Association among presence of cancer pain, inadequate pain control, and psychotropic drug use

Introduction Pain is a common symptom in cancer patients, and its control and management are complex. Despite the high concomitant use of psychotropic drugs among such patients, the association among pain, inadequate pain control, and psychotropic drug use has not been fully determined. This study examined the prevalence of cancer pain and inadequate pain control and the association with psychotropic drug use. Materials and methods In this cross-sectional study, we investigated 402 medical records obtained by simple random sampling of oncology patients at a hospital in northern Spain from July 2012 to July 2014. Adjusted odds ratios (ORs) were estimated together with their 95% confidence intervals (95% CIs) by unconditional logistic regression for each type of psychotropic drug (anxiolytics, hypnotics, and antidepressants). Results The mean patient age was 61.17 (standard deviation ± 13.14) years; 57.5% were women, 42.5% men. Pain was present in 18.4% of patients and inadequate pain control in 54.2%. We found a statistically significant association between the presence of cancer pain and anxiolytic use (adjusted OR, 3.15; 95% CI, 1.49–6.68) and hypnotic use (adjusted OR, 5.19; 95% CI, 1.77–15.25). Inadequate pain control was associated to a greater extent with the use of those drugs: adjusted OR for anxiolytic use, 4.74 (95% CI, 1.91–11.80); adjusted OR for hypnotic use, 6.09 (95% CI, 1.74–21.32). By contrast, no association was found between pain and antidepressant use (adjusted OR, 0.99). Conclusion The presence of pain and (to a greater extent) poor pain control were associated with increased use of certain psychotropic drugs, such as anxiolytics and hypnotics. There appeared to be no association between pain and antidepressant use.


Results
The mean patient age was 61.17 (standard deviation ± 13.14) years; 57.5% were women, 42.5% men. Pain was present in 18.4% of patients and inadequate pain control in 54.2%. We found a statistically significant association between the presence of cancer pain and anxiolytic use (adjusted OR, 3.15; 95% CI, 1.49-6.68) and hypnotic use (adjusted OR, 5.19; 95% CI, 1.77-15.25). Inadequate pain control was associated to a greater extent with the use of those drugs: adjusted OR for anxiolytic use, 4.74 (95% CI, 1.91-11.80); adjusted OR for hypnotic use, 6.09 (95% CI, 1.74-21.32). By contrast, no association was found between pain and antidepressant use (adjusted OR, 0.99).

Conclusion
The presence of pain and (to a greater extent) poor pain control were associated with increased use of certain psychotropic drugs, such as anxiolytics and hypnotics. There appeared to be no association between pain and antidepressant use. PLOS  Introduction

Materials and methods
The study protocol was approved by the Cantabria Clinical Research Ethics Committee on 29 May 2014; that was before data acquisition, which began in February 2015. The relevant authorization was also obtained in line with hospital regulations. The data were treated anonymously and confidentially under Spanish Organic Law 15/1999 of 13 December on Personal Data Protection. Because the data were analysed anonymously, it was not necessary to obtain the consent of patients to investigate their data. This study received the 16th National Nursing Research Award Valdecilla, which funds social work in Caja Cantabria (Spain). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Study design and participants
In this cross-sectional study, the eligibility criteria were patients aged over 18 years with oncological disease treated at the Oncology Unit of Hospital University Marqués of Valdecilla (National Public Health System of Spain) in northern Spain between 1 July 2012 and 1 July 2014. In all, 420 Oncology Unit patients were selected by simple random sampling. We considered the medical records incomplete when information about all the variables related to pain and the use of each type of psychotropic drug was missing. We considered the records complete if they contained information about >80% of the remaining variables. If the medical records were incomplete, we withdrew the patient from the study. Of the 420 original subjects, we removed 18 patients (4.3%) in this way (Fig 1).
To determine the sample size, we used the following criteria. With respect to prevalence, the literature is scarce, and so we calculated the sample size by assuming maximum indeterminacy (P = 0.5). Under that assumption, with a 95% confidence interval for a maximum error of 5%, we estimated a sample of 400 patients. That sample size (n = 400) would have sufficient power (1β >80%) to detect associations !1.3 as significant for a 50% risk of the effect in the unexposed sample, considering an unexposed/exposed ratio of 1, using a two-tailed chi-square test with an alpha level of 0.05.

