​Individualized care in chronic pain research   June 7th, 2023

Chronic pain is defined as pain that persists for more than three months and currently affects over 20% of people worldwide, making it the leading cause of disability and disease. Despite its prevalence, optimal treatment for chronic pain is still lacking [1], [2]. A recent hype in pain research is to improve treatment outcomes based on a more personalized approach, in which terms like stratified care, matched care, individualized care, personalized medicine, precision medicine, etc. are becoming increasingly popular [3]. They all include a way of tailoring a treatment to (a group of) individuals. Reading these terms out loud sounds like they are almost synonyms, but are they really?

Different terms of tailored treatment

To date, the definitions of aforementioned terms are still under debate, but the main goal of all is to optimize treatment benefit of individuals [3]. An attempt to differentiate between these terms will be made in this paragraph. Stratified and matched care mean that individuals are divided into subgroups based on similar characteristics first, but the difference lays in tailoring the treatment to the subgroups. In stratified care, individuals within the same subgroup receive the same treatment based on their specific characteristics, while in matched care every individual in the subgroup will be assessed thoroughly and treated with individualized care accordingly [4]. As such, individualized care means that caregivers first learn about the patient as a unique individual and tailor their intervention to their specific needs, abilities and wants [5].

Next, also precision and personalized medicine have a lot of overlap. Personalized medicine is seen as an ‘older’ term and changed to precision medicine later on, because personalized medicine was thought to denote a unique treatment for every individual, while precision medicine is based on finding the most effective treatment based on genetic, environmental and lifestyle factors of individuals [6].

Despite attempts to differentiate these terms, overlap and confusion remains present …

How to divide individuals in subgroups in research?

As mentioned before, in order to perform stratified or matched care,individuals need to be classified into subgroups based on shared characteristics. This process is often referred to as phenotyping. To date, several statistical methods for phenotyping individuals already exist and can include cluster analysis or latent class/profile analysis.

The two most popular cluster analysis techniques are hierarchical and K-means clustering. The main difference between these two is that in K-means clustering the number of phenotypes is chosen first, based on advance knowledge of the number of phenotypes to be expected, and then individuals are divided in that number of phenotypes. In contrast, hierarchical clustering has no predetermined number of phenotypes and determines phenotypes based on a visual dendrogram, which has a ‘tree’ structure. Both methods are based on similarities between observed continuous data of individuals, but K-means clustering requires prior knowledge of the number of phenotypes, while hierarchical clustering can determine the appropriate number of phenotypes based on the data itself.

In latent class or profile analysis individuals will be grouped based on their unobserved heterogeneity into ‘latent (i.e. hidden variable) classes (i.e. phenotypes)’. Latent class analysis is based on categorical variables, while latent profile analysis can include both continuous and categorical variables. Advantages of latent class/profile analysis compared to cluster analysis include the allowance for the estimation of model fit (based on e.g. Bayesian Information Criterion), the handling of missing data, the estimate of probability of group membership per individual and the inclusion of covariates. This is of course a very concise description, but can give you a first impression when choosing a specific phenotype-method for your research [7], [8].

Recommendations for chronic pain patient phenotyping

The Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) has provided recommendations for the variables that should be used in phenotyping chronic pain patients. These variables include psychosocial factors, symptom characteristics, sleep patterns, responses to noxious stimulations, endogenous pain-modulatory processes and response to pharmacologic challenge. A detailed overview of the recommendations can be found in the article of Edwards et al. [9].

In summary, when it comes to treating chronic pain, there seems to be no one-size-fits-all solution. But with stratified, matched and individualized care, we can at least make sure the treatment fits the individual or subgroup...and not the other way around. However, further research in individualized care and its efficacy is still necessary.

Sophie Vervullens

Physiotherapist and PhD student at the University of Antwerp (MOVANT) and the University of Maastricht (CAPHRI).

2023Pain in Motion

References and further reading:

[1] R.-D. Treede et al., ‘A classification of chronic pain for ICD-11’, Pain, vol. 156, no. 6, pp. 1003–1007, Jun. 2015, doi: 10.1097/j.pain.0000000000000160.

[2] G. 2016 D. and I. I. and P. Collaborators, ‘Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016’, Lancet Lond. Engl., vol. 390, no. 10100, p. 1211, Sep. 2017, doi: 10.1016/S0140-6736(17)32154-2.

[3] S. Erikainen and S. Chan, ‘Contested futures: envisioning “Personalized,” “Stratified,” and “Precision” medicine’, New Genet. Soc., vol. 38, no. 3, pp. 308–330, Jul. 2019, doi: 10.1080/14636778.2019.1637720.

[4] S. J. Linton, M. Nicholas, and W. Shaw, ‘Why wait to address high-risk cases of acute low back pain? A comparison of stepped, stratified, and matched care’, PAIN, vol. 159, no. 12, p. 2437, Dec. 2018, doi: 10.1097/j.pain.0000000000001308.

[5] L. E. Radwin and K. Alster, ‘Individualized nursing care: an empirically generated definition’, Int. Nurs. Rev., vol. 49, no. 1, pp. 54–63, 2002, doi: 10.1046/j.1466-7657.2002.00101.x.

[6] ‘What is the difference between precision medicine and personalized medicine? What about pharmacogenomics?: MedlinePlus Genetics’. https://medlineplus.gov/genetics/understanding/pre... (accessed Apr. 04, 2023).

[7] D. Spurk, A. Hirschi, M. Wang, D. Valero, and S. Kauffeld, ‘Latent profile analysis: A review and “how to” guide of its application within vocational behavior research’, J. Vocat. Behav., vol. 120, p. 103445, Aug. 2020, doi: 10.1016/j.jvb.2020.103445.

[8] J. Magidson and J. Vermunt, ‘Latent class models for clustering: A comparison with K-means. Canadian J Mark Res’, Prof. Mark. Res. Soc., vol. 20, pp. 36–43, Jan. 2002.

[9] R. R. Edwards et al., ‘Patient phenotyping in clinical trials of chronic pain treatments: IMMPACT recommendations’, PAIN Rep., vol. 6, no. 1, p. e896, 2021, doi: 10.1097/PR9.0000000000000896.