Obstetric brachial plexopathy occurs in approximately 1.5 of 1000 live births1,2, and this prevalence seems to be increasing despite medical progress in obstetrics2,3. Although the natural history of these lesions is not well defined, most of these children recover well without surgery4. However, some have development of a severe lifelong motor impairment, and the indications for and timing of brachial plexus surgery for these patients are a matter of controversy. Patients who receive an early intervention have a better prognosis5-9, but it can also mean that some patients have unnecessary surgery10. The most frequently used clinical criterion, which was defined by Gilbert et al. twenty years ago, is based on the absence of biceps function at three months of age5; however, there is some confusion regarding the exact definition of so-called biceps function. Gilbert et al. used palpable biceps contraction, while others use elbow flexion9,11,12. The presence of voluntary biceps electrical activity in electromyography is not considered biceps function5,8. The Gilbert criterion has been criticized because of its lack of specificity11-16; therefore, some surgeons prefer to wait longer in order to avoid unneeded surgery17,18.
Most surgeons do not rely on electromyography for the selection of infants for brachial plexus surgery5,7,8,19-22. Needle electromyography is an uncomfortable procedure, which is difficult to perform in an uncooperative patient, and it requires a well-trained neurophysiologist with pediatric expertise. Many authors have reported that needle electromyography is usually overly optimistic in relation to clinical evaluation5,7,19,21, and this was confirmed in a recent study by our group23. On the other hand, nerve conduction studies may still be helpful. The side-to-side comparison of the amplitudes of compound muscle action potentials can provide an estimation of the amount of motor axonal degeneration of the terminal nerves arising from the brachial plexus. We have shown that the ratio between the amplitude of the compound muscle action potential of the affected side and that of the healthy side is related to clinical recovery in these patients24.
The objectives of this study were to determine whether this index can be useful for the selection of infants with a poor prognosis, to establish neurophysiologic criteria, and to compare the performance of this index with the clinical criterion proposed by Gilbert et al.
From July 2000 to February 2007, seventy-seven patients with obstetric brachial plexopathy who were less than sixty days old were referred to our hospital. Thirteen patients showed remarkable clinical recovery after a few days and were not considered for neurophysiologic evaluation. We also excluded one patient who had cerebral palsy develop, two patients with bilateral lesions, and seven patients who were lost to follow-up. The remaining fifty-four patients are the subjects of this prospective cohort study. Sample size was determined by convenience.
The birth weight ranged from 2695 to 5515 g (median, 3820 g). There were thirty-four normal vaginal deliveries (63%), including two breech presentations (4%); nineteen forceps-assisted deliveries (35%); and one cesarean section delivery (2%). Thirty patients (56%) were boys, and twenty-four were girls. The right side was affected in thirty-eight patients (70%) and the left side, in sixteen (30%). Thirty-five subjects (65%) had clinical evidence of a lesion of the C5-C6 levels, thirteen (24%) had involvement of the C5-C7 levels, five (9%) had complete lesions, and one (2%) had C7-T1 involvement. This last patient with Klumpke palsy was evaluated two days after a breech delivery and showed hand paralysis and wrist drop with preserved shoulder abduction and elbow flexion.
The patients underwent bilateral motor-nerve-conduction studies between ten and sixty days of age (median, twenty days), with forty patients (74%) who were seen during the first month. The hospital ethics committee approved the study, and parents provided written informed consent. The nerves (and muscles) evaluated included the axillary (deltoid), musculocutaneous (biceps), proximal radial (triceps), distal radial (extensor digitorum communis), median (thenar eminence), and ulnar (hypothenar eminence). Technical data concerning electrode type and placement are provided in our previous study24. The placement of recording electrodes should be as symmetrical as possible, as this is the main source of variability. We compared the side-to-side amplitudes of compound muscle action potentials and defined the axonal viability index as the amplitude of the affected side expressed as a percentage of the corresponding value of the healthy side.
All of the patients were followed on a monthly basis by the same pediatric neurologist (C.O.H.), who used a standard evaluation protocol that included power assessment of shoulder abduction and external rotation, elbow flexion and extension, forearm supination, wrist extension and flexion, extension and flexion of the fingers, and thumb opposition. We used a modified Medical Research Council motor scale, ranging from 0 (no contraction) to 5 (normal findings) for each movement (see Appendix). This modification incorporates some concepts from the motor scale of The Hospital for Sick Children19. For infants older than six months, we also used a modification of the so-called towel test25, in which the child should be able to remove a blindfold from the face, while in the sitting position, using the affected limb and without flexing the neck.
