Objective Evaluation of Chronic Low-Back Pain Using Serum Lipids: The Role of the Doctor-Patient Relationship

Statistical data show that pain intensity in patients with low back pain is associated with a higher BMI, total serum cholesterol, and triacylglycerol levels. The objective of our study was to evaluate how these associations are dependent on the nature of the patient-doctor relationship. Eighty-nine patients hospitalized with chronic low-back pain (50 women, 39 men; average age: 64.5 ± 12.7 years) were assessed over a 3-year period. A serum lipid analysis was conducted (LDL-C, HDL-C, and total cholesterols) at admission in parallel with a subjective evaluation of pain intensity, which was assessed using a numeric rating scale. The participating physician assigned, based on their personal interaction with the patient, an attribute of affinity (positive, neutral, and negative) towards them. Current serum lipid levels and pain intensity were correlated relative to these attributes. Pain intensity did not differ between the groups assigned positive or negative attributes of affinity. In patients belonging to the “positive” group, pain intensity correlated positively with total cholesterol (p=0.01) and LDL cholesterol (p=0.007). No correlations were found in the “negative” group or when the patient-doctor relationship was ignored. We found a significant association between subjectively assessed low back pain intensity and serum levels of total and LDL cholesterol in patients with whom the physician had a positive affinity. A positive affinity with the patients having chronic pain and the patient's trust in their physicians may ultimately mean that the patient's statement about pain is more credible, which may retroactively affect the outcome of therapy.


Introduction
Chronic pain is an indirectly observable phenomenon. Tere are no quick and accurate objective methods for evaluating its intensity. Objective electrophysiological methods, neuroimaging, and mapping the activity of the autonomic nervous system have all been used to assess acute as well as chronic pain, but they are not practical for use in everyday practice [1][2][3][4][5]. In common clinical practice, questionnaires are often used to assess chronic pain, which is reproducible and easy to administer [6,7]. Te McGill University Questionnaire is a commonly used self-assessment questionnaire, where patients provide their physicians with subjective information about the intensity and quality of the pain they experience [7,8]. Te Brief Pain Inventory (BPI) and the clinically popular short-form BPI-SF are other common methods which assess, in addition to the pain intensity, the location of pain indicated (on a drawing or image of a human fgure), afective descriptors of pain, and medications used for pain treatment [9]. Te simplest and the most commonly used assessment is the numeric rating scale  or the visual analog scale (VAS), where pain intensity is visually marked on a line with two extremes, no pain and maximum possible pain [10]. Both methods involve some simplifcation of the actually experienced pain, but the simplicity makes them easy to understand.
Te relationship between obesity and chronic vertebrogenic pain has been shown, and analyses on representative sets of patients also indicated that patients with high BMIs also had elevated cholesterol and triacylglycerol levels [16,17].
Because blood chemistry is a routine, reproducible, and particularly easy evaluation, the goal of our work was to study the association between subjectively experienced pain and serum lipid levels in patients with chronic vertebrogenic pain but without signifcant obesity.
A patient's assessment of their pain, in the presence of an attending physician, can be signifcantly infuenced by their relationship with the physician. We also have the patient's motivation to be treated, which can simulate or dissimulate pain, which can be afected by their relationship with the doctor. Empathy is the ability to share the afective states of others, or the ability to understand the pain of others, and is important in the formation of a positive doctor-patient relationship [18]. It is also associated with patient satisfaction and compliance with recommended treatments [19]. In good doctor-patient relationships, pain self-assessments by patients are considered more accurate and reliable by the physician and thus contribute to the choosing of the best therapy; it may also better refect objectively measured clinical values.
We tested the hypothesis that patients with more intensive vertebrogenic pain would have higher plasma lipid levels, especially total cholesterol, and LDL cholesterol. Additionally, we looked at how this association could be infuenced by the patient-doctor relationship, also called therapeutic alliance [20]; an association that has not been previously studied.

Clinical Subjects.
Tis was a multiyear observational study. Over three years, 89 patients (39 men, 50 women) with an M51-4 or G54.2-4 diagnosis, based on the 10th ICDrevision and chronic vertebrogenic pain were evaluated at the inpatient unit of the Department of Neurology. Te vertebrogenic symptom complex includes local and referred pain and autonomic refex dysfunction within the lumbosacral zones of the head [21]. Te pain can have many qualities and a number of causes as well as just as many treatment options; in addition to pharmacological treatment, it may also include physical therapy or spinal manipulation and, in justifed cases, surgical treatment.
Study exclusion criteria included the following: dependence on tobacco, alcohol, or narcotics, hormonal therapy treatment, medical treatment for an endocrine disease, and psychiatric diagnosis. All patients were taking nonsteroidal anti-infammatory drugs, 20 patients were taking antidepressants, and 32 patients were taking statins. All patients signed informed consent, and the study was approved by the local Ethics Committee.