Data sources and variables
We collected data between February and May 2015. From patient medical records and nursing assessments, we obtained information for each cancer patient included who underwent active treatment from 1 July 2012 to 1 July 2014. We obtained full oncological details. We collected the following variables: • Sociodemographic: age, sex, and education level.
• Family background: marital status and number of children [25].
• Oncological process: type of cancer (location), cancer treatment (chemotherapy, radiotherapy, hormone therapy, biological therapy, surgery), and side effects of cancer treatment. The symptoms considered side effects were those established by the National Cancer Institute [26]. S1 Appendix.
• Pain: presence of pain, location of pain, intensity of pain, and pain control. The presence of pain was treated dichotomously (yes/no). In patients whose pain was treated with analgesics, the type of analgesic, route of administration, and intensity of pain were recorded. The latter was measured following the recommendations of the European Society for Medical Oncology using the Visual Analogue Scale (VAS) [13], from 0 ('no pain') to 10 ('extreme pain'). Self-perceived pain control, reported by the patients themselves, was recorded in the nursing history; this information was obtained directly from the medical record. The variable 'pain control' was categorized ordinally into 'no pain', 'adequate pain control', and 'inadequate pain control'. The Oncology Unit uses the latest recommendations of the European Society for Medical Oncology to determine the prescription of analgesic treatment. The use of psychotropic drugs was considered according to those recommendations [13]. The cancer pain characteristics of the study population are presented in relation to sex, which may influence pain expression [27,28]. Inadequate cancer pain control and psychotropic drug use • Use of psychotropic drugs: the use of anxiolytics, hypnotics, and antidepressants was classified dichotomously (yes/no) if a prescription for those treatments (both continuous and on demand) was reported in the patient's medical record.

Data analysis
The data analysis included an initial descriptive analysis. Means with their standard deviation (SD; or median and interquartile range in the case of asymmetric distributions) were estimated for the continuous variables using the Student t test for comparisons or the Mann-Whitney U test in the case of non-normal distributions. We estimated proportions for the categorical and discrete variables by means of the Pearson chi-square test, likelihood ratio (LR) chi-square test, Wilcoxon two-sample test, and Fisher exact test as appropriate for comparisons. The use of each type of psychotropic drug was treated as a dependent variable; we performed separate analyses for anxiolytics, hypnotics, and antidepressants. Finally, we calculated a variable that included the joint use of anxiolytics and hypnotics. As a measure of association, we estimated the odds ratios (ORs) together with their 95% confidence intervals (95% CIs) by unconditional logistic regression.
To control for confounding bias in the association between presence of pain, pain control, and use of each psychotropic drug, we predefined the potential confounders. They were as follows: age; sex; marital status; education level; having children; type of cancer (location); chemotherapy; radiotherapy; hormone therapy; biological therapy; surgery; side effects of cancer treatment; and fatigue. We calculated the adjusted ORs for all the predefined potential confounders (maximum model). Furthermore, to obtain better precision in the adjusted estimates, we predefined three additional criteria based on the simulation studies of Mickey and Greenland [29] and Maldonado and Greenland [30]: 1. Associated with the presence or absence of pain with significance levels of P 0.20.
2. Associated with use of the psychotropic drug under analysis with an OR of 1.5-0.67 as criterion.
3. Causes a major change in the OR when adjusted: a change greater than 10% with the equation (ORc-ORa)/ORa >10% was considered a criterion.
