra et al,  involved SAE. It is noteworthy that neither of the two clinical trials [17, 18] resulted in any SAE.
Limitations of Vohra et al Systematic Review 
No studies, including systematic reviews, are free of limitations or indeed errors which may affect their results and subsequent contribution to the scientific literature. Nevertheless it is clear that regardless of the limitations, this systematic review is important and provides useful information related to adverse events and pediatric spinal manipulation. The following is a discussion of the major limitations related to the Vohra et al  systematic review that is relevant to this current paper.
Serious Adverse Events (SAE)
The results of the Vohra et al  comprehensive literature search covering a 58 year period, involving eight major databases and including all language reports, resulted in nine serious adverse events (SAE) related to Ped MT. However, on further analysis, of the nine SAE, one involved an examination of a patient's traumatized cervical spine by a medical practitioner . Examination of the cervical spine is not done with therapeutic intent thereby excluding it by definition from Ped MT [11, 12]. Consequently a total of eight SAE instead of nine SAE should have been reported.
Misdiagnosis and Delayed Diagnosis
Of the SAE reported, at least three and possibly up to five of these cases had underlying pathologies or conditions (spinal cord astrocytoma, osteogenesis imperfecta, congenital occipitalization) which are clearly contraindications for Ped MT [25, 28, 29]. SAE understandably resulted from inappropriate application of spinal manipulation in these cases due to a misdiagnosis. The clinical point here is that the association of an adverse event to pediatric SMT arises from a misdiagnosis.
Equally, a delayed diagnosis of a serious underlying pathology or condition needing referral to the appropriate medical practitioner may also result in a SAE or MoAE. A study of 90 chiropractors in the Boston metropolitan area in the United States identified that 17% would continue to be the sole treatment provided in a hypothetical neonate case with a two week history of fever . Clearly this conflicts with current medical guidelines for treatment of pediatric cases with fever and likely will result in a delayed diagnosis and possibly some type of adverse event. As above, the clinical point is that a delayed diagnosis leads to the inappropriate application of SMT and the AE.
Consequently the author suggests that it would be more informative for clinicians and researchers if in addition to the current classification system, delayed diagnosis would be identified if appropriate. Consequently Ped MT given to a patient with a serious underlying pathology leading to a SAE would be classified as SAE (misdiagnosis) rather than simply SAE. Similarly if Ped MT were performed in a case of misdiagnosis which led to a moderate AE, it would be classified as MoAE (misdiagnosis) and so forth. An SAE alone is the result of the treatment without serious underlying pathology. Currently adverse events are lumped together and only by sorting through the details can one arrive at a better understanding of the AE. This is unhelpful to clinicians as it hides the underlying cause of the SAE.
Update of Research Literature on Possible Adverse Events and Ped MT
One of the common criticisms lodged against the paucity of adverse events and Ped MT in the literature is that the spontaneous reporting of AE leads to underreporting [15, 28].
Since the publication of the Vohra et al  systematic review, a search of the literature (MEDLINE, PubMed, PubMed Central) from July 2004 until present (January 2010) retrieved three reports related to Ped MT and adverse events. All three studies made an attempt at providing information on prevalence or rates of adverse events in pediatric patients undergoing manual therapy although with variable quality.
Hayes and Bezilla  conducted a retrospective review of adverse events (aggravation and complications) in the medical records for pediatric patients receiving Osteopathic SMT (OPSMT). A total of 346 out of 502 files reviewed met their inclusion criteria. Although no serious complications were found associated with OPSMT, 9% (31 patients out of 346) had file entries of clinical aggravation after OPSMT. These could be categorized as minor adverse events such as worsening symptoms, irritability, soreness, headache, behavior problems and pain. The authors' reported that their findings support OPSMT as a safe treatment for the pediatric population . However, a retrospective file review may suffer from numerous methodological flaws and the results should be viewed with caution.
Miller and Benfield  published a three year retrospective case file review of all patients younger than three years of age attending the outpatient clinic of the Anglo-European College of Chiropractic in the UK. A total of 781 pediatric cases were retrieved. Of these 699 (89.5%) pediatric cases representing 5242 treatment visits were included with over 77% having received Ped MT. The other 82 (10.5%) were referred for other care. Most of the pediatric patients (n = 574; 73.5%) were 12 weeks of age or younger. The most common age group was between five and eight weeks with the most common presenting complaint being attributed to spinal biomechanical dysfunction manifesting as 'irritability or colic' often attributed to birth trauma. Miller and Benfield  reported that over 85% of parents reported improvement in their child's complaints with treatment while just over 14% reported no change.
For adverse events, negative side effects as identified by the parent were reported. A total of seven adverse events out of 697 pediatric patients (two were lost to follow-up) were elicited. All seven involved minor adverse events (crying or increased crying or restlessness or sleeping disturbance) which were transient and did not require medical care.