Assessment.
On the day of admission, blood samples were drawn from fasting patients. Serum lipids, i.e., total cholesterol (TC), LDL-cholesterol, HDL-cholesterol, and triacylglycerols (TG), were measured using an automatic biochemical analyzer in a certifed laboratory. In the presence of a physician, the patients assessed their pain intensity using a numeric rating scale (NRS-11) by assigning numbers 0-10 to their pain level, where 0 was a pain-free state, and 10 was the maximum imaginable pain intensity.
At this same time, the attending physician assigned an attribute of afnity (or fellowship) based on their experience with the patient, where "+" means a high degree of afnity with the patient, "0" a neutral afnity, and "−" a negative afnity with the patient. Te attitude towards the patient was based primarily on the subjective experience of mutual verbal and nonverbal communication.

Statistical Analyses.
All quantitative variables are given as means and standard deviations, and qualitative variables were given as frequencies and percentages. Tese parameters were calculated both for all patients together and for two groups of patients, which were created according to the patient-doctor relationship. In the frst group were the patients with positive afnity attribute and in the second group were the patients with negative or neutral attribute.
To fnd the relationship between pain intensity NRS-11 and serum lipids, i.e., TC, LDL-C, HDL-C, TG, and LDL/ HDL ratio, as well as between NRS-11 and age and BMI of the patients, the parametric Pearson's correlation coefcients were calculated. Te required sample size for all signifcant correlations was calculated.
After merging the indiferent and negative groups of patients, we obtained two groups of patients with positive and negative rapport. For the comparison of age, BMI, NRS-11, and all observed serum lipids between these two groups of patients, the parametric Student's t-test was used. To compare the qualitative variables, i.e., sex (M/F) and medication with statins (Yes/No) and antidepressants (Yes/ No), the analysis of contingency tables was used; Pearson's chi-square test was calculated.
In order to evaluate the efect of all observed variables in the assessment of the severity of chronic pain (NRS-11), the general multivariable regression analysis was performed. We created three regression models-for all patients together and for each group of patients-with positive and negative rapport separately.
In order to fnd whether our observed variables discriminate well enough between the two groups of patients (positive vs. negative rapport), the multivariate logistic regression model was performed. Te list for predictive factors was the same as in the multivariate regression model except for the addition of NRS-11.
TIBCO Statistica version 14.0 and IBM SPSS Statistics version 23 were used for statistical analysis. A p value less than 0.05 was considered to be statistically signifcant.

Gender Diferences.
Te mean pain intensity was the same in men and women (5.62 vs. 5.66, respectively; p � 0.92). Men had higher TG compared to women (1.39 vs.

2
Pain Research and Management 1.13 mmol/l, respectively; p � 0.04), but they did not difer in levels of TC, LDL-C, and HDL-C.

Efect of Statins and Antidepressants.
Tirty-two patients (21 men and 11 women) were taking statins for lipidlowering therapy. Te statin group had lower TC and LDL-C in comparison with untreated patients (TC: 4.86 vs. 5.42 mmol/l, respectively, p � 0.025; LDL-C: 2.7 vs. 3.34 mmol/l, respectively, p � 0.003). Te groups did not difer relative to pain intensity. Twenty patients (6 men and 14 women) were taking antidepressants. Neither pain intensity nor lipid levels differed from patients not taking antidepressants. Te lower weight and BMI in this group refects the greater number of women in the group. (Table 1). Forty-three patients (18 men and 25 women) received a positive afnity attribute, thirty-six patients (18 men and 18 women) received a neutral attribute, and ten patients (3 men and 7 women) received a negative attribute. Due to the small number of subjects in the last category, these patients were merged with the neutral group, creating a group that contained 21 men and 25 women.

Patient-Doctor Relationships
Although positive and neutral or negative patients did not difer signifcantly in pain intensity, BMI, TC, or HDL-C, they did difer signifcantly in the LDL-C which was lower in positive afnity patients (2.86 vs. 3.34 mmol/l, respectively, p � 0.028) and the LDL/HDL ratio which was also signifcantly lower in the positive afnity patients (1.82 vs. 2.27, respectively, p � 0.004). According to the results of contingency tables analysis (Table 1 lower part), we found no statistically signifcant diference in distributions of men and women, usage of statins (Yes/No), and antidepressants (Yes/ No) between positive and negative groups of patients. (Table 2). Analyzing the relationships between pain intensity and lipid levels in the entire sample of patients, it was shown that pain did not correlate with any particular variable. In contrast, we found a signifcant positive correlation between TC and pain intensity (r � 0.393, p � 0.010) (Figure 1), LDL-C and pain intensity (r � 0.409, p � 0.007) (Figure 2), and LDL/HDL ration and pain intensity (r � 0.348, p � 0.024) in the group of positive afnity patients only. Required sample size conditions were almost fulflled only for TC (required sample size 48) and LDL-C (required sample size 45). Neither HDL nor TG nor age and nor BMI correlated with pain intensity in any group.