For each psychotropic drug, we constructed an ad hoc optimal sub-model based on fulfilment of the above three criteria. We calculated the following: crude OR (ORc); adjusted OR (ORa) in the maximum model; and adjusted OR in the optimal sub-model.
We set the alpha error at 0.05, and all P were bilateral. All statistical analyses were performed using IBM SPSS Statistics, version 22.0. Table 1 shows the pain characteristics in the sample of 402 patients according to sex.
Among the 74 patients who reported pain, treatment was with analgesics in 72 (97.3%). With respect to the type of analgesic, strong opioids were prescribed in 41.7% (n = 30) and weak opioids in 5.6% (n = 4). Non-opioid analgesics were prescribed in 52.8% of patients (n = 38). The most common route of administration was oral, which used in 63.9% of cases (n = 46). With respect to pain control, among the 72 patients treated with analgesics, 45.8% (95% CI, 33.6-58.0) (n = 33) reported adequate pain control, whereas 54.2% (95% CI, 42.0-66.4) (n = 39) reported inadequate pain control; there were no statistically significant differences with sex (P = 0.108). Table 2 shows the use of psychotropic drugs according to the presence of pain. In all, 40.5% (95% CI, 28.7-52.4) (n = 30) of patients with pain reported taking some type of psychotropic drug. With respect to the type of psychotropic drug, 32.4% used anxiolytics, 12.2% hypnotics, and 4.1% antidepressants. The use of anxiolytics (P <0.001) and hypnotics (P = 0.004) was Inadequate cancer pain control and psychotropic drug use significantly higher among the patients with pain. Table 2 shows the potential confounders according to the presence of pain. Sex, the type of cancer (location), hormone therapy, and side effects of cancer treatment were statistically significantly associated with the presence of pain. Marital status, radiotherapy, and surgery to remove cancer were associated to a lesser degree, presenting significance levels of P 0.20. All these variables therefore met the first criterion for being included as confounders in the optimal models. The association of each of these variables with taking psychotropic drugs can be seen in S1 Table. Table 3 shows the associations among the presence of pain, pain control, and anxiolytic use. The presence of pain was statistically significantly associated with anxiolytic use: ORc, 3.10 (95% CI, 1.7-5.5). This association was strengthened after adjusting for all potential confounders in the maximum model: ORa maximum model, 3.64 (95% CI, 1.6-8.4; S2 Table). The association was also strengthened after the selection of confounders included in the optimum model based on established methodological criteria (sex, marital status, type of cancer [location], and side effects of cancer treatment): ORa optimum model, 3.15 (95% CI, 1.5-6.7). After ordinal categorization based on pain control, the associations showed a statistically significant dose-response pattern: ORc for 'inadequate pain control', 4.03 (95% CI, 2.0-8.3; linear P trend, <0.001); ORa maximum model, 4.98 (95% CI, 1.8-13.7; linear P trend = 0.001 S2 Table); and ORa optimum model, 4.74 (95% CI, 1.9-11.8; linear P trend = 0.001). Table 4 shows the associations among presence of pain, pain control, and hypnotic use. Statistically, the presence of pain was significantly associated with hypnotic use: ORc, 4.91 (95% CI, 1.9-12.8). This association was also strengthened after adjusting for all confounders in the maximum model (ORa maximum model, 5.57; 95% CI, 1.6-19.0; S2 Table) Table); and ORa optimum model, 6.09 (95% CI, 1.7-21.3; linear P trend = 0.003). By contrast, we found no association between the presence of pain and antidepressant use: ORc, 0.95 (95% CI, 0.27-3.39); ORa maximum model, 1.03 (95% CI, 0.2-4.8; S2 Table); and ORa optimum model, 0.99 (95% CI, 0.2-4.2). Details appear in Table 5.
When we considered joint use of anxiolytics and hypnotics, we were able to determine the associations between pain and pain control with greater precision: ORc, 3