Miller and Benfield  estimate 1% of pediatric patients (one in 749 treatments) suffered from a minor and self limited AE. A detailed description of the AE allowed the authors to review and classify them. Based on their analysis, three of the events attributed as AE may have been incorrectly attributed as a result of treatment (i.e. pre-existing constipation, common behavior of child, etc.). If these are excluded, the AE rate becomes one reaction per 1310 treatments.
The study by Miller and Benfield  is a higher quality study compared to the one by Hayes and Bezilla  because it incorporated a more rigorous methodological design. However, as a retrospective file review, there were likely to have been some methodological limitations such as reporting bias, different time periods over which the data was collected, and the difficulty of identifying adverse events experienced by infants but as reported by their parents.
Alcantara, Ohm and Kunz  conducted a retrospective cross-sectional survey of 577 pediatric patients (5,438 treatment visits) ranging from less than one day old to 18 years of age (mean age = 7.45; mode age = 1; median age = 7) attending for chiropractic treatment. All patients received treatment to at least one spinal region. Spinal manipulation was the most common treatment employed with 492 of the 577 pediatric patients received either manual (449) or instrumented mechanical manipulation (43).
In terms of adverse events, the parents of the pediatric patients reported only three events out of 5,438 treatments, all of which were minor (muscle stiffness, spine soreness, stiff and sore) and time limited. Based on these results, Alcantara et al  reported that 0.83% of pediatric patients or one in 1,812 patient visits resulted in a minor adverse event after chiropractic treatment. However this study was poorly designed with many possible sources of bias and errors. It suffers from similar methodological design flaws as the study by Hayes and Bezilla .
The three recent studies identified [14, 16, 30] were all retrospective file reviews and as such may have suffered from flaws associated with this methodology. Carnes et al.  suggest that several factors such as unclear definitions of Ped MT, different time periods of reporting, whether the patient or practitioner reports on the adverse events, confidentiality issues, missing data and missing data to follow-up, misinterpretation of data, quality assurance and bias of file reviewers, may adversely affect the validity of the study. Therefore the results of these three studies should be viewed with caution.
A Recent Systematic Review of Adverse Events and Manual Therapy
Carnes et al  recently published a systematic review of adverse events and manual therapy (MT), irrespective of age. Inclusion criteria were studies which used manual therapy only; therapy was delivered by a registered therapist; the intervention was clearly described and adverse events were reported.
Carnes et al  identified eight prospective cohort studies and 31 manual therapy randomized controlled trials (RCTs). Although none of these studies specifically investigated Ped MT and adverse events, the results may provide a useful benchmark for pediatric adverse events in the absence of other research evidence. In addition, Carnes et al  performed a meta-analysis on the pooled data from both the cohort studies and the RCTs.
Carnes et al  were able to confirm that as reported by other authors, approximately half of adult patients treated by manual therapy are likely to experience a minor to moderate adverse event after treatment, and particularly after the first treatment [31–36]. These adverse events typically begin within 24 hours after treatment and most resolve within 48 hours [31, 34]. However the risk of a SAE is small with no catastrophic adverse event such as death or stroke reported in any of the eight cohort studies or 31 RCTs included in the review.
In the meta-analysis, Carnes et al  were able to pool their results and compare MT to other therapies. MT was found to produce, in general, more adverse events compared to general medical practitioner care, about the same number compared to exercise but less adverse events than drug therapy (Non-steroidal anti-inflammatory drugs such as diclofenac or amitriptyline, an antidepressant).
Even though there were no reports of catastrophic or SAE in the 39 studies, one cannot assume that they did not occur as underreporting of adverse events is possible. Using the Exact method (according to binomial theory) Carnes et al  estimated the risk of an SAE after manual therapy at the upper 95% confidence interval to be approximately 0.13%. Thiel et al  using Handley's rule of three calculated an upper 95% confidence limit of SAE following chiropractic care in their RCT to be approximately 0.01%.
Carnes et al  concluded that the results of their meta-analysis showed that the relative risk of having a minor or moderate adverse event after high velocity thrust spinal manipulation was significantly less than taking medication that is often prescribed for these painful conditions.
Limitations of the Systematic Review
The Carnes et al 2010 systematic review and meta-analysis  had some methodological limitations related to the methodological quality of the included studies. In particular, the included cohort and RCT studies may have suffered from unclear definitions of manual therapies, different time periods over which the data was collected, bias due to patient or practitioner reporting, confidentiality issues, patient satisfaction issues and loss of patients to follow-up. Patient reporting bias and patient selection bias may have also affected the findings, along with concurrent treatment with other healthcare professionals or self-medication by patients [12, 33].