General Regression Model.
In order to evaluate the efect of serum lipids, age, BMI, sex, and medication with statins and antidepressants on the assessment of the severity of chronic pain, general multivariable regression analysis was performed. As the independent variable, we used all serum lipids, i.e., TC, LDL-C, HDL-C, TG, and LDL/HDL ratio, and then age, BMI, sex, and medication with statins and antidepressants; as the dependent variable, the pain intensity was used. We created three regression models-for all patients together and for each group of patients-with positive and negative rapport separately. All three models were statistically insignifcant, and neither model was signifcant for predicting the severity of chronic pain, NRS-11.
Te values of coefcients of determination R 2 are as follows: positive rapport group: R 2 � 0.249, negative rapport group: R 2 � 0.169, and the entire sample of patients, as the input factor, the rapport between the patient and the doctor (positive vs. negative) was added. Te coefcient of determination R 2 � 0.103, the only variable for which the p value is less than 0.1, was the LDL/HDL ratio (p � 0.087).

Logistic Regression Model.
In order to fnd whether our observed variables discriminate well enough between the two groups of patients (positive vs. negative rapport), the multivariate logistic regression model was performed. Te list for predictive factors was the same as in the multivariate regression model except for the addition of NRS-11.
Te results are as follows: 2LL statistics is insignifcant if p � 0.352, Nagelkerke R 2 is only 0.146, and for the Hosmer-Lemeshow test, the p value is 0.490; the model adequately interpolated the data. Te classifcation ability is only 59.3%. Te area under the ROC curve (AUC) which determines the discrimination power of the logistic model reached the value 0.695; discrimination quality according to Tape [22] is "Poor." No input variable is statistically signifcant.

Pain and Serum
Lipids. In our study, we found a positive correlation between the subjectively evaluated intensity of chronic vertebrogenic pain and serum total and LDL cholesterol levels in adult patients with whom the attending physician had a relationship describable as positive afnity.
Experiments with animals have shown that serum lipids might refect a nonspecifc stress efect of acute and chronic pain [11]. After acute painful stimulation, HDL-cholesterol, triacylglycerols, glucose, and free fatty acids were elevated, whereas total cholesterol levels did not change and longterm repeated painful stimulation resulted in an increase in LDL-cholesterol and HDL-cholesterol. Te question is, to what extent this relationship also applies to humans.
Higher serum levels of triglycerides and HDL-cholesterol were detected in the acute pain of patients with fractures and acute pancreatitis [12]. Tese values subsequently decreased during hospitalization and treatment. Even in hospitalized pain-free controls, higher levels of LDL-cholesterol and triacylglycerols have been reported, which is believed to be the efect of immobilization stress during hospitalization.
In several studies, atherogenic lipids, LDL cholesterol, and triglycerides were associated with clinical manifestations of lumbalgia (low back pain) to support the lumbar atherosclerosis hypothesis as a cause of intervertebral disc degeneration or chronic hip pain [23,24]. On the other hand, the recent studies have questioned this hypothesis and consider it insufciently substantiated [25].

Pain Research and Management
Conversely, there is a work showing a decrease in total cholesterol in acute severe trauma. Patients with good outcomes in intensive care units showed improvement in TC levels during treatment, while a further reduction was observed in patients with infections, organ dysfunction, or death [26].
It is also important to note that each disease has its own dynamics. During illnesses, lipid levels may change and diferent active substances, not only lipids, may diferently afect pathological pain processes during the transition from acute to chronic pain [15].

Gender
Diferences. Compared to women, men have higher triglycerides and total cholesterol levels and lower HDL-C-levels [27]. Also, total-and LDL-cholesterol appears to be more important in determining cardiovascular diseases in men, while high triacylglycerols and low HDL-cholesterol are more signifcant in women [28].
In men, the total cholesterol and triacylglycerol levels correlated with the fve-year probability of developing lumbar vertebrogenic algic syndrome. Tis relationship has not been demonstrated in women [29]. Another study describes an inverse association between the prevalence of low back pain with HDL-C and a positive association with triacylglycerols, with stronger associations in women than in men [24,30]. A combination of higher total-and LDL-cholesterol levels has also been reported in patients with myofascial pain [31].

Statins and Antidepressants. Te role of statins in infuencing pain
is not yet fully clear. Attenuation of thermal hyperalgesia in an animal model of neuropathic pain, induced by partial ligation of the sciatic nerve, was independent on the statin-induced hypolipidemic efect [32].
In our study, patients taking statins had lower TC and LDL-C in comparison with patients without this treatment.

