A lateralized model of the pain-depression dyad.

While distress encompasses many emotional constructs, in this section we focus on depression, which is ranked by the World Health Organization as the leading cause of disability around the world 48 , and its shared neural circuitry 10 , 11 with pain. Specifically, we scrutinize the neural correlates of these two disorders through the lens of laterality. In the localization of neural activation in response to pain, a group of bilaterally distributed cortical and subcortical regions (e.g., S1, insula, ACC, thalamus, among others) have been shown to respond to pain-provoking stimuli to the body 49 . Yet, these regions have also been shown to be recruited in the processing of non-noxious stimuli 50 , 51 , suggesting these regions might simply be engaged in the processing of salient stimuli 52 – 54 or of emotion related to the stimuli 55 – 58 . Also, in a case study of individuals with a congenital mutation that rendered them unable to perceive pain, there was atypical response in these same brain regions, suggesting such activations may not necessarily be associated with nociception 59 per se. Regardless, there are regions that show reliably lateralized neural correlates of evoked experimental pain irrespective of stimulation side ( Table 2 ) as well as lateralized neural correlates of various clinical spontaneous pain conditions ( Table 3 ). As for experimentally evoked pain, several regions show greater effects in the RH compared to the LH ( Table 2 ). Whereas the S1 is typically activated contralateral to the stimulated side consistent with its role in the sensory-discriminative dimension of pain, regions including the right thalamus, insula, and ACC (more on these regions below) are activated regardless of side of stimulation. Only one study showed strictly left-lateralized response in the insula 61 ; in the other study with left-lateralized response 64 (in putamen, pregenual cingulate cortex, supramarginal area, parahippocampus), there were also right-lateralized response in other regions (cerebellar lobes). On the other hand, the lateralization of the neural correlates of spontaneous clinical pain is heterogeneous ( Table 3 : 50% of the studies had right-lateralized response to painful disorders). In fact, even similar conditions could show opposite lateralization (e.g., migraine with both left 66 and right 67 neural correlates, albeit at different regions). It is important to note that experimental and clinical pain do not always have the same neural substrates 68 , 69 , let alone lateralization of such, which is perhaps because of the emotional and autonomic reactions in pain disorders. Moreover, heterogeneity may arise from individual differences in the manifestations of pathology, in contrast to the carefully controlled setting of an experiment. Nevertheless, the key finding from these studies is the plausibility of having a lateralized neural response to clinical spontaneous pain irrespective of location (e.g., Table 3 , widespread pain in fibromyalgia evokes right-lateralized activation). Below, we review pain-related regions with the aim of showing whether these regions exhibit lateralized effects in relation to depression. There are a few constraints we considered in delimiting our scope. First, we chose regions in the forebrain based on a meta-analysis of acute and chronic pain 9 that showed consistent activation of the following cortical and subcortical regions: S1, insula, ACC, and thalamus. We excluded the S1 given its main role of encoding the sensory-discriminative (as opposed to emotional-affective) dimension of pain. Second, we only included papers published from 2015 onwards to ensure we were surveying the most recent findings related to depression. Third, we excluded papers on depression with concomitant conditions (e.g., post-stroke depression) or studies that had other extenuating circumstances (e.g., study of neural changes in children who themselves did not have depression, but whose mothers were diagnosed with depression). The following subsections are not meant to be an exhaustive review of laterality of neural correlates of depression but rather to demonstrate the possible overlap of the lateralized neural correlates for both pain and depression. For each region, we present neural correlates of depression in the RH first, and then at the end of each sub-section also present findings from the LH. The thalamus is a key relay station through which nociceptive signals are sent to the cerebral cortex 78 . It exhibits lateralized involvement in depression in terms of functional 79 – 82 or structural connectivity 83 and volume 84 , 85 of the right thalamus. In a study of patients with treatment-resistant depression, non-responders had higher spontaneous activity in the right thalamus than responders 79 . Also, among patients who responded to treatment, those who had higher activity in the right thalamus exhibited lower clinical improvements at 6-week follow-up 79 . In patients with migraine and comorbid depression, the low intrinsic activity in the right thalamus may be indicative of the abnormal developmental trajectory compared to those with migraine or depression separately 80 . In college students presenting with subclinical depression 81 , although the connectivity profile of the bilateral thalamus was altered after stress paradigms (e.g., speech and mathematical calculation tasks), only the changes in the right thalamus predicted the severity of depressive symptoms at one-year follow-up. Last, changes in structural nodal degree of the right thalamus were correlated with the percent change in depressive symptoms following cognitive behavioral therapy in adolescents with depressive symptoms 83 . Interestingly, the aforementioned neural correlates of depression in the right thalamus concern mostly its function. In contrast, alterations in morphological properties of the LH, such as lower gray matter volume of the left thalamus 86 – 88 , were predictive of depressive symptoms. The insula is a region deep within the lateral sulcus of the brain 89 that is known to serve as an integration hub of sensory and affect processing 90 . Compared to healthy controls, patients with major depressive disorder exhibit abnormal functional 91 – 98 and structural 99 connectivity of the right insula. Even at the pre-clinical stage, greater depressive symptoms in healthy individuals were associated with the functional connectivity of the right insula 100 . Likewise, response to interventions including behavioral activation treatment 101 , cognitive behavioral therapy 102 or antidepressants 103 was also predicated on the functional connectivity of the right insula. In pain disorders such as fibromyalgia 104 and classical trigeminal neuralgia 105 , levels of depressive symptoms were associated with functional connectivity and activation of the right insula. Similarly, individuals with tinnitus who exhibited higher levels of distress also had higher activation of the right insula, compared to those with low distress 106 . Together, these findings corroborate the right insula’s role as a key neural substrate of depression mostly in terms of function, similar to the thalamus. Nonetheless, there are also left-lateralized abnormalities of the insula in terms of functional 107 , 108 or structural 109 connectivity and volumetric measures 110 – 114 . The ACC is part of the corticolimbic circuitry that is crucial for pain-related emotion 115 . Compared to healthy controls, patients with depressive symptoms exhibit altered functional connectivity 116 , 117 and lower stability of intrinsic function 118 of the right ACC. Patients with pharmacological treatment-resistant depression had lower gray matter volume of the right ACC compared to treatment responders 119 . Last, the right ACC gray matter volume was predictive of symptoms at 5 years follow-up in patients with depressive symptoms 120 . Like the other two regions, there are also left-lateralized effects observed in the ACC in terms of functional connectivity 121 , 122 and cortical thickness 123 related to depression. Unlike the other two regions, the depression-related effects of the ACC are equally distributed between hemispheres in terms of functional and morphological measures. These studies show that the thalamus, insula, and ACC, on top of being responsive to pain (and in some instances showing lateralized responses), also exhibit lateralized effects in depression. In fact, the lateralization elegantly fits into a pattern if we were to categorize the neural signatures into active versus passive properties defined as follows: active properties are related to neural functions such as activation and connectivity and passive properties are related to morphology such as cortical thickness and volumetric measures. We intentionally do not differentiate between functional and structural connectivity, because we argue that both types of connectivity are active properties by the following logic. Functional connectivity is positively correlated with the strength of structural white matter connections between two regions 124 . In instances when there is strong functional connectivity in the absence of a direct structural connection, it is supposed that the former is a result of polysynaptic transitive structural connections (i.e., there are structural connections to a third, separate region, which acts as a waypoint). In short, structural connectivity can be considered to be in service of functional connectivity. Of note, connectivity, an active property, relies on coactivation or connection between at least two areas (i.e., connectivity is a diffuse biomarker), in contrast to the possibly localized nature of morphological changes (i.e., passive properties are focal biomarkers). To summarize, in the regions scrutinized above, by proxy of the number of studies showing the respective lateralized effects, the RH neural correlates of depression were chiefly in terms of the active properties (i.e., activation and connectivity) while the neural correlates of the LH were in terms of passive properties (i.e., morphology). This pattern is consistent with the right hemispheric dysfunction thesis of depression 12 – 14 , in that the RH is associated with active effects (i.e., [dys]function). By adopting this dichotomy between active and passive neural correlates, we can now proceed to proposing a model for a lateralized pain-depression dyad in the next section.

As early as 1859, a study involving individuals with hysteria showed a higher frequency of left- versus right-sided hemianesthesia 15 [authors’ note: we recognize the sexist history of hysteria and we refer the reader to Mai and Merskey 16 for an historical perspective]. However, the concept of symptom laterality somehow lay dormant until more than a century later; in their seminal review, Merskey and Watson 17 reported a higher frequency of pain in the left hemibody of individuals with comorbid psychological disorders (e.g., conversion disorder). Consistent with Merskey and Watson’s findings, further studies 18 – 20 showed a high frequency of left-sided symptoms in individuals in pain with a psychological component (psychogenic pain hereafter). Notably, that Merskey and Watson 17 had made their initial observations of pain in individuals with psychopathological comorbidities was seemingly lost in some of the later reiterations, thus perpetuating the impression that pain would be more common in the left hemibody in any condition, or even in the general population. However, studies involving patients of various disorders and with symptoms in different parts of the body (e.g., head, chest, arms, legs) showed equal frequencies of left- and right-sided pain 21 – 24 . These findings suggest no difference in frequency of lateralized clinical spontaneous pain in the absence of confounding psychological comorbidities (although see Roelofs et al 25 for equal frequency of pain on either side in individuals with conversion disorder). Notwithstanding the conflicting evidence, the idea of pain being more common in the left side of the body without psychopathology is still referenced, even as recently as the last decade 26 , 27 . Logically, it follows that in order to conclude definitively whether any type of pain would indeed be more common on either side, researchers should look to pain disorders that have inherently lateralized symptoms. Moreover, the symptoms should not be attributable to psychological comorbidities. Two candidates that meet these criteria are migraine and complex regional pain syndrome (CRPS). A diagnostic feature of migraine is the unilateral pain during a given episode 28 , although the side of headache may shift from one attack to the next. CRPS is a chronic pain condition 29 that usually affects a unilateral distal extremity after some precipitating physical trauma 30 , 31 . The laterality of pain symptoms in these conditions make them ideal for adjudicating the hypothesized prevalence of left-sided pain and the data show equal frequency of pain occurring on either side in both migraine 32 and CRPS 33 – 35 . These findings suggest pain in either the left or right hemibody is statistically equal. Indeed, even for the case of psychogenic pain, a systematic review of the literature 36 showed the frequency of left-sided functional and motor symptoms was higher only in studies where the term laterality was featured in the title (i.e., “headlined”), and there were no incidental findings of laterality in non-headlined studies. Now, rather than highlighting potential differences in lateralization frequency, we instead surveyed the literature on pain-related affective experience. We found studies showing greater levels of psychological distress when pain is in the left versus the right hemibody in clinical spontaneous pain conditions ( Table 1 ). In contrast with ambiguous results regarding laterality of frequency, these studies have shown that pain from various disorders is typically accompanied by poorer emotional outcomes when symptoms are in the left hemibody. These measures include higher depressive and anxiety scores and lower health-related quality of life, among others, generally pointing to greater psychological distress with left-sided pain. Unfortunately, there is a paucity of literature on the subject (i.e., laterality of pain and mental health measures together), likely because painful disorders, especially chronic conditions, are nonetheless debilitating, regardless of location of pain. These studies indicate: (1) there is no robust evidence that left-sided pain would be observed more frequently than right-sided pain in the general population and (2) pain in the left hemibody evinces greater distress than does pain in the right hemibody. Of note, as well, is the apparently higher frequency of left-sided psychogenic pain. A prevailing explanation for this finding implicates the RH for its preferential role in pain and affective processing, as is discussed below, but it remains to be seen how such hemispheric specialization brings about the lateralized pain effects.

We reviewed literature on pain as it relates to depression and presented a model for lateralized pain-depression dyad. Importantly, there is no robust difference in the frequency of left- versus right-sided pain in the general population. Instead, and more pertinent, there are differential effects of laterality on mental health, such as left-sided pain being concomitant with greater psychological distress. Our lateralized pain-depression dyad model explains the apparently higher frequency of left-sided pain of psychogenic origin. It also provides the framework for the risk for depression in individuals with pain. Our model suggests that in individuals with depression, careful consideration should be given to novel symptoms in the left hemibody. Conversely, in patients with pain (except those with acute pain of known origin), referral for psychological evaluation should become standard practice and these psychological evaluations should be incorporated into treatment regimens. Often, comorbid depression is left undertreated despite being referred for specialist treatment 138 . Despite the fact that patients can only benefit from the addition of mental health professionals to their healthcare teams 139 , 140 , the implementation of a truly interdisciplinary approach, where patients and health experts from different disciplines work together in the pursuit of a common goal, remains a challenge (as opposed to multidisciplinary approach where individual experts may each have singular, but non-overlapping treatment goals) 141 . However, there are promising interdisciplinary approaches available to pediatric pain populations 142 – 144 after which pain centers for adults can be modeled.

We propose an updated model of the pain-depression dyad based on the evidence presented in the previous sections. Our model addresses existing gaps in the lateralization of the neural basis of pain and depression. First, we do not have a clear explanation as to why psychogenic pain more frequently manifests in the left hemibody. Second, we do not have a critically laid out thesis for how pain patients develop depression and vice versa; we simply observe an increased likelihood for one given the other. Here, we present two-way model ( Fig. 1 ) that accounts for the high incidence of left-sided psychogenic pain and also formulates the risks for pain and depression based on neural signatures. Fig. 1A is simply a graphical representation of the classic pain-depression dyad; that is, it shows the higher risk for comorbid pain and depression compared to the risk for pain or depression, separately. This graphic also works as a representation of the severity of symptoms. As to depression causing lateralized psychogenic pain ( Fig. 1B ), our model emphasizes a disruption in homeostasis of the RH. The changes to the active properties as a result of depression introduce widespread noise in the RH. Signaling in the right S1 is also affected, thereby creating the sensation of pain in the left hemibody. This dynamic is not unlike the nature of sensorimotor incongruence, where inconsistent signaling between the sensory and motor systems elicit sensory disturbances in healthy adults 125 . Consistent with our model, there are other frameworks that propose noisy signaling as the basis of disease burden 126 , 127 . Similarly, changes to the passive properties of the LH implicated in depression could also lead to right-sided psychogenic pain. However, passive properties tend to be localized (e.g., changes in cortical thickness of a specific region). Once a large enough volume of the cortex is affected, the alterations to the passive properties could ultimately lead to changes in active properties. This cascading effect in the LH, to a lesser degree and possibly more gradual progression than the RH, may then cause psychogenic pain in the right hemibody. Together, the differences in active versus passive property changes in each hemisphere may explain the differential incidence of left-versus right-sided pain. As to pain causing depression ( Fig. 1C ), our model relies on having lateralized pain and corresponding idiosyncratic neural changes to either hemisphere. Pain in the left hemibody must be accompanied by changes to the active properties of the RH or pain in the right hemibody must be accompanied by changes to the passive properties of the LH; such combinations (left side, Fig. 1C ) would increase the risk for depression. However, in contrast, if pain in the left hemibody is concomitant with changes to the passive properties of the RH or pain in the right hemibody with changes to the active properties of the LH, there would still be a risk for depression, albeit at a lower level (right side of Fig. 1C ). These changes, at different levels of risk, could elicit a depressive response, especially if compounded over time. In our model, we focused on the neural correlates of lateralized pain and depression. We attributed the preponderance of psychogenic pain in the left hemibody to differences in active and passive properties in the forebrain. It is worth noting here that we excluded the typically right-lateralized trigeminal neuralgia 27 in our discussions. We posit that the reversed laterality of orofacial pain may be explained in part by the differences in decussation of different nerves; whereas the somatic/spinal nerves have contralateral projections, cranial nerves, on the other hand, have ipsilateral branches. Thus, we limited this review to pain in either the left or right hemibody. Here, we address some concepts related to our model and other ideas related to neuroimaging of pain and depression. There could be explanations for the over-representation of pain in the left hemibody that are not cerebral in nature; one is physiological — the higher sensitivity to pain 128 , 129 that perhaps predisposes the left hemibody to experiencing painful sensations. This predisposition is manifested in terms of lower pressure-pain threshold (i.e., the minimum pressure which elicits a painful response) in the left than the right side of the body. Another explanation is the convenience hypothesis 130 , which suggests that the left and typically nondominant side is sacrificed to injury with the hopes of preserving the right dominant side. These two explanations raise the related issue of handedness. It has been shown that sinistrals exhibit more variable hemispheric specialization for language 131 , but it remains to be seen whether a reversal of lateralization for other functions and cognitive domains exist. This alternative hemispheric specialization in sinistrals opens up our model to scrutiny as it relates to pain and depression in left-handers, given the majority of the studies in this review involved mostly right-handed participants. Also, we intentionally did not factor in sex, the study of which is in itself wide-ranging and for which we refer the reader to recent reviews 132 – 134 . Looking ahead, our model can be tested and refined in the context of a carefully controlled prospective studies of lateralized disorders such as migraine and CRPS. In neuroimaging studies, not all studies use the same parcellation, or even the same granularity of parcellation. These differences in spatial resolution complicates integration of results, especially across domains (here, pain and depression), as there is an added layer of nuance to the roles regional subdivisions play. For example, noxious thermal stimuli elicited a response not in the entirety of the ACC but just in the anterior portion as well as the ventral portion of the posterior ACC 135 . Even in encoding the location of pain, only the posterior insula and the lateral thalamus, but not the anterior insula or the medial thalamus, showed somatotopic organization 136 . It is conceivable that different functions are localized to different subregions, and thus the neural substrate for pain may not even be strictly co-localized with that for depression. Nevertheless, it is known that functions and regions of the brain follow a many-to-one and also a one-to-many relationship (i.e., multiple regions subserve a single function, and a single region can be a substrate of multiple functions). In this regard, our model should fare reasonably well as we did not emphasize the role of specific regions, but rather focused on the differential roles of each hemisphere. Finally, it is not uncommon for published studies to present findings from a limited number of regions based on a priori hypotheses. However, the consistent involvement of key pain-related regions in earlier studies can set a strong precedent resulting in a proliferation of studies showing effects in always the same regions. A recent study involving clinical deficits in CRPS (e.g., poor tactile acuity due to pain) investigated the connectivity between, unsurprisingly, the S1 and the thalamus 137 . In this case, a hypothesis-driven approach limited the analysis to circumscribed areas of the brain. While performing whole-brain analyses has the potential to reveal effects in regions other than those widely documented, the difficulty in integrating and interpreting results from vast swathes of the brain in relation to pathology must be recognized.

The revised definition of pain from the International Association for the Study of Pain (“an unpleasant sensory and emotional experience associated with, or resembling that associated with, actual or potential tissue damage” 1 ) underscores the multidimensional nature of pain and inextricably links the physical symptoms with psychological well-being. Indeed, the prevalence of comorbid pain and depression is significantly higher than that of pain or depression when either condition is considered independently 2 , 3 . Importantly, a bidirectional longitudinal relationship between pain and depression exists based on the high likelihood of developing depression secondary to chronic pain, and vice versa 4 – 6 . This interconnected nature of pain and depression — or the pain-depression dyad 7 , 8 , as it has been called — can be facilitated through a shared neural circuitry. The primary somatosensory cortex (S1) in each hemisphere encodes the location of pain experienced in the contralateral hemibody; that is, pain in the left hemibody is represented in the somatotopically organized S1 of the right hemisphere (RH) while pain in the right hemibody is similarly represented in the S1 of the left hemisphere (LH). Other regions consistently associated with pain processing include the thalamus, insula, and anterior cingulate cortex (ACC), based on a meta-analytic study of acute and chronic pain 9 . Interestingly, these regions have also been implicated in depression 10 , 11 . While these regions are bilaterally distributed, depression is generally considered to be a dysfunction of the RH, manifested in terms of its functional insufficiency or physiological hyperactivity 12 – 14 relative to the LH. This hemispheric asymmetry begs the question: what are the consequences of pain on either side of the body in relation to depressive symptoms? Here, we review existing literature on pain and depression, with particular emphasis on the laterality of the somatic and the neural phenotypes and present a theoretical framework for the involvement of forebrain regions in pain and depression. First, we evaluate this thesis in the context of left and right hemibody pain with differential observations of frequency and of psychological distress. Second, we scrutinize the overlapping neural substrates of pain and depression. We review the hemispheric lateralization of the neural correlates of experimentally evoked pain and of clinical spontaneous pain. We also present neural correlates of depression in specific regions known to be typically responsive to pain. Finally, we propose how these themes can be integrated. We synthesize pertinent findings into an idiosyncratic model of how pain gives rise to depression, and vice versa.