Our literature review showed that the genes for many molecules associated with motor control circuits have common polymorphisms. Such polymorphisms might conceivably contribute to excessive guarding shown by some patients with pain.

The present study of 14 of these candidate genes revealed an association of a GABA-A β1 subunit haplotype with the amount of motor limitation (Table 5). Two copies of the haplotype reduced movement limitation scores almost 30% compared with no copies. The nominal p value of 0.0025 is small enough to be interesting even if one corrects for the 22 haplotype blocks examined, yielding an overall p value = 0.055. If confirmed by subsequent studies, we would hypothesize that this variant of the gene serves as an "endogenous muscle relaxant", with greater GABAergic motor inhibitory activity that reduces spasm and permits more activities.

We cannot speculate about the specific biochemical mechanism by which this haplotype might affect motor function. The finding that the haplotype analysis was much more sensitive than single SNP analyses suggests that a movement-facilitating SNP that we did not directly test may be present on the haplotype we identified. Compared to other GABA receptors, the GABA-A receptor β1 subunit has been little studied [20], and there is no commonly used specific antagonist, agonist, or knockout mouse. The published literature states that benzodiazepine muscle relaxant effects are primarily mediated through GABA-A α2, α3, and α5 receptor subunits [21]. However, interest in the β1 subunit is growing because of possible associations with autism [22] and bipolar disorder [23], so the gene may soon be better understood.

The nominally significant associations of two ADRA2C SNPs and a haplotype with motor limitation scores are not persuasive because of the modest size of the effect and the many genes tested. However, this gene is known to have common polymorphisms with strong effects on cardiovascular regulation and mood, such as a deletion affecting about 40% of chromosomes in African Americans [24,25], not tested for in the current study of a Caucasian population. Alpha-2 adrenergic receptors mediate inhibition of motor tone at the spinal level [26], so this gene should be revisited in future studies.

Because of several limitations in our study methods, we consider these findings tentative. The first limitation is that the Roland-Morris scale items that we used are retrospective and subjective assessments of the patient's ability to carry out various movements of daily life. Moreover, we chose a subset of these items to represent changes in movement without any formal validation studies. Objective measures of motor function would have been more convincing, but these were not available for this cohort.

Secondly, we tested only 14 genes, chosen by an anatomical analysis the emphasized the final two or three neurons in the motor control pathway. Molecules influencing more rostral levels of control might also be candidates. For example, Kleim et al. [27] have reported a BDNF polymorphism to be associated with modified plasticity in the human motor cortex. A more thorough search could be done using a whole genome association array, but this would require a sample size in the thousands [23] to provide adequate power.

Thirdly, we would have been interested in searching for an interaction between gene effect and pain level on movement, similar to the gene, pain, and depression analysis we have done on the same cohort [15]. For example, some polymorphic alleles might produce a major reduction in movement only in the presence of severe pain. However, this analysis was not appropriate in this case because the scale items mention both pain and various movements.

Finally, long-term decrease in movement is more directly relevant to the economic burdens of pain than the acute effects that we examined above. The analytical approach demonstrated above might also be used to search for effects of genes on a chronic decrease in movement. We chose to examine acute pain and movement data as an initial demonstration of this method for several reasons: (1) Almost all of our 290 patients who gave a DNA sample had pain at the baseline measurement, while a minority had persistent pain at one year or later time points. Sample size is crucial in screening multiple candidate genes [19]. (2) A shorter time window in which patients are studied under similar methods and treated by the same clinicians may be more sensitive to genetic effects than a longer period during which many types of environmental variability overshadow gene effects. (3) The processes involved in chronic pain and movement limitation are probably more complex than acute pain, including deconditioning, muscle atrophy and contractures, chronic psychiatric morbidity, and occupational factors.

Although physicians commonly prescribe "muscle relaxants" such as cyclobenzaprine, carisoprodol, and methocarbamol for acute pain, the database supporting their effects on muscle and clinical usefulness is limited [28]. Development of new treatments will require expansion of animal research on the neurochemical mechanisms of pain-related guarding. We propose that, as has been demonstrated in genetic studies of pain [14], genomic screens in humans may be useful in to prioritizing targets identified in animal studies.