The infants were divided into three groups. Patients in group A showed complete recovery at six months of age, with no muscle strength asymmetry and a normal towel test with the affected limb. Patients in group B had a good outcome at twelve months of age. These patients were able to perform the towel test with the affected limb at this age and fulfilled the criteria for a good outcome according to Narakas (see Appendix)26. Patients in group C were considered to have a poor outcome. These patients were not able to perform the towel test with the affected limb and did not fulfill the Narakas criteria for a good outcome at twelve months of age.
We calculated the average axonal viability index for each group and used the Kruskal-Wallis test to verify group differences for each motor nerve evaluated.
Our objective was to detect patients in group C with use of motor nerve-conduction studies. For each nerve, we constructed an empirical receiver operating characteristic curve using four different axonal viability index cutoff points: 0% (no motor potential), <10%, <20%, and <50%. This curve plots the sensitivity versus the proportion of false positives (or 1 — specificity) at different cutoff values27,28. Usually there is an inverse relation between sensitivity and specificity. If the axonal viability index cutoff value is too high, sensitivity would be high, but specificity would be low. On the other hand, if the axonal viability index cutoff is too low, sensitivity would be low, but specificity would be high. The ideal test (100% sensitivity and 100% specificity) is located in the upper left corner of the plotting area. The lower left corner and the upper right corner are not truly cutoff points, but are respectively considered the starting and ending points of the curve. The lower left corner represents a test rejecting all patients (0% sensitivity and 100% specificity), and the upper right corner represents a test selecting all patients (100% sensitivity and 0% specificity). From the start to the end, the other four dots represent sequentially the axonal viability index cutoff values of 0%, 10%, 20%, and 50%. The best cutoff point was defined as the curve point closest to the upper left corner27. After defining the best axonal viability index cutoff point, we calculated the 95% confidence interval for its sensitivity and specificity.
We registered the biceps function at three months of age from each patient, using the Medical Research Council motor scale. In a similar way, we constructed an empirical receiver operating characteristic curve to define the best cutoff point to detect patients from group C. After defining the best cutoff point on the Medical Research Council scale, we also calculated the 95% confidence interval for its sensitivity and specificity.
In order to compare the methods, we divided the sensitivity and specificity obtained from the neurophysiologic criteria by that of the clinical criterion. We also calculated the 95% confidence interval of that ratio. If the confidence interval included the number one, there was no significant difference between the two methods.
Source of Funding
There was no external funding source for this study.
Twenty patients (37%) had a fast recovery, showed no muscle strength or movement asymmetry at six months of age, and were assigned to group A (a complete recovery). These included seventeen patients (85%) presenting initially with C5-C6 involvement, two patients (10%) with C5-C7 involvement, and one (5%) with Klumpke palsy. At six months of age, ten of these patients showed only minor deep tendon reflex asymmetries during the clinical evaluation. In the remaining patients, it was not possible to determine by clinical evaluation which was the originally affected limb.
Seventeen patients (31%) had a good recovery and were assigned to group B. All were able to perform the towel test with the affected limb, while in the sitting position, without assistance at twelve months of age. Five of these patients showed no clear movement asymmetry. All five of them presented initially with C5-C6 involvement. These patients were not assigned to group A, since such delayed recoveries are probably not truly complete. Seven patients, including six patients presenting initially with C5-C6 involvement and one with C5-C7 involvement, showed minor asymmetries or scapular winging. Five patients showed obvious motor deficits or the trumpet sign. These motor deficits were related to shoulder external rotation and abduction or forearm supination. Four of these children initially had C5-C6 involvement, and one had C5-C7 involvement.
Seventeen patients (31%) were assigned to group C, as they were unable to perform the towel test at twelve months of age. The follow-up period for this group actually ranged from eighteen to seventy-three months, and none fulfilled the Narakas criteria for a good outcome after the twelve-month observation period. All five patients presenting initially with global paralysis, as well as nine patients with C5-C7 involvement and three with motor deficits restricted to the C5-C6 levels, were included in this group. Eight patients in this group underwent external neurolysis between six and twelve months of age, but none showed muscle power deterioration after surgery. Eight patients in group C had biceps muscle power of less than grade 3 at one year of age, and another patient had transient finger-biting behavior develop.
The compound muscle action-potential amplitudes obtained from the control side of the patients are shown in a table in the Appendix. The distribution is not gaussian, so median values are presented instead of means.
