ChatGPT and its Role in the Decision-Making for the Diagnosis and Treatment of Lumbar Spinal Stenosis: A Comparative Analysis and Narrative Review

Abstract

Study Design

Comparative Analysis and Narrative Review.

Objective

To assess and compare ChatGPT’s responses to the clinical questions and recommendations proposed by The 2011 North American Spine Society (NASS) Clinical Guideline for the Diagnosis and Treatment of Degenerative Lumbar Spinal Stenosis (LSS). We explore the advantages and disadvantages of ChatGPT’s responses through an updated literature review on spinal stenosis.

Methods

We prompted ChatGPT with questions from the NASS Evidence-based Clinical Guidelines for LSS and compared its generated responses with the recommendations provided by the guidelines. A review of the literature was performed via PubMed, OVID, and Cochrane on the diagnosis and treatment of lumbar spinal stenosis between January 2012 and April 2023.

Results

14 questions proposed by the NASS guidelines for LSS were uploaded into ChatGPT and directly compared to the responses offered by NASS. Three questions were on the definition and history of LSS, one on diagnostic tests, seven on non-surgical interventions and three on surgical interventions. The review process found 40 articles that were selected for inclusion that helped corroborate or contradict the responses that were generated by ChatGPT.

Conclusions

ChatGPT’s responses were similar to findings in the current literature on LSS. These results demonstrate the potential for implementing ChatGPT into the spine surgeon’s workplace as a means of supporting the decision-making process for LSS diagnosis and treatment. However, our narrative summary only provides a limited literature review and additional research is needed to standardize our findings as means of validating ChatGPT’s use in the clinical space.

Introduction

Spine surgeons are tasked with making polychotomous decisions every day when it comes to providing effective patient care. As spine research and technology continue to advance, the role of guidelines has become prominent in the surgeon’s decision-making process. ¹ Degenerative lumbar spinal stenosis is one such spinal etiology that continues to advance in terms of accepted practices for its diagnosis and treatment. ² In 2011, The North American Spine Society (NASS) Clinical Guideline for the Diagnosis and Treatment of Degenerative Lumbar Spinal Stenosis attempted to alleviate differences in practices by evaluating the highest quality literature and formulating evidence-based recommendations through expert consensus and analysis. ³ However, as new research continues to emerge and outpace the development of updated guidelines, a gap has formed between what is recommended and what is possible with the latest advancements.

Recently, Chat Generative Pre-trained Transformer (ChatGPT) has emerged as a potential resource to address these challenges. ChatGPT is an artificial intelligence (AI) language model that utilizes deep learning to generate responses to human prompts on command. The latest version of ChatGPT (GPT-3.5) was trained on 45 terabytes of text data, including news articles, books, journals, and other sources that allow it to readily respond to a wide array of queries. Thus, the advent of ChatGPT has led to its interest in its ability to synthesize medical knowledge and make clinical decisions.^4-6

The purpose of this paper is to analyze and compare ChatGPT’s recommendations to the clinical questions and recommendations proposed by the NASS Clinical Guideline for the Diagnosis and Treatment of Degenerative Lumbar Spinal Stenosis. Our study aims to qualitatively evaluate ChatGPT’s performance when responding to clinical questions related to lumbar spinal stenosis and provide a review of the literature to corroborate or refute its findings.

Methods

16 questions were obtained from the most recent NASS Clinical Guideline for the Diagnosis and Treatment of Degenerative Lumbar Spinal Stenosis, revised in 2011. All questions were screened and then imputed into ChatGPT with no modifications. Two questions from the guidelines were excluded since the NASS guidelines did not report an answer or any evidence supporting their claim.

For the included 14 questions, a new ChatGPT window was created for each question in order to avoid any biases from the prior questions. After ChatGPT generated a response, it was recorded verbatim onto our database. The answers from ChatGPT were then compared to those presented by the guidelines and were assessed for accuracy based on what the guidelines had presented.

For our narrative review, we searched the PubMed, Ovid, and Cochrane research databases between January 2012 and April 2023 for articles about diagnosis and treatment of lumbar spinal stenosis that answered the questions proposed by the NASS guidelines. Our search strategy included key words such as lumbar spinal stenosis or spinal stenosis, surgery, decompression, fusion, medication, opioids, epidural injections, steroids, physical therapy, exercise, manipulation, bracing, diagnostic test, history, and clinical guidelines. Eligible studies (1) evaluated adult patients (≥18 years) who had undergone an operative or non-operative intervention for LSS; (2) reported postoperative outcomes that were related to the guidelines’ questions; and (3) included data that analyzed changes in function or outcomes compared to baseline. All studies before 2012 were excluded since these findings were already incorporated into the 2011 NASS guidelines. We excluded animal studies, case reports, and articles not available in English or that did not provide data related to the questions posed by the guidelines. Two reviewers independently screened the titles and abstracts of the 4487 citations retrieved from the initial search based on our inclusion criteria. Afterwards, two reviewers independently screened the included full-text articles. Conflicts between the reviewers were resolved by a third investigator. After performing full-text screening, 40 studies were considered eligible.

Informed consent was not required for this project, and there was no need for Institutional Review Board approval as no patient data was involved.

Guideline Questions Characteristics

There were a total of 14 questions from the NASS Evidence-Based Clinical Guidelines for Lumbar Spinal Stenosis that were assessed in this study. There were three questions related to the definitions and history of progression, one question about diagnostic tests, seven questions that assessed non-surgical interventions, and three questions that summarized surgical interventions. Nine questions proposed by the guidelines were given recommendations by both the guidelines and ChatGPT. (Table 1) There were five questions proposed in which the guidelines did not have sufficient evidence to make a recommendation, but were provided recommendations by ChatGPT. (Table 2)

Table 1.

NASS Guideline recommendations for Lumbar Stenosis vs ChatGPT. Questions and recommendations from the NASS guidelines recorded for responses that had sufficient evidence to make a claim. Responses from ChatGPT were also recorded and graded for precision.