Pain Research and Management
Contrary to the expectation, triglycerides in these patients were increased. Tis observation can be explained by statins being efective at decreasing triglyceride levels but only in hyper-triglyceridemic patients [33].
Even though statins also exert a pleiotropic nonlipid efect, they possess anti-infammatory properties and antioxidant and neuromodulatory efects; we did not observe any efect on pain intensity in patients treated with statins [34]. Te controversial results following from animal and clinical studies did not permit simple conclusions about whether statins have a pain-inducing or pain-attenuating role [35].
Several studies have demonstrated frequent cooccurrences of pain and depression. Chronic pain is associated with changes in brain physiology and anatomy, and the positive impact of antidepressants might result in a reduction of these pathological processes and in the amelioration of symptoms, which can improve the quality of the life of patients [13,36]. Terefore, the antidepressant use was another analyzed factor. Although we did not evaluate the presence of clinical depression, almost one-fourth of our patients were taking antidepressants as adjuvant therapy. We found no diferences either in pain or in other serum lipid biomarkers between these two groups of patients. Te lower weight and BMI of the patients taking antidepressants can be explained by the greater occurrence of women in this group.

Afnity.
Te most important result of our study was fnding a positive correlation between the subjectively evaluated intensity of chronic vertebrogenic pain and serum total and LDL cholesterol levels in adult patients with whom the attending physician had a relationship describable as a positive afnity.
Here is the distinction between afnity, which can be understood as an afrmative connection with the feelings of another person and empathy, which represents an act of understanding without the need to agree with the feelings of another [37]. Empathy as a psychological phenomenon can be investigated using questionnaires and thus ofers the potential to objectify certain physiological relationships in the future [38]. Te attending physician in our study (T.B.) was a neurologist, and according to a comparative study conducted on a group of healthcare professions, he is in the category of specialists that have higher levels of empathy, as evaluated using the Jeferson scale of empathy [37].
Te positive correlation between pain and serum lipids is certainly infuenced by the patient's own assessment of pain. Tis means that the self-assessment of pain intensity may be somewhat afected by a therapeutic relationship in which it can function as a placebo or nocebo efect. According to Fabrizio Benedetti, meeting the doctor involves many psychological responses in the patient's brain, which are responsible things such as expectations, trust, and hope. Similarly, many mechanisms are at work in a doctor's brain, such as empathy and compassion. In turn, these led to the fnal step of providing therapy, which regardless of its efectiveness or inefectiveness, triggers a placebo response" [39].

Limitation of the Study.
Many studies have shown that physical activity is a safe method for improving patients' physical performance and alleviating symptoms. Both highintensity aerobic exercise and long-term low-intensity exercise were found to reduce pain, the disability rate, and psychological stress and enhance quality of life in patients with low back pain [40,41]. In our study, the physical activity of the monitored patients was not controlled.
Te impact of exercise on the lipid profle is somewhat controversial; a number of data confrm the benefcial efects of the regular activity on cholesterol levels; however, signifcant inconsistency in the blood lipid response has been observed. Te most frequently observed change was an increase in HDL-C with less frequently observed decrease in TC, LDL-C, and TG [42,43].
We consider the quality of the therapeutic relationship to be crucial for an objective verbal assessment of pain. Although the number of patients in our work is limited and the verifcation of the relationship presented in this study deserve more extensive research in relation to gender, age, and cultural habits, we consider a good therapeutic doctorpatient relationship a good starting point for more objective pain assessment.
We would like to believe the research results will be similar even in selecting and assessing other patients by another therapist. Rapport is not a personality trait, and individuals experience rapport as the result of a combination of qualities that emerge from each individual during interaction [21]. Tis hypothesis should be the aim of further research.
In conclusion, we found a positive correlation between subjectively evaluated chronic vertebrogenic pain intensity and serum total and LDL cholesterol levels in adult patients treated by a physician who described their relationship as having the attribute of positive afnity. Tis positive relationship between a physician and a patient with chronic pain, as well as the patient's confdence in the physician, may result in a more credible pain assessment by the patient, which may retroactively afect the outcome of the therapy.

Data Availability
Te data set used to support the fndings of this study is available from the corresponding author (yamamoto@ lf3.cuni.cz) upon request.

Disclosure
Te preliminary data of this paper were presented at FENS Congress in Berlin, 428 Germany, 2018. 6 Pain Research and Management

Conflicts of Interest
Te authors declare that there are no conficts of interest.

Authors' Contributions
Tomas Bruthans and Anna Yamamotová made substantial contributions to the conception of the paper and are responsible for the formulation of the hypothesis and interpretation of data. Tomas Bruthans was responsible for patient examination and selection of patients according to the inclusion criteria. Jana Vránová performed all statistical analyses and contributed to data interpretation.