We analyzed the distribution of axonal viability index values in each group for every motor nerve studied. Figure 1 shows an example of a scatterplot of the values. The axonal viability index median and range for each group are shown in Table I. A significant difference was detected among the groups for all motor nerves, except for the median nerve. The median nerve also showed marked dispersion of axonal viability index values, with values of up to 300%. Therefore, the axonal viability index for the median nerve was excluded from the receiver operating characteristic curve analysis.
The empirical receiver operating characteristic curves for each motor nerve are shown in Figures 2 through 6. From the lower left to the upper right corners (excluding the corners themselves), the dots sequentially represent the axonal viability index cutoff values of 0%, 10%, 20%, and 50%. In the receiver operating characteristic curve of the axonal viability index for the ulnar nerve, the 10% and the 20% dots overlapped, so there are only three dots. The best axonal viability index cutoff values for each motor nerve, their sensitivity, specificity, and 95% confidence intervals are shown in Table II.
We also analyzed the distribution of the Medical Research Council scores for the biceps at three months of age among the groups, as shown in the Appendix. There was a significant difference among the groups (p < 0.01). We also constructed an empirical receiver operating characteristic curve to select the best cutoff point to define patients with a poor prognosis (Fig. 7). Despite the good specificity for the absence of biceps contraction, the best cutoff value was considered the absence of elbow flexion, which showed a much higher sensitivity. The sensitivity, specificity, and 95% confidence interval are shown in Table II.
The comparison of sensitivity between the neurophysiologic criteria and the clinical criterion is shown in Figure 8. The sensitivity was higher for the clinical criterion, but this was significant only for the ulnar nerve (p < 0.05), with a sensitivity of 37% less than that of the clinical criterion. The comparison of specificity between the neurophysiologic criteria and the clinical criterion is shown in Figure 9. The specificity was usually higher for the neurophysiologic criteria, and this was significant for the proximal radial, distal radial, and ulnar nerves (p < 0.05), with the specificity being 20% higher than that of the clinical criterion.
Seven patients were lost to follow-up, and six of the seven were evaluated until at least three months of age. At the time of the last evaluation, two patients had active elbow flexion and four did not. If we assume all lost patients had a poor outcome, the sensitivity of motor conduction studies would be 75% (95% confidence interval, 53% to 90%) for the axillary nerve, 87% (95% confidence interval, 68% to 97%) for the musculocutaneous nerve, 67% (95% confidence interval, 45% to 84%) for the proximal and the distal radial nerve, and 42% (95% confidence interval, 22% to 63%) for the ulnar nerve. The sensitivity of absent elbow flexion at three months of age would be 87% (excluding the patient lost before this age), which would be significantly higher than that of the motor conduction study of the ulnar nerve. If we assume all of the lost patients had a good outcome, the specificity of motor conduction studies would be 84% (95% confidence interval, 70% to 93%) for the axillary nerve, 64% (95% confidence interval, 48% to 78%) for the musculocutaneous nerve, 93% (95% confidence interval, 81% to 99%) for the proximal and the distal radial nerve, and 98% (95% confidence interval, 88% to 100%) for the ulnar nerve. The specificity of absent elbow flexion at three months would be 74%, which is still substantially less than that of the motor conduction studies of the proximal radial, distal radial, and ulnar nerves.
Our results show that comparative motor-nerve conduction studies performed between ten and sixty days of life may be useful in the selection of infants for brachial plexus surgery. Timing is an important issue for motor conduction studies. wallerian degeneration is a relatively slow process, and it takes as long as nine days in adults to be unable to elicit a compound muscle action potential in a transected nerve29. In newborns, this period is probably shorter30, but there is little evidence of this in studies of humans31. However, this is the reason we waited at least ten days to perform the neurophysiologic study. On the other hand, motor conduction studies should be done as soon as possible after wallerian degeneration, if our goal is to provide an estimation of axonal viability. This estimation relies on the assumption of equivalent motor unit sizes on both sides. After a partial axonal injury, the remaining axons provide distal collateral sprouting to reestablish synaptic connections with the denervated muscle fibers, and the resulting motor unit is larger than it was originally32. The side-to-side comparison of amplitudes of compound muscle action potentials underestimates the amount of axonal degeneration if it is done after the collateral reinnervation process has already began29. The upper limit of sixty days is arbitrary, as we do not know when collateral sprouting begins to affect the compound muscle action-potential amplitude.
Group C included eight patients who were managed surgically between six and twelve months of age, and all had external neurolysis. Since we did not observe loss of muscle power after surgery and no patient had a satisfactory recovery, surgery was not responsible for the poor outcome, and we believe that this outcome represents the natural history of these lesions. Other authors have also reported poor results after neurolysis33,34. We now have adopted a more aggressive approach, and nerve grafts are usually used in these patients after intraoperative monitoring of nerve action potentials. Patients having nerve graft surgery were not included in this series.