Question	NASS	Chat GPT
What is the best working definition of degenerative lumbar spinal stenosis?	Degenerative lumbar spinal stenosis describes a condition in which there is diminished space available for the neural and vascular elements in the lumbar spine secondary to degenerative changes in the spinal canal. When symptomatic, this causes a variable clinical syndrome of gluteal and/or lower extremity pain and/or fatigue which may occur with or without back pain. Symptomatic lumbar spinal stenosis has certain characteristic provocative and palliative features. Provocative features include upright exercises such as walking or positionally-induced neurogenic claudication. Palliative features commonly include symptomatic relief with forward flexion, sitting and/or recumbency	Degenerative lumbar spinal stenosis refers to a condition where there is a narrowing of the spinal canal in the lower back due to degenerative changes in the spine. This narrowing can put pressure on the spinal cord and nerves, leading to a variety of symptoms, such as pain, numbness, weakness, and difficulty walking
What is the natural history of degenerative lumbar spinal stenosis?	“In the absence of reliable evidence, it is the work group’s opinion that the natural history of patients with clinically mild to moderately symptomatic degenerative lumbar stenosis can be favorable in about one-third to one-half of patients In the absence of reliable evidence, it is the work group’s opinion that in patients with mild or moderately symptomatic degenerative lumbar stenosis, rapid or catastrophic neurologic decline is rare In the absence of reliable evidence, it is the work group’s opinion that information in the literature is insufficient to define the natural history of clinically or radiographically severe degenerative lumbar stenosis.”	“The symptoms of lumbar spinal stenosis typically develop gradually and may include lower back pain, weakness, numbness, and tingling sensations in the legs, buttocks, and feet. These symptoms often worsen when the person is standing or walking, and may be relieved when they sit or lie down In some cases, lumbar spinal stenosis may remain stable for many years, with minimal or no symptoms. In other cases, the condition may progress and cause more severe symptoms, such as difficulty walking, loss of bladder or bowel control, and even paralysis Treatment for lumbar spinal stenosis typically starts with non-surgical options, such as physical therapy, anti-inflammatory medications, and steroid injections. If these methods fail to provide relief, surgery may be considered to decompress the spinal canal and relieve pressure on the nerves It is important to note that the natural history of lumbar spinal stenosis can be influenced by various factors, such as age, overall health, and lifestyle habits. Regular exercise, weight management, and smoking cessation may help to slow the progression of the condition and improve overall health outcomes.”
What are the most appropriate diagnostic tests for degenerative lumbar spinal stenosis?	“In patients with history and physical examination findings consistent with degenerative lumbar spinal stenosis, MRI is suggested as the most appropriate, noninvasive test to confirm the presence of anatomic narrowing of the spinal canal or the presence of nerve root impingement In patients with a history and physical examination findings consistent with degenerative lumbar spinal stenosis, for whom MRI is either contraindicated or inconclusive, CT myelography is suggested as the most appropriate test to confirm the presence of anatomic narrowing of the spinal canal or the presence of nerve root impingement In patients with a history and physical examination, findings consistent with degenerative lumbar spinal stenosis for whom MRI and CT myelography are contraindicated, inconclusive or inappropriate, CT is the preferred test to confirm the presence of anatomic narrowing of the spinal canal or the presence of nerve root impingement MRI or CT with axial loading is suggested as a useful adjunct to routine imaging in patients who have clinical signs and symptoms of lumbar spinal stenosis, a dural sac area (DSA) of less than 110mm2 at one or more levels, and suspected but not verified central or lateral stenosis on routine unloaded MRI or CT. It is suggested that readers use well-defined, articulated and validated criteria for anatomic canal narrowing on MRI, CTM and CT to improve interobserver and intraobserver reliability There is insufficient evidence to make a recommendation for or against a correlation between clinical symptoms or function with the presence of anatomic narrowing of the spinal canal on MRI, CTM or CT. In the absence of reliable evidence, it is the work group’s opinion that imaging studies be considered as a first-line diagnostic test in the diagnosis of degenerative lumbar spinal stenosis Electromyographic paraspinal mapping is suggested to confirm the diagnosis of degenerative lumbar spinal stenosis in patients with mild or moderate symptoms and radiographic evidence of stenosis There is insufficient evidence to make a recommendation for or against the use of F wave, H reflex, motor evoked potential (MEP), motor nerve conduction studies, somatosensory evoked potentials (SSEP), dermatomal sensory evoked potentials (DSEP) and lower extremity EMG in the confirmation of lumbar spinal stenosis. These studies may be used to help identify other comorbidities”	“The most appropriate diagnostic tests for degenerative lumbar spinal stenosis include: Imaging tests: Imaging tests are used to visualize the spinal canal and identify any abnormalities. The most common imaging tests used to diagnose degenerative lumbar spinal stenosis include:  X-rays: These can show changes in the shape and alignment of the spine, such as narrowing of the spinal canal, bone spurs, or a loss of disc height  Magnetic resonance imaging (MRI): This is a non-invasive test that uses powerful magnets and radio waves to create detailed images of the spinal cord and surrounding structures. It can show the extent of the stenosis and whether there is any compression of the nerves  Computed tomography (CT) scan: This is a test that uses X-rays and computer technology to create detailed images of the spine. It can show bony abnormalities, such as bone spurs or herniated discs, that may be causing the stenosis Nerve function tests can help evaluate the function of the nerves in the legs and determine if there is any nerve damage or compression. The most common nerve function tests used to diagnose degenerative lumbar spinal stenosis include:  Electromyography (EMG): This is a test that uses small needles to measure the electrical activity of the muscles and nerves in the legs. It can show if there is any nerve damage or compression and the severity of the damage  Nerve conduction studies (NCS): This is a test that measures how quickly electrical impulses travel through the nerves in the legs. It can show if there is any nerve damage or compression and the severity of the damage Overall, a combination of imaging tests and nerve function tests is typically used to diagnose degenerative lumbar spinal stenosis and evaluate its severity. It is important to consult with a healthcare professional to determine the most appropriate diagnostic tests based on individual circumstances.”
What is the role of contrast-enhanced, fluoroscopic guidance in the routine performance of epidural steroid injections for the treatment of lumbar spinal stenosis?	Contrast-enhanced fluoroscopy is recommended to guide epidural steroid injections to improve the accuracy of medication delivery	“Contrast-enhanced, fluoroscopic guidance is an imaging technique used to guide epidural steroid injections (ESIs) for the treatment of lumbar spinal stenosis. This technique allows for real-time visualization of the needle and the injected medication, ensuring accurate placement of the needle in the epidural space and the precise delivery of the medication ESIs are commonly used to treat this condition, as they can help relieve pain, reduce inflammation, and improve function. However, accurate needle placement is crucial for the success of the injection, and fluoroscopic guidance helps ensure this accuracy During the procedure, a contrast agent is injected into the epidural space to help visualize the space and ensure accurate needle placement. Fluoroscopy is then used to guide the needle to the correct location, and the steroid medication is injected. Contrast-enhanced, fluoroscopic guidance has become a routine part of the performance of epidural steroid injections for the treatment of lumbar spinal stenosis. It has been shown to improve the accuracy and safety of the procedure, reducing the risk of complications such as nerve injury, infection, or bleeding. It also increases the effectiveness of the injection by ensuring that the medication is delivered to the right location.”
What is the role of epidural steroid injections (ESI) in the treatment of spinal stenosis?	“Interlaminar epidural steroid injections are suggested to provide short-term (two weeks to 6 months) symptom relief in patients with neurogenic claudication or radiculopathy. There is, however, conflicting evidence concerning long-term (21.5-24 months) efficacy A multiple injection regimen of radiographically-guided transforaminal epidural steroid injection or caudal injections is suggested to produce medium-term (3-36 months) relief of pain in patients with radiculopathy or neurogenic intermittent claudication (NIC) from lumbar spinal stenosis.”	“Epidural steroid injections (ESI) are a common non-surgical treatment option for spinal stenosis. Spinal stenosis is a condition in which the spinal canal narrows, putting pressure on the spinal cord and nerves. This can cause pain, numbness, and weakness in the legs or arms, depending on the location of the stenosis ESI involves injecting a corticosteroid medication into the epidural space, which is the space around the spinal cord and nerve roots. The corticosteroid medication helps to reduce inflammation and swelling in the area, which can relieve pressure on the nerves and provide pain relief ESI is typically used as a short-term treatment option for spinal stenosis, as it can provide temporary relief of symptoms. The injections are usually given as a series of two to three injections spaced out over a few weeks to months, depending on the patient's response and the severity of their symptoms It is important to note that ESI is not a cure for spinal stenosis and does not address the underlying cause of the condition. It is generally used in conjunction with other non-surgical treatments such as physical therapy, exercise, and pain medications to manage symptoms and improve function. Surgery may be necessary in more severe cases of spinal stenosis that do not respond to conservative treatments.”
What is the long-term (two to 10 years) result of medical/interventional management of spinal stenosis?	Medical/interventional treatment may be considered to provide long-term (2-10 years) improvement in patients with degenerative lumbar spinal stenosis and has been shown to improve outcomes in a large percentage of patients	The long-term result of medical/interventional management of spinal stenosis can vary depending on various factors, including the severity of the condition, the specific treatment approach, and the individual patient's health and lifestyle In general, conservative treatment options for spinal stenosis, such as physical therapy, pain management, and exercise, may provide some relief of symptoms but do not necessarily address the underlying cause of the condition. These approaches may be effective in managing symptoms for some patients over the long term, but for others, they may not provide significant or sustained relief Interventional treatments for spinal stenosis, such as epidural steroid injections or surgical decompression, may provide more substantial relief of symptoms in some cases. However, these interventions also carry some risks and are not appropriate or necessary for all patients In general, the long-term outcome of medical/interventional management of spinal stenosis is difficult to predict and can vary widely depending on the individual patient's circumstances. Some patients may experience significant improvement in symptoms and quality of life, while others may experience only modest or temporary relief of symptoms. It is important for patients to work closely with their healthcare provider to determine the best course of treatment for their specific condition and individual needs.”