Discussion
In our sample, 18% of patients reported cancer pain. Pain is the symptom most commonly reported by cancer patients with treatment, and it causes major suffering to patients and their families [31]. Our findings are in agreement with those of Fischer et al. [5], who found the prevalence of pain to be 20%-50% in a cancer population. However, our results are at variance with those of other studies, which reported the prevalence to be 40%, 50%, and even 90% of patients [6,22,32,33], depending on the type of cancer or disease stage. The findings of Fischer et al. [5] are used as a reference by the National Cancer Institute in their latest guidelines on cancer pain [21].
Since pain is such a widespread problem, it is particularly important to achieve effective treatment and symptom control. However, patients are often undertreated, which negatively affects their quality of life [21,33]. Complete, individualized patient assessment is therefore essential; it should address their pain history and involve a physical and psychosocial examination [20,34].  Inadequate cancer pain control and psychotropic drug use Pain control was inadequate in more than 50% of our patients. This proportion is similar to that reported by Greco et al. [3] in a review of 20 articles: control was found to be inadequate in up to 68% of patients, with a mean of 31.8%. Their data were collected from 2007 to 2013. Greco et al. found improvement in pain control with respect to a similar review published in 2008, where undertreatment was reported in up to 82% of patients [35].
In our sample, 20% of cancer patients took some type of psychotropic drug. This result is in line with the findings of Grassi et al. [36] in a review on mental disorders in cancer patients. Grassi et al. determined that 25%-30% of patients met the criteria for depression, anxiety, and insomnia (among other conditions).
Among our patients with pain, 32.4% took anxiolytics. This figure is higher than that reported by Desplenter et al. [37] in a general cancer population. However, our figure is lower than that determined by Syrowatka et al. [14] in a study of women with breast cancer; the authors found that 50.6% of patients received anxiolytics in the course of active treatment. Our figure is also lower than that reported by Barry et al. [15] among patients with metastatic cancer and pain; the authors observed anxiolytic use among 51% of patients. It should be noted that Barry et al. [15] reported that 38% of patients consumed three or more psychotropic drugs.
Although insomnia is often related to pain, only 12.2% of our patients with pain consumed hypnotics. That is because in the present study we investigated the use of hypnotics, not the presence of insomnia. Thus, it is possible that in our sample, many patients suffered from insomnia but were not receiving pharmacological treatment; some studies have reported a higher prevalence of insomnia in patients with pain [38,39]. However, that 12.2% figure is significantly higher than the proportion of patients without pain who took psychotropic drugs, which was 2.7% (P = 0.004).
We found that 4.1% of our patients with pain took antidepressants; that figure did not differ from the group of patients without pain. These results are much lower than those reported by Syrowatka et al. [14], Barry et al. [15], and Ng et al. [40]; those authors found the prevalence to be 22.4%, 64%, and 10.8%, respectively. However, Kierner et al. [41], who studied patients with cancer in late stages of the disease, reported that more than 75% and 90% of patients, respectively, took psychotropic and analgesic drugs. Their finding suggests that consumption may increase as the disease progresses We found that the presence of pain and (to a greater extent) inadequate pain control were associated with a higher risk of taking anxiolytics and hypnotics, but not antidepressants. When adjusted for the optimal model, with respect to patients with no pain, the possibility (odds) of taking anxiolytics and hypnotics increased 4.74-fold and 6.09-fold, respectively, in patients with inadequate pain control. There are several explanations for this association. Lack of pain control results in poorer patient quality of life and social relationships as well as in less tolerance to cancer treatment [42]. This situation, together with the associated sleep disorders, fatigue, anxiety, and depression [42][43][44][45], could result in increased psychotropic drug use. It may also be the case, however, that certain psychotropic drugs are prescribed as analgesic adjuvants for better pain control [11,24,46].
The cross-sectional nature of our study does not permit conclusions to be drawn in the above regard. Future longitudinal studies should investigate this association in greater depth. They should determine to what extent pain leads to more psychiatric symptoms (since that is the reason for taking psychotropic drugs) and to what extent psychotropic drugs are prescribed as adjuvants in the treatment of pain, without assessing psychiatric symptoms. However, the above association suggests that if pain can be reduced-following analgesic recommendations or using interventions based on relaxation or adaptive coping strategies [24]-the patient will require fewer psychotropic drugs and obtain better quality of life.
The present study shows that despite the latest recommendations [12,13], cancer patients still have uncontrolled pain. This may be because of the following factors: the treatment and control of cancer pain is very complex; the lack of clear consensus among experts [7,9]; the need to re-evaluate the patient to ensure efficacy of treatment [7,10]; the need for individualized treatment for each patient [2]; patients' fear of addiction; misconceptions about pain and illness [47]; ignorance about medical equipment in evaluating and managing pain; and lack of access or limited access to certain analgesics [48]. All those factors contribute to less pain control. There is therefore a need to continue research in this area.
In this study, we investigated only the presence of pain and its management and control independently of the duration. Our objective was to determine the association between pain control and psychotropic drug use. However, we believe that future studies should address any differences between acute and chronic pain.
In retrospective studies based on secondary information (records), the low quality of the information-owing to either incomplete records or lack of agreement among different records-can be a major limitation. For this reason, we selected variables that could be collected homogeneously in secondary records. However, we were unable to obtain information about symptoms of stress, anxiety, depression, and sleeping problems or identify the reasons for patients having been prescribed psychotropic drugs. We only derived information about the whole oncological situation at the time of data collection; oncological processes may, though, in some cases continue for several years.
The protocol in this study involved checking for discrepancies (contradictory information) individually with the physicians or nurses responsible for the patient; we conducted a sensitivity analysis by incorporating the information separately. We found 100% agreement regarding the analysed variables.
One limitation of this study was that the small number of patients who took antidepressants did not enable their association for inadequate pain control to be determined with sufficient precision. One of the strengths of the study was the control of confounding bias on the associations: applying the optimal model allowed the confounding bias to be minimized with the highest possible precision in the estimates.
In conclusion, we found the prevalence of pain in cancer patients to be almost 20%. Despite receiving analgesic treatment, pain was inadequately controlled in over 50% of patients. Both the presence of pain and (to a greater extent) inadequate pain control were associated with use of anxiolytics and hypnotics. This association did not appear to exist for antidepressant use.
Supporting information S1 Table. Association between potential confounders and psychotropic drug use. (DOCX) S2 Table. Association between cancer pain,effectiveness of analgesia and anxiolytic, hynotic and antidepressant consumption adding to the maximum model. (DOCX) S3 Table. Association between existence of pain, pain control, and anxiolytic and/or hypnotic use.