The Effects of Frequent Coffee Drinking on Female-Dominated Healthcare Workers Experiencing Musculoskeletal Pain and a Lack of Sleep

Previous research has demonstrated that chronic diseases can occur due to musculoskeletal (MS) pain and poor sleep. It is also worth noting that the caffeine in coffee can reduce overall sleep duration, efficiency, and quality. Thus, the present study examines the effects of frequent coffee drinking (two cups per day) on individuals experiencing MS pain and a lack of sleep during the COVID-19 period. This observational and cross-sectional study recruited 1615 individuals who completed the self-reported (Nordic musculoskeletal) questionnaire. Long-term, frequent coffee drinking and a sleep duration of less than 6 h per day were significantly associated with neck and shoulder pain among healthy individuals. The mediation model demonstrated that the shorter sleep duration and drinking multiple cups of coffee per day had a two-way relationship that worsened such pain over the long term. Specifically, individuals who experienced such pain frequently drank multiple cups of coffee per day, which, in turn, shortened their sleep durations. In summary, long-term coffee drinking creates a vicious cycle between MS pain and sleep duration. Therefore, the amount of coffee should be fewer than two cups per day for individuals who sleep less than 6 h per day or suffer from MS pain, especially neck and shoulder pain.


Introduction
Coffee is one of the most popular beverages for people of many ages. It is a complex chemical mixture that contains caffeine, which is a purine alkaloid that is naturally found in coffee beans [1] and contributes to its bitterness [2]. Caffeine stimulates the central nervous system, which can increase alertness, blood circulation, and respiration [3]. However, caffeine has biphasic effects, i.e., lower doses can provide some behavioral stimulation, whereas higher doses can lead to anxiety, aversion, irritability, and discomfort [4]. Despite clinical studies demonstrating the adjuvant analgesic effects of caffeine [5], long-term coffee drinking can negatively affect health and musculoskeletal (MS) pain. In health, individuals drinking more than five cups of coffee per day can have an increased risk of myocardial infarction or unstable angina [6]. In MS pain, related research showed that drinking more than seven cups of coffee per day was associated with a higher risk of knee osteoarthritis among Korean men [7]. Interestingly, patients with chronic back pain tend to drink two times as much caffeine as those without such pain [8], whereas individuals with related factors in the preceding year. The pain sites on the NMQ were classified as the neck, left or right shoulder, upper back, waist or lower back, left or right elbow, left or right wrist, left hip/thigh/buttock, right hip/thigh/buttock, left or right knee, and left or right ankle. The options for the frequency of each pain site were every day, once a week, once a month, once every half a year, and at least once every half a year, scored as 100, 80, 60, 40, and 20 points, respectively. Factor analysis was also adopted in the NMQ to determine the underlying variables that could effectively explain most of the questionnaire items. Through varimax rotation, the standardized scoring coefficients constituted new factor loadings and were defined according to their corresponding significance. The new factors that featured vector values exceeding 1 were retained according to the principle proposed by Hair et al. [50].
In the questionnaire, the basic response options included male or female for gender; age; "married" or "other" for marriage; and "without child," "one child," "two children," "three children," and "more than three children" for having children. The survey also asked if the participants engaged in leisure activities with family/friends during vacation time. The response options included "always," "often," "sometimes," "seldom," and "never." Regarding their education, the response options were "master's degree or above" and "university degree or below," while the response options for self-reported sleep duration per day included "less than 5 h," "between 5 and 6 h," "between 6 and 7 h," "between 7 and 8 h," and "more than 8 h." As for their coffee intake per day, the response options were "more than 2 cups per day," "2 cups per day," "1 cup per day," "occasionally," and "never." Regarding their alcohol use, the response options included "alcohol use in a month" and "no alcohol use in a month," while the response options were "yes" and "no" for exercising at least once a week. As for their overtime work, the response options were "seldom," "fewer than 45 h per month," "45 to 80 h per month," and "more than 80 h per month," while "irregular," "regular," "night," and "day" were the response options for shift schedules. Finally, the participants were classified as physicians, nurses, professional and technical personnel, and administrative staff. They were also asked about the presence of chronic diseases. In this regard, the presence of one or more diseases was classified as a "yes" response.
Regarding the statistical methods, factor analysis [50] was adopted for the NMQ to determine new underlying variables, while a t-test or one-way ANOVA was adopted to examine the differences between the continuous variables. Additionally, a chi-square test or Fisher's exact test was conducted to determine the significant differences in the categorical variables, while simple/multiple linear or logistic regression was used to examine the correlation between the dependent variable (DV) and the independent variable (IV), in the absence (or presence) of the controlled variables. The mediation effects among the IV, DV, and mediator were based on the following strategy proposed by Baron and Kenny [51]: 1) in the presence of the first-stage effect, the IV significantly affects the mediation factor; 2) in the absence of the mediation factor, the IV significantly affects the DV; 3) in the presence of the second-stage effect, the mediation factor has a significant effect on the DV; and 4) the effect of the IV on the DV weakens upon the addition of a mediation factor in the model.
A mediation model suitable for combining the categorical and continuous variables was developed by Iacobucci (2012) [52]. The formulas are as follows: If the mediation factor and dependent variables are continuous variables, then the original formula of the Sobel test is applicable: If the mediation factor or dependent variables are categorical variables, then the original formula of the Sobel test is rederived into a new formula: Among them, a is the simple linear or logistic regression coefficient for the independent variable against the mediation factor, while b is the regression coefficient for the mediation factor against the dependent variable in the binary linear or logistic regression model. Additionally, s a and s b represent the standard deviations of a and b, respectively, while the results exceeding |1.96|, |2.57|, and |3.90| (for the two-tailed test) are significant at α = 0.05, 0.01, and 0.0001, respectively. In this study, the analyses were performed using SAS Enterprise Guide 7.1 software (SAS Institute Inc., Cary, NC, USA), and the significance was set at p < 0.05.

Results
Regarding the detailed description, the description of the basic demographics, sleep duration per day, and coffee intake of 1615 participants are shown in the Supplementary Information Tables S1-S3. The results demonstrated that marriage (p = 0.016), engaging in leisure activities with family/friends (p < 0.0001), coffee intake per day (p < 0.0001), exercise at least once a week (p = 0.008), overtime work in a month (p < 0.0001), shift schedules (p < 0.0001), and profession (p = 0.005) were associated with sleep duration per day. In addition, gender (p = 0.024), age (p < 0.0001), marriage (p < 0.0001), having children (p < 0.0001), education (p < 0.0001), alcohol use (p < 0.0001), exercise at least once a week (p = 0.002), and profession (p = 0.001) were related to coffee intake. Table 1 illustrates that the common pain sites included both shoulders (43.09%), neck (36.22%), waist or lower back (27.93%), and upper back (16.90%). According to the principle proposed by Hair and Anderson (1995) [50], Factors 1 and 2 were retained because their vector values exceeded 1. In addition, the factor loadings were converted into standardized scoring coefficients through varimax rotation. The relatively large factor loading values for Factors 1 and 2 corresponded to the neck and both shoulder pain and both ankle pain sites, respectively. Thus, Factors 1 and 2 were redefined into two new variables: the neck and both shoulder pain (NBSP) score and the both ankle pain (BAP) score. According to Table 2, there were significant differences in the NBSP scores for gender (p < 0.001), age (p = 0.003), marriage (p = 0.003), having children (p = 0.006), education (p = 0.034), sleep duration per day (p < 0.001), coffee intake per day (p = < 0.001), alcohol use (p = 0.001), exercise at least once a week (p = 0.001), overtime work per month (p < 0.0001), profession (p = 0.036), and suffering from chronic diseases (p < 0.0001). There were no significant differences in the BAP scores among the survey variables, except for education (p < 0.0001). Regarding the other survey variables, the females obtained higher NBSP scores than the males (0.04 ± 0.93 vs. −0.17 ± 0.84). Moreover, individuals who were 38-45 years of age (0.15 ± 0.96), were married (0.07 ± 0.96), were parents (0.07 ± 0.96), had a master's degree or above (0.11 ± 0.99), had a sleep duration of less than 5 h (0.26 ± 1.04), drank more than two cups of coffee per day (0.61 ± 1.25), used alcohol in a month (0.10 ± 0.97), had no weekly exercise (0.09 ± 0.97), worked overtime more than 45 h per month (0.54 ± 1.35/ 0.44 ± 1.14), were nurses (0.08 ± 0.94), or suffered from chronic diseases (0.20 ± 1.03) achieved higher NBSP scores than the others. Simple multiple linear or logistic regression was also used to examine the correlation between the dependent and independent variables in the absence (or presence) of the controlled variables. Since the number of individuals with a sleep duration of less than 5 h per day was only 63 (Table 2), the variables of sleep duration of less than 5 h per day and sleep duration between 5 and 6 h per day were combined into a new variable: sleep duration of less than 6 h per day (SLD < 6 h). Furthermore, since the number of individuals who drank more than two cups of coffee per day was only 26, the variables of drinking more than two cups of coffee per day or drinking just two cups of coffee per day were combined into a new variable: drinking multiple cups of coffee (MCC) per day.
As shown in Table 3, which presents the effects of sleep duration per day and drinking MCC per day on NBSP, drinking MCC per day was significantly associated with an increased level of NBSP in the simple or multiple linear regression models (B = 0.32, p = 0.001; 0.23, p = 0.016), while SLD < 6 h was significantly associated with an increased level of NBSP in the simple or multiple linear regression models (B = 0.21, p < 0.0001; 0.15, p = 0.001). The results in Table 3 confirm Hypotheses 1 and 2 (illustrated in the introduction), that coffee intake (more than two cups per day) and a shorter sleep duration (less than 6 h per day) are associated with MS pain (especially neck and both shoulders pain).
Finally, this study used mediation analysis to determine the existence of a mutual relationship between coffee, MS pain, and sleep. Figure 1.1 demonstrates that drinking MCC per day mediated the effect of SLD < