The number of central nervous system-driven symptoms predicts subsequent chronic primary pain. Response to Br J Anaesth 2025; 135: 493–4 and 135: 495–6

Editor

We thank Kou and colleagues¹ and Zeng and Zhou² for their thoughtful responses to our recent article exploring how CNS-driven symptoms including sleep disturbance, mood dysfunction, and cognitive impairment predict subsequent development of chronic primary pain in a pain-free cohort from the UK Biobank.³

Kou and colleagues raise three key methodological considerations: subgroup analysis, representativeness, and missing data handling. We appreciate their suggestion to examine potential effect modification by demographic variables. Although our models adjusted for age, sex, and socioeconomic status, this assumes these factors confound the relationship rather than modify it. Age, for example, exhibits a complex relationship with pain, with fibromyalgia often peaking in midlife,⁴ and degenerative conditions more prevalent later in life. However, given the relatively narrow recruitment age in the UK Biobank of 40–70 (median 58) yr, we have limited power to detect interactions across broader age ranges, such as in much younger or older adults.

Sex differences in pain perception, reporting, and CNS symptomatology are well documented. Males and female differ in pain-induced connectivity between the periaqueductal grey and amygdala, brain regions implicated in affective and cognitive pain processing.⁵ These differences warrant future exploration.

The role of socioeconomic status is complex. Socioeconomic deprivation is associated with increased risk of nociplastic pain,⁶ potentially through cumulative life stressors, yet pain itself might also impair educational and occupational attainment, raising questions about directionality.⁷ Although subgroup analyses might yield further insight, they also risk overcomplicating interpretation or identifying spurious associations in a large dataset.

We also agree with Kou and colleagues that the demographic composition of the UK Biobank limits the generalisability of our findings. Participants tend to be healthier, wealthier, and more educated than the general population.⁸ However, multiple studies have shown that although baseline prevalence estimates can differ, exposure–outcome relationships within the UK Biobank are broadly consistent with those observed in more representative samples.⁸ Notably, the prevalence of chronic pain and the predominance of female participants in our cohort align with external datasets.⁹ Although inverse probability weighting might help adjust for selection bias, this method sacrifices statistical power,¹⁰ and the qualitative direction of our findings is likely generalisable.

On the issue of missing data, we conducted a complete case analysis as <5% of eligible participants had missing baseline or outcome data owing to incomplete data collection. However, a larger proportion were lost to follow-up, which could introduce selection bias. Although multiple imputation by chained equations (MICE) is often used to address missingness, it relies on data being missing completely at random (MCAR) or at random (MAR), assumptions unlikely to hold in this setting. For example, participants who did not complete the follow-up pain questionnaire were more deprived and reported worse health, suggesting that missingness might not be random. Imputation techniques such as MICE do not resolve bias when data are missing not at random (MNAR). Sensitivity analyses in other UK Biobank studies that addressed both selection bias and missing data have shown minimal impact on associations.⁸

Zeng and Zhou offer a more mechanistic and conceptual response, which we also welcome. They rightly suggest that the relationship between CNS symptoms and primary pain is likely bidirectional. Although our analysis excluded individuals with chronic or recent pain at baseline and applied a 6-month lag to mitigate reverse causality, we acknowledge that prior painful experiences could shape CNS vulnerability. Indeed, work from our group using the ABCD study has demonstrated similar predictive relationships between attentional issues and sleep disturbance and new multisite pain in adolescents.¹¹ Nevertheless, the clinical implication remains clear: among adults who do not report current pain, the burden of CNS symptoms predicts future nociplastic pain, and might thus represent a modifiable risk factor.

Zeng and Zhou also raise the question of how pharmacological treatments for CNS symptoms might influence pain outcomes. A recent preclinical study in mice suggests that adolescent exposure to fluoxetine, a selective serotonin reuptake inhibitor, might increase thermal pain sensitivity in adulthood.¹² However, the preponderance of evidence in humans supports an analgesic effect of antidepressants in clinical use.¹³ Disentangling the effects of these medications from those of the underlying conditions they are prescribed for remains challenging, particularly in non-randomised settings. Nonetheless, many agents, especially serotonin–norepinephrine reuptake inhibitors (SNRIs) and pregabalin, have demonstrated efficacy in alleviating symptoms, including pain, in fibromyalgia and other chronic pain conditions.^14,15 We agree that further research is needed to understand how both pharmacological and nonpharmacological symptom management strategies influence the risk of central sensitisation over time. We also support their emphasis on nonpharmacological approaches such as cognitive behavioural therapy and exercise, which have shown consistent benefit in reducing both symptom burden and central sensitisation.

We also agree with Zeng and Zhou that shared pathophysiology, rather than direct causality, might underlie CNS symptoms and nociplastic pain. CNS dysfunction might not merely be a consequence of chronic pain, but rather a core driver (which we refer to elsewhere as ‘top-down’ mechanisms¹⁶). This does not undermine the predictive value of these symptoms but reinforces their role as early markers of an at-risk neurobiological state.

Finally, Zeng and Zhou propose incorporating tools such as the Central Sensitization Inventory (CSI)¹⁷ or the Fibromyalgia Rapid Screening Tool (FiRST)¹⁸ into routine screening. Although these instruments are valuable for identifying nociplastic pain in people with established symptoms, they have not been validated in asymptomatic individuals. Our study suggests that a simple count of three CNS symptoms (sleep disturbance, mood problems, and cognitive dysfunction) serves as a pragmatic and clinically useful risk indicator in the general population.

In conclusion, we thank both groups for their thoughtful contributions. Our findings suggest that CNS symptoms, even in the absence of current pain, identify individuals at increased risk for future nociplastic pain. Although we agree that the relationships are complex, likely bidirectional, and shaped by underlying neurobiology, our work supports the value of early identification and targeted intervention. Future research should focus on prospective interventional studies aimed at treating sleep, mood, and cognitive dysfunction to test whether such approaches can prevent or delay the onset of chronic primary pain.

Funding

National Institute for Health Research (NIHR) Biomedical Research Centre, Oxford, UK; NIHR Pfizer Doctoral Fellowship (NIHR301808 to EK). This paper presents independent research funded by the National Institute for Health Research (NIHR) and Pfizer. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, the Department of Health and Social Care, or Pfizer.

Declarations of interest

EK was supported by a National Institute for Health Research (NIHR) Pfizer Doctoral Fellowship for this research project (NIHR301808). DK, AS, and IT have no interests to declare. DC and CK are supported by a US National Institutes of Health (NIH) grant (Bethesda, MD, USA). AI received funding from UCB, unrelated to the current project, during the time of the project.

Handling Editor: Hugh C Hemmings Jr