An axonal viability index of >100% seems unusual, but this only reflects a compound muscle action-potential amplitude on the affected side that is larger than that on the healthy side. The normal asymmetry of compound muscle action-potential amplitudes in adults can range up to 50%29,35, but there are no similar studies in newborns. Assuming an asymmetry similar to that in adults, axonal viability index values of up to 200% would be expected. This was true for all except the median nerve. The great dispersion of axonal viability index values in the median nerve was due to low compound muscle action-potential amplitudes, even on the healthy side. Our electrodes had a diameter of 1 cm, and this usually covered all of the thenar muscles and possibly caused phase cancellation of the motor unit potentials or short circuits beneath the electrode. For larger muscles, the electrode size was not a problem. Reproducibility of compound muscle action-potential amplitudes in newborns was not addressed in our study. This is also a concern in similar tests such as the facial motor-nerve-conduction study for prognostic assessment of facial nerve palsy in adults36,37, despite the fact that it has been commonly used for a long time. Perhaps most of the so-called normal 50% asymmetry between sides is, in fact, related to minor differences in the placement of the recording electrodes, which also would affect reproducibility. This precludes motor nerve-conduction studies as a diagnostic test, since minor differences could not be noted. However, in the case of prognostic assessment, we are dealing with major differences between the two arms.
Analyzing the area under the receiver operating characteristic curves, we concluded that the proximal radial motor-nerve-conduction study was the best test of all, but axillary and distal radial nerves were also good. The musculocutaneous nerve study had a low specificity, while the ulnar nerve study had a low sensitivity. The low sensitivity of the ulnar nerve study was related to the pathophysiology, since we were analyzing the lower part of the brachial plexus and most obstetric lesions do not affect that portion. An absent compound muscle action potential was always associated with a poor prognosis, except for the musculocutaneous nerve, and the low specificity of this nerve may be related to stimulation problems. The musculocutaneous nerve was stimulated at the axilla, very close to the median nerve. High currents could not be used since the current spread would activate the median nerve and cause a volume-conducted potential from the forearm muscles. Conversely, selective stimulation of the axillary and proximal radial nerves was impossible (both were stimulated at Erb's point), but volume-conducted potentials from other than the target muscles were not a problem. Variability in test performance could affect the result in borderline cases, so we recommend that several nerves should be evaluated and that all data be analyzed together, instead of relying on a single axonal viability index.
In our previous study, we found that an axonal viability index of >10% was related to recovery of the corresponding level of the brachial plexus24. This continues to be our opinion, but our approach was different this time since we were interested in the recovery of the patient as a whole. An axonal viability index of the ulnar nerve of 30% would mean that the chance of recovering finger flexion is good, but the overall prognosis is poor. A lesion severe enough to cause ulnar axonal degeneration probably is associated with a major axonal degeneration of the upper levels, assuming the usual cranial-caudal gradient of the lesion.
We found good specificity of the clinical criterion in our study. In fact, it was better than has been previously reported by other authors11-13,16. Nevertheless, nerve conduction studies provided higher specificity values, with similar sensitivity. The patients lost to follow-up did not affect this result. Some may say that the clinical criterion proposed by Gilbert et al. was actually the absence of biceps contraction, and if this criterion were used, the specificity would be 100%. This is true, but this interpretation would decrease the sensitivity to only 47%. If absent compound muscle action potentials were used as the neurophysiologic criteria, the specificity of motor conduction studies of the axillary, proximal radial, distal radial, and ulnar nerves would also be 100%. The receiver operating characteristic curve analysis is particularly useful in this matter, since it provides sensitivity and specificity with different cutoff values. We selected the most accurate cutoff value, but different values could be used depending on specific needs, for example, to maximize sensitivity or specificity.
Perhaps the great advantage of motor conduction studies in relation to clinical evaluation is the timing of the prognostic estimation. Motor conduction studies can provide prognostic information with high specificity as soon as ten days after birth, whereas we would have to wait at least three months using the standard approach. We are not proposing surgery before three months of age yet31, but this period could be used to perform elective imaging studies in selected patients. This extra time could be helpful for parents who are considering surgery, and it could provide time to refer the child to other centers with expertise in obstetric brachial plexopathy management for a second opinion.
The value of clinical examination, which remains a hallmark surgical indication, should not be diminished. However, nerve conduction studies can be very helpful in the treatment decision process. Our protocol is easy to perform, is widely available, and is more objective than the standard neurophysiologic approach.