A study looking at peripheral neuropathy in people living with HIV has found links to both known and new variables including older age, longer duration of HIV infection, exposure to neurotoxic antiretrovirals and other chronic conditions. Dr Wei Tu’s research team at the University of Alberta used machine learning to improve data analysis. They found that the associated factors for peripheral neuropathy in those diagnosed with HIV more than 15 years ago were different to those diagnosed less than 15 years ago.
Peripheral neuropathy is a common condition associated with HIV, affecting 21% in this study, which is a similar rate to other studies globally. Peripheral neuropathy refers to nerve damage that starts furthest from the brain in the feet or hands, but can also affect the legs and arms. Symptoms may be milder in the form of tingling and numbness, or may include walking difficulties and severe pain. There are multiple different types of damage that can occur, which makes it difficult to determine which factors are causing peripheral neuropathy. However, the resulting chronic pain and disability can worsen a person’s quality of life and ability to function, making peripheral neuropathy an important issue.
In this study published in AIDS, machine learning allowed for huge amounts of information to be analysed. This enabled better understanding of the variables associated with different types of peripheral neuropathy. Understanding these variables could lead to better ways of diagnosing, preventing and treating peripheral neuropathy in people living with HIV.
People living with HIV at the Southern Alberta Clinic in Canada were regularly assessed for nerve disorders at their routine clinical appointments. In total 519 participants were enrolled between 2013-2019. The participants’ demographics and clinical data were then analysed using univariate and multivariate methods, and then compared to machine learning analysis.
Overall 111 (21%) were found to have some form of peripheral neuropathy which was split into two groups: those with damage to multiple nerves affecting their sensation (17%) and those with damage to specific single nerves (4%). Damage to multiple nerves included those with more generalised burning or stabbing pain, loss of feeling and weakness. Specific nerves were identified by those with symptoms consistent with things like carpal tunnel syndrome, which affects one nerve in the hand.
Univariate analysis indicated that peripheral neuropathy was associated with 28 different variables, including age, diabetes, substance misuse and exposure to nerve damaging antiretroviral therapy. The antiretrovirals known to cause peripheral neuropathy were the ‘d-drugs’ didanosine (ddI), zalcitabine (ddC) and stavudine (d4T), which are no longer used. However, past use impacted the likelihood of peripheral neuropathy in this study, despite not being apparent in some previous studies. Links were also found to higher peak viral loads, lower lowest ever CD4 counts and less sleep.
A longer duration of HIV infection increased the likelihood of peripheral neuropathy when using any model of analysis. There was a higher prevalence of nerve damage (40%) in those living with HIV for more than 15 years, compared to those for less than 15 years (11%). Analysis with machine learning showed that risk factors associated with peripheral neuropathy were different for those diagnosed more than 15 years ago, compared to those diagnosed more recently.
For people living with HIV for more than fifteen years, associations included diabetes, syphilis, cardiovascular disease, increased viral load and use of nerve damaging ART medications.
For those diagnosed less than fifteen years ago associated variables were worse self-rating scores on mental health questionnaires, cigarette use, less sleep and less education. Medications like vincristine (used for chemotherapy) and nitroglycerin (often used for cardiovascular disease) were also linked to peripheral neuropathy in this group.
Neuropathic pain, use of more than five medications (not including antiretrovirals) and exposure to a medication to treat neuropathic pain (pregabalin) were linked to peripheral neuropathy irrespective of the time since HIV diagnosis.
Given the amount of available data, machine learning could be used going forward to spot patterns that would be difficult to detect using other analysis techniques. While peripheral neuropathy can affect anyone and is linked with many conditions, it is a common condition for people living with HIV and contributes to significant disability and pain. The researchers noted that “a deeper understanding of both pathogenesis and clinical factors defining peripheral neuropathy is required for addressing this issue.”
Wei T et al. Predictive variables for peripheral neuropathy in treated HIV-1 infection revealed by machine learning. AIDS, online ahead of print, 24 May 2021.