Does surgical decompression alone improve surgical outcomes in the treatment of spinal stenosis compared to medical/interventional treatment?	“Decompressive surgery is suggested to improve outcomes in patients with moderate to severe symptoms of lumbar spinal stenosis Medical/interventional treatment may be considered for patients with moderate symptoms of lumbar spinal stenosis In the absence of evidence for or against any specific treatment, it is the work group’s recommendation that medical/interventional treatment be considered for patients with mild symptoms of lumbar spinal stenosis There is insufficient evidence at this time to make a recommendation for or against the placement of an interspinous process spacing device in patients with lumbar spinal stenosis.”	“Surgical decompression is often considered the gold standard treatment for spinal stenosis when conservative management fails to relieve symptoms. The primary goal of surgical decompression is to remove the compression on the spinal cord or nerve roots, which can lead to pain, numbness, or weakness in the affected area Several studies have compared the outcomes of surgical decompression with conservative management, such as medications, physical therapy, or injections. Overall, surgical decompression has been found to be more effective in relieving symptoms and improving function in patients with spinal stenosis For example, a 2013 systematic review of randomized controlled trials found that surgical decompression was superior to non-surgical treatments in improving pain, function, and quality of life in patients with lumbar spinal stenosis. Similarly, a 2015 study found that surgery was associated with greater improvements in physical function and less pain compared to non-surgical management However, it is important to note that surgery carries potential risks and complications, and it is not appropriate for every patient with spinal stenosis. The decision to pursue surgical decompression should be made on an individual basis, taking into account factors such as the severity of symptoms, the presence of other medical conditions, and the patient's preferences In summary, surgical decompression has been found to be more effective than conservative management in improving symptoms and function in patients with spinal stenosis. However, the decision to pursue surgery should be made in consultation with a qualified healthcare professional and should consider individual factors and risks.”
Does the addition of lumbar fusion, with or without instrumentation, to surgical decompression improve surgical outcomes in the treatment of spinal stenosis compared to treatment by decompression alone?	“Decompression alone is suggested for patients with leg predominant symptoms without instability”	“The addition of lumbar fusion, with or without instrumentation, to surgical decompression, may improve surgical outcomes in certain cases of spinal stenosis compared to treatment by decompression alone. However, the decision to perform fusion should be based on individual patient factors, such as the extent of spinal instability or the presence of spondylolisthesis, and the potential risks and benefits of the procedure should be carefully weighed There is some evidence to suggest that fusion may improve outcomes in patients with spinal stenosis who have associated spondylolisthesis or significant spinal instability. A systematic review and meta-analysis of randomized controlled trials published in the Journal of Bone and Joint surgery found that patients with spondylolisthesis who underwent fusion in addition to decompression had greater improvements in pain and function compared to those who underwent decompression alone However, in patients without significant instability, the addition of fusion may not improve outcomes and may even lead to increased risks and complications. A systematic review and meta-analysis published in spine found that in patients with spinal stenosis without spondylolisthesis or instability, the addition of fusion did not improve outcomes and was associated with a higher risk of complications In summary, the decision to perform fusion in addition to decompression for the treatment of spinal stenosis should be made on a case-by-case basis, taking into account individual patient factors and weighing the potential risks and benefits of the procedure.”
What is the long-term result (4+ years) of surgical management of spinal stenosis?	“Surgical treatment may be considered to provide long-term (4+ years) improvement in patients with degenerative lumbar spinal stenosis and has been shown to improve outcomes in a large percentage of patients Surgical decompression may be considered in patients aged 75 or greater with lumbar spinal stenosis.”	“The long-term results of surgical management of spinal stenosis can vary depending on several factors, including the severity of the stenosis, the age and overall health of the patient, and the type of surgery performed In general, studies have shown that surgery for spinal stenosis can provide significant long-term relief of symptoms, including pain, numbness, and weakness in the legs. For example, a study published in the new England Journal of Medicine in 2008 found that patients who underwent surgery for spinal stenosis had greater improvement in their physical function and less pain than those who received nonsurgical treatment, even after 4 years of follow-up However, there are also risks associated with spinal surgery, including infection, nerve damage, and spinal fluid leaks. In some cases, patients may require additional surgeries or experience ongoing symptoms despite surgery Overall, the long-term results of surgical management of spinal stenosis are generally positive, but it is important for patients to carefully consider the potential benefits and risks of surgery and discuss their options with their healthcare provider.”

Table 2.

Insufficient Data to Make Conclusions Based on NASS Guidelines vs ChatGPT Recommendations. Questions and Recommendations From the NASS Guidelines Recorded for Responses That had Insufficient Evidence to Make a Claim. Responses From ChatGPT Were also Recorded and Graded for Precision.

Question	NASS	ChatGPT
What is the role of physical therapy/exercise in the treatment of spinal stenosis?	There is insufficient evidence to make a recommendation for or against the use of physical therapy or exercise as stand-alone treatments for degenerative lumbar spinal stenosis In the absence of reliable evidence, it is the work group’s opinion that a limited course of active physical therapy is an option for patients with lumbar spinal stenosis.”	“Physical therapy and exercise can play a crucial role in the treatment of spinal stenosis, which is a condition that occurs when the spinal canal narrows and puts pressure on the spinal cord or nerves. The goal of physical therapy and exercise is to improve strength, flexibility, and range of motion, reduce pain and discomfort, and enhance overall function Some specific ways that physical therapy and exercise can be helpful for spinal stenosis include:  Strengthening exercises: Strengthening exercises can help to build up the muscles that support the spine and improve posture. This can help to take pressure off the spinal canal and reduce pain Stretching exercises: Stretching exercises can help to improve flexibility and range of motion in the spine, which can also reduce pressure and alleviate pain  Aerobic exercise: Aerobic exercises, such as walking or cycling, can help to improve cardiovascular health and overall fitness, which can make it easier to perform daily activities without pain or discomfort  Manual therapy: Manual therapy, such as massage or mobilization, can help to release tension in the muscles and joints and improve overall spinal alignment, which can reduce pressure on the spinal canal Overall, physical therapy and exercise should be tailored to the individual needs and goals of each patient and should be performed under the guidance of a qualified healthcare professional, such as a physical therapist.”
What is the role of manipulation in the treatment of spinal stenosis?	There is insufficient evidence to make a recommendation for or against spinal manipulation for the treatment of lumbar spinal stenosis	“The role of manipulation in the treatment of spinal stenosis is controversial and depends on the specific case and the preferences of the patient and their healthcare provider Manipulation, also known as spinal manipulation or spinal adjustment, is a manual therapy technique that involves the application of controlled force to joints in the spine to improve function and reduce pain. It is often performed by chiropractors, osteopaths, and physical therapists Some studies suggest that manipulation may be effective in reducing pain and improving function in patients with spinal stenosis. However, the evidence is limited and conflicting, and there is a risk of injury or worsening of symptoms if manipulation is not performed correctly or inappropriately Therefore, the decision to use manipulation as a treatment for spinal stenosis should be based on a careful assessment of the patient's individual needs and preferences, as well as the potential risks and benefits of the technique. Other non-surgical treatments such as physical therapy, medications, and corticosteroid injections may also be considered as part of a comprehensive treatment plan for spinal stenosis. In some cases, surgery may be necessary to relieve symptoms and improve function.”
What is the role of pharmacological treatment in the management of spinal stenosis?	There is insufficient evidence to make a recommendation for or against the use of pharmacological treatment in the management of spinal stenosis	“Pharmacological treatment can be useful in managing the symptoms of spinal stenosis. Here are some common pharmacological treatments that can be used: Nonsteroidal anti-inflammatory drugs (NSAIDs): These drugs help to reduce pain and inflammation. Examples include ibuprofen, naproxen, and aspirin Acetaminophen: This medication is used for pain relief and is an alternative for those who cannot take NSAIDs Opioids: These medications are reserved for severe pain that is not relieved by other treatments. They can be effective, but they are associated with a risk of addiction and other side effects Corticosteroids: These medications can help to reduce inflammation and swelling in the affected area. They can be given orally or as an injection into the affected area Antidepressants and anticonvulsants: These medications can help to relieve nerve-related pain It is important to note that medication alone may not be sufficient for managing spinal stenosis, and other treatments such as physical therapy and surgery may be necessary depending on the severity of the condition. A healthcare professional should be consulted to determine the best course of treatment for an individual's specific situation.”
What is the role of ancillary treatments such as bracing, traction, electrical stimulation and transcutaneous electrical stimulation (TENS) in the treatment of lumbar spinal stenosis?	The use of a lumbosacral corset is suggested to increase walking distance and decrease pain in patients with lumbar spinal stenosis. There is no evidence that results are sustained once the brace is removed There is insufficient evidence to make a recommendation for or against traction, electrical stimulation, or TENS for the treatment of patients with lumbar spinal stenosis There is insufficient evidence to make a recommendation for or against acupuncture, in for the treatment of patients with lumbar spinal stenosis.”	“Lumbar spinal stenosis is a condition that occurs when the spinal canal in the lower back narrows and compresses the spinal cord and nerves. Ancillary treatments such as bracing, traction, electrical stimulation, and transcutaneous electrical stimulation (TENS) can play a role in the treatment of lumbar spinal stenosis Bracing: Bracing can provide support to the lower back and improve posture, which can reduce pain and improve function. A back brace can also help to reduce pressure on the spinal cord and nerves, which can alleviate symptoms. However, bracing is not a long-term solution and should be used in combination with other treatments Traction: Traction involves the use of weights or a mechanical device to pull the spine and create space between the vertebrae. This can help to reduce pressure on the spinal cord and nerves and alleviate symptoms. However, traction is not recommended for everyone and should only be performed under the guidance of a healthcare professional Electrical stimulation: Electrical stimulation can be used to reduce pain and improve muscle strength in the lower back. This type of treatment involves the use of a device that sends electrical impulses to the nerves in the affected area. Electrical stimulation can be performed in a clinical setting or at home using a portable device Transcutaneous electrical stimulation (TENS): TENS involves the use of a small device that sends electrical impulses to the nerves in the affected area through electrodes placed on the skin. TENS can help to reduce pain and improve muscle strength in the lower back. TENS can be performed at home and does not require a healthcare professional Overall, ancillary treatments such as bracing, traction, electrical stimulation, and TENS can be effective in the treatment of lumbar spinal stenosis. However, these treatments should be used in combination with other treatments such as physical therapy, medication, and lifestyle modifications. It is important to discuss the use of these treatments with a healthcare professional to determine the best course of treatment for individual cases.”
What are the most appropriate historical and physical findings consistent with the diagnosis of degenerative lumbar spinal stenosis?	“The diagnosis of lumbar spinal stenosis may be considered in older patients presenting with a history of gluteal or lower extremity symptoms exacerbated by walking or standing which improves or resolves with sitting or bending forward. Patients whose pain is not made worse with walking have a low likelihood of stenosis There is insufficient evidence to make a recommendation for or against the use of self-administered questionnaires to improve the accuracy of the diagnosis of spinal stenosis There is insufficient evidence to make a recommendation for or against certain physical findings for the diagnosis of degenerative lumbar spinal stenosis including an abnormal Romberg test, thigh pain exacerbated with extension, sensorimotor deficits, leg cramps, and abnormal Achilles tendon reflexes There is insufficient evidence to make a recommendation for or against the diagnostic reliability of patient-reported dominance of lower extremity pain and low back pain.”	“The most appropriate historical and physical findings consistent with the diagnosis of degenerative lumbar spinal stenosis include: Historical findings: Gradual onset of symptoms Symptoms worsen with standing or walking and improve with sitting or bending forward Pain and numbness in the lower back, buttocks, and legs, which may be unilateral or bilateral Weakness in the legs or difficulty with walking, particularly on inclines or stairs Loss of bladder or bowel control (in severe cases) Physical findings: Limited range of motion in the lumbar spine Tight hamstrings and calf muscles Positive straight leg raise test (pain or tingling in the leg with leg raised) Decreased or absent ankle reflexes Muscle weakness or atrophy in the legs Sensory deficits in the legs, such as decreased sensation to light touch or pinprick Positive Lhermitte's sign (electric shock-like sensation down the back and legs with neck flexion) It is important to note that not all patients with degenerative lumbar spinal stenosis will exhibit all of these findings, and other conditions may present with similar symptoms. A thorough medical history, physical examination, and diagnostic imaging (such as X-rays, MRI, or CT scan) are needed to confirm the diagnosis and develop an appropriate treatment plan.”

Assessment of ChatGPT’s Performance

Our study revealed that ChatGPT has the potential to provide accurate recommendations for interventions and treatments in the management of lumbar stenosis. Our findings can be categorized into two themes: (1) The precision of ChatGPT in providing recommendations compared with the NASS guidelines, and (2) The capability of ChatGPT to provide accurate recommendations in the absence of evidence presented within the NASS guidelines.

ChatGPT vs Guidelines Recommendations

Definition and History

The NASS guidelines recommended answers for two of the three questions that fell under the category of definitions and natural history of the condition. When comparing the responses to the question regarding the working definition of lumbar stenosis, ChatGPT was able to generate a definition that closely resembled what the guidelines presented, including the correct anatomical aberrations and associated symptoms. However, when asked about the natural history of lumbar spinal stenosis, ChatGPT failed to mention some of the criteria presented in the guidelines. For instance, ChatGPT failed to mention the favorability rate of the diagnosis, specifically that mild to moderate stenosis can be favorable in one-third to one-half of patients. These findings have been recorded across numerous studies in the literature throughout the past decade.^7-10 Furthermore, the guidelines mentioned the rarity of rapid neurological impairments following a lumbar stenosis diagnosis, a symptom that ChatGPT failed to comment on. There have only been a handful of case reports of neurological decline within patients with spinal stenosis, suggesting that it is indeed an atypical finding. ¹¹

Diagnostic Tests

Within the NASS guidelines, there was only one question that assessed the role of appropriate diagnostic tests for lumbar stenosis. We found that ChatGPT gave a comprehensive response that adequately matched many major themes highlighted by the guidelines. When discussing imaging tests that can be implemented, ChatGPT accurately presented the utilization of X-rays, MRIs, and CT scans in the diagnosis of spinal stenosis. Several studies in the current literature have discussed the diagnostic benefits of using each of the aforementioned tests.^12-15 ChatGPT also accurately presented information on the importance of nerve function tests that were highlighted by the guidelines, including electromyography and nerve conduction studies. These findings by both the guidelines and ChatGPT have been tested across multiple papers in the last few years and have been found to be an accurate diagnostic tool.^16-18 It should be noted that although ChatGPT did not rank the different tests in terms of their efficacy as the guidelines did, it communicated the importance of seeking a healthcare professional to help make a decision on the most appropriate diagnostic test that should be utilized.

Non-Surgical Interventions

The NASS guidelines had adequate evidence to provide recommendations for only three of the seven questions related to non-surgical interventions in the treatment of lumbar stenosis. Two of the questions assessed the role of epidural steroid injections (ESI), while the third question explored the long-term results of medical management of stenosis. When assessing the role of contrast-enhanced, fluoroscopic guidance for the injection of epidural steroids, both the guidelines and ChatGPT agreed that it is recommended in order to improve the accuracy of medication delivery. Although there are only a couple of studies in the literature that specifically assess the effectiveness of fluoroscopically guided epidural steroid injections in lumbar stenosis, they all reached the same consensus as highlighted by the guidelines.^19,20

The second question explored the role of ESI in the management of spinal stenosis, and our results showed that ChatGPT was able to accurately able to comment on the location and the benefits of the injection. There are a couple of randomized trials that have assessed the efficacy of ESI in spinal stenosis and found that a slight benefit was detected in patients in the short term.^21,22 Interestingly, both the guidelines and ChatGPT emphasized that most ESIs were only analyzed in short-term studies and there wasn’t enough evidence to comment on the long-term effects, suggesting ChatGPT is capable of adapting their recommendations to include short and long-term analyses of interventions. The results from the final question found that ChatGPT was able to accurately conclude that long-term results following the medical management of stenosis may improve patient outcomes, but that it is dependent on a case-to-case basis. These results are a further suggestion that ChatGPT is capable of understanding that broad questions like these cannot be answered in a general sense and that it is important to have more information about the individual patient in order to make a definitive conclusion.

Surgical Interventions

Within the NASS guidelines, all three questions pertaining to the surgical interventions involving spinal stenosis were provided with recommendations based on evidence-based literature. One of the most common procedures performed is spinal decompression, which the guidelines recommend as the intervention for patients with at least moderate symptoms of lumbar spinal stenosis. Within the literature, there is a multitude of systematic reviews and meta-analyses in the past five years that have associated spinal decompression with favorable outcomes for patients with spinal stenosis.^23-26 ChatGPT provided a very similar recommendation to the guidelines and importantly, took into account that this procedure should be assessed on a patient-to-patient basis. These results suggest that if given adequate information, ChatGPT may be capable of recommending even more specific interventions based on the condition of the patient. However, the precision of ChatGPT’s answer, in this case, was graded as imprecise due to its citation of false evidence from studies that don’t exist in the literature in its response.

The second question presented in the NASS guidelines explored whether the addition of fusion with spinal decompression improves patient outcomes. Both the guidelines and ChatGPT highlighted that decompression alone is the superior approach, especially for patients who don't present with instability. These findings have been recently corroborated in the literature by multiple studies.^25,27,28 Once again, ChatGPT presented a conservative recommendation to this question and emphasized that only patients that present with instability tend to get this operation. However, it once again presented within its response fictitious papers that could not be found in the literature.

Finally, the NASS guidelines investigated what the long-term results were for the surgical management of stenosis and concluded that there had been significant improvements in a large percentage of patients after >4 years postoperatively. They also recommend that surgical decompression be considered in patients who are ≥75 years or older. A handful of long-term studies that addressed this specific question were found to be in favor of the recommendations from the guidelines.^29-31 ChatGPT was also able to reach the same conclusions, but once again, had within its response fabricated evidence from false articles. In all, it is important to note that ChatGPT can be used to assess the benefits and risks associated with different surgical interventions in lumbar spinal stenosis, but may fall short in providing accurate evidence-based literature within its responses.

ChatGPT recommendations vs insufficient data in guidelines

Due to insufficient evidence at the time of writing, the NASS guidelines did not provide recommendations for five questions, including physical exercise and therapy, manipulation, pharmacological treatment, ancillary treatments, and physical findings that are associated with lumbar spinal stenosis. However, ChatGPT provided recommendations that were cross-referenced with the contemporary literature.

Physical Exercise and Therapy

When assessing the role of physical therapy and exercise in the treatment of spinal stenosis, the NASS guidelines didn’t make a recommendation about its utilization and broadly recommended that any active therapy can be considered an option. On the other hand, ChatGPT provided specific exercises that could be implemented for patients, including strengthing, stretching, and aerobic exercises. The conclusions reached by ChatGPT have been studied in the recent literature. A randomized controlled trial (RCT) of 104 patients with spinal stenosis found 82% of participants who were part of a 6-week structured comprehensive training program that consisted of 18 exercises and stretches targeted for the lower back reached a minimally clinically important difference (MCID) in mean walking capacity compared to 63% in the self-directed program. ³² Another RCT that analyzed 86 stenosis patients found supervised physical therapy that included stretching and strengthening exercises, cycling, and treadmill walking to be significantly associated (P = .01) with higher MCIDs compared to a home-exercise group. ³³ The findings from these studies match the recommendations generated by ChatGPT, suggesting that it is capable of providing accurate recommendations to clinicians when it comes to exercise and therapy for lumbar stenosis.

Manipulation

With regards to manipulation techniques, which involve manual therapy usually performed by chiropractors or physical therapists, the NASS guidelines abstained from recommending both in favor or against this approach to the management of spinal stenosis due to a lack of evidence. However, ChatGPT did make certain recommendations regarding the impact that manipulation may have on spinal stenosis, taking into consideration the controversies behind some of the practices. There have been multiple studies that analyzed the effects of manipulation techniques in stenosis patients with varying results. For instance, Choi et al ³⁴ found that flexion-distraction spinal manipulation was effective at reducing pain in patients with spinal stenosis to a greater degree compared to those treated with conservative physical therapy. This study aligns with the statement given by ChatGPT with regard to pain outcomes when using manipulation in spinal stenosis patients. Another study by Oh et al ³⁵ reported that flexion-distraction reduces pain to a greater extent than conservative physical therapy, indicating that manual manipulation therapy is an effective intervention for pain management. Furthermore, Smith et al ³⁶ concluded that flexion-distraction manipulation improves both pain and function in lumbar spinal stenosis (LSS) patients, recommending its utilization in the clinical setting. The results from these studies are in agreement with ChatGPT’s recommendations of manipulation with regard to function since it was found that manipulation therapy improved function in patients burdened with spinal stenosis.

Other studies have assessed the quality of the existing literature with regard to the role of manipulation in LSS and concluded that there is “moderate-quality evidence” supporting the positive role that manipulation has in managing LSS. ³⁷ This systematic review found a single-randomized clinical trial testing the effects of manipulation on LSS and the referenced study found no significant improvements in motion or function following manipulation in patients with LSS. ³⁸ In accordance with the output from ChatGPT, the literature does have conflicting evidence regarding the benefits of manipulation. Thus, ChatGPT gave an accurate representation of the current body of literature regarding manipulation and outlined documented benefits in terms of pain and function, while still being conservative and acknowledging the conflicting body of evidence.

Pharmacological Treatment

The NASS guidelines have made no recommendations for or against pharmacological treatment for LSS. However, ChatGPT did generate an output regarding pharmacological treatment and denoted it as “useful” in the treatment of spinal stenosis. Specifically, ChatGPT gave a broad pain control recommendation including both NSAIDs and acetaminophen, both of which have been studied in the literature. Chou et al ³⁹ conducted a systematic review on the effectiveness of pharmacological treatments for the treatment of lower back pain and found that acetaminophen was ineffective for chronic lower back pain (CLBP), a condition with symptomatology similar to LSS. Another systematic review by Enthoven et al ⁴⁰ identified weak evidence that favors NSAID use for the long-term management of CLBP, concluding that no firm recommendation can be made about their implementation in the clinic. Despite the weak evidence in the literature regarding NSAID or acetaminophen use in the treatment of CLBP or LSS, ChatGPT still deemed these pharmacological treatments as useful in the management of LSS symptomatology.

Additionally, ChatGPT’s recommendation failed to incorporate new evidence with regard to pharmacological treatments that have historically been used for neuropathic pain. One medication of this nature is gabapentin which is typically used in diabetes-associated neuropathic pain, but some studies have explored its effects in the management of LSS. A study by Haddadi et al. reported a decreased number of patients with paraesthesia and claudication 8 weeks after starting Gabapentin. Furthermore, a systematic review by Ammendolia et al ⁴¹ found that gabapentin was effective in reducing pain and improving walking distance. Other pharmacological treatments that have been documented in the literature include epidural corticosteroids which have been tested for their effectiveness in LSS management. A recent randomized clinical trial by Friedly et al ²² found no benefit of epidural corticosteroids in terms of disability and pain over a period of 12 months in patients with LSS. A more rigorous randomized trial by Friedly et al ²¹ in 2014 similarly concluded that over 12 months epidural corticosteroid injections show no significant benefit in terms of disability and pain associated with LSS. Other medications like duloxetine have been tested in patients with chronic lower back pain (CLBP) which has symptomatology similar to that of LSS. Duloxetine is thought to work by increasing neurotransmitters involved in the descending pain inhibitory pathway promoting analgesia via that mechanism. Another study by Konno et al ⁴² reported improvements in CLBP at week 14 compared to placebo (P = .0026) as well as improvements in impression of severity (P = .0019), and disability (P = .0439). ChatGPT failed to present information available on gabapentin, epidural corticosteroids, and duloxetine which all have supporting evidence for their potentially positive role in the management of LSS. In terms of pharmacological recommendations for LSS, ChatGPT inappropriately recommended NSAIDs and acetaminophen for LSS and did not include other pain medications for neuropathic pain widely explored in the body of literature for their potential benefits.

Ancillary Treatments

With regard to ancillary treatments, the NASS guidelines did not make a definitive recommendation about treatments such as bracing, electrical stimulation, or acupuncture. ChatGPT commented on the role bracing can play in reducing pressure on the thecal sac and nerve roots, while also mediating support to the lower back. There is very limited literature that specifically assessed the roles of bracing in the context of stenosis. One study by Ammendolia et al ⁴³ developed a prototypical lumbar stenosis belt and conducted a two-arm, double-blinded RCT that assessed walking distance in 104 patients who either wore the prototype vs patients who wore standard lumbar support (Tensor 3M Canada). The authors concluded that both the prototype and standard lumbar support significantly improved the walking ability of patients with lumbar stenosis, but there was no significant difference found between the two groups. Based on the improved outcomes of patients in this study, ChatGPT’s response was deemed accurate when it comes to bracing recommendations.

ChatGPT further commented on the role of electrical stimulation and specifically Transcutaneous Electrical Stimulation (TENS), recommending their utilization in either the hospital or home setting for pain reduction and muscle strengthening, although there is also very limited literature on this data. A recent RCT conducted by Harmsen et al ⁴⁴ assessed the role of neuromuscular electric stimulation in patients presenting with leg cramps secondary to their diagnosis of lumbar spinal stenosis. Thirty-two participants were grouped into four cohorts that received either 85%, 55%, 25%, or 0% of their maximum tolerated stimulation intensity and found that the first three groups reported a significantly reduced amount of leg cramps. These findings suggest that even at low stimulation levels, the application of electrical stimulation can alleviate pain in patients with lumbar stenosis. Another RCT by Ammendolia et al ⁴⁵ explored how TENS can impact neurogenic claudication and walking ability by randomizing 104 patients either to an active (n = 49) or detuned TENS (n = 51) group. The results revealed that 71% of the active TENS and 74% of the detuned TENS participants achieved a 30% improvement in walking distance, priming the authors to recommend against the utilization of TENS in clinical practice. Since ChatGPT did not comment in its statement about this study, and because there is a lack of evidence to make these recommendations when it comes to ancillary treatments, this response was marked as imprecise.

Furthermore, the NASS guidelines had commented on the role of acupuncture but ChatGPT did not mention this intervention. However, based on the most recent literature search, it is plausible to assume that most of ChatGPT’s recommendations are backed by limited evidence-based studies and can serve as a useful tool when it comes to assessing ancillary treatments.

Physical Findings Associated With LSS

The NASS guidelines did not provide a recommendation for one of the three questions that fell under the category of definitions and natural history of the condition. In this case, the guidelines did not have sufficient evidence to make any recommendations related to the history and physical findings associated with a diagnosis of lumbar spinal stenosis other than older patients who present with a history of lower extremity pains that are exacerbated by walking or standing which can be characterized as having spinal stenosis. ChatGPT failed to categorize the patients by different demographic factors such as age in its response when assessing the physical findings, making its conclusion less accurate.

The NASS guidelines also commented on the utilization of self-administered questionnaires in the diagnosis of spinal stenosis. However, ChatGPT did not mention qualitative measures such as questionnaires within its response, further decreasing its preciseness compared to the guidelines. There have been more recent studies that have been published on the benefits of such qualitative measures. For instance, a study by Tominaga et al ⁴⁶ assessed the effectiveness of two measures, the LSS-diagnosis support tool (LSS-DST) and the LSS-self-administered self-reported history questionnaire (LSS-SSHQ), and found that both tools were more sensitive in screening patients for spinal stenosis than traditional measures presented by the NASS guidelines. Another recent study by Wilartatsmai evaluated the reliability of the Swiss Spinal Stenosis (SSS) questionnaire in one hundred and seven patients and found it to be a strongly valid and reliable tool in the evaluation of physical functions and symptom severities in patients with spinal stenosis. ⁴⁷ Taken together, the conclusions from these studies suggest the implementation of qualitative measures in the diagnosis of stenosis. The remainder of the response reported by ChatGPT provided general physical findings that have been found to be associated with spinal stenosis while noting that it is best to get a thorough history, physical examination, and diagnostic imaging to make an accurate decision.

Future Directions

ChatGPT is a non-generative AI model that is mostly trained to recognize patterns and generate predictions when it is presented with a prompt. However, with the rise of new generative AI models that are capable of assessing and synthesizing new content that is fine-tuned from medical databases, it is possible to generate stronger results than what our study has shown. Future research is needed to assess these new generative models to provide clinicians with the most recent and productive tools that can help make decisions in the field.

Limitations

A limitation of this study is that the NASS Evidence-Based Clinical Guidelines for Lumbar Spinal stenosis have not been updated since 2011, while ChatGPT has derived data from up to 2021. Thus, it is plausible that some of the recommendations presented by the guidelines are outdated. However, we conducted a contemporary literature search about the most recent evidence-based studies as a means of assessing ChatGPT’s reliability to reach conclusions based on new data. Another limitation is that some questions presented by the guidelines were too broad, priming ChatGPT to generate responses that were generally non-specific. By priming the questions with more specific details, ChatGPT may better serve clinicians with more accurate responses, although further work is needed to prove this. Furthermore, a third limitation is that ChatGPT have been trained to generate definitive responses to direct questions, even if it is sometimes better to provide a response that claims that there is not enough evidence to make a recommendation. This phenomenon could be costly to patients and has been labeled as “artificial hallucination”, which highlights the issue with ChatGPT trying to cite references that do not exist in support of its argument. Additionally, ChatGPT may have been trained on NASS guidelines when first developing its algorithm, possibly introducing a response bias to the specific questions it was presented with in our study. Finally, clinically fine-tuned models such as Med-PaLM 2 are not publicly available, which is why ChatGPT was utilized in this study.

Conclusions

The implementation of AI models in the medical setting is becoming a promising tool that can help physicians make better decisions when it comes to patient care. Our study suggests that ChatGPT may accurately answer questions and provide evidence about the diagnosis and treatment options for lumbar spinal stenosis that is line with the current available evidence. However, it has shown to fabricate evidence from time to time when it comes to specific questions pertaining to spinal stenosis. Thus, we recommend patients and clinicians utilize ChatGPT with caution. Further study is required to understand the full impact ChatGPT can provide in the clinical setting for spine surgery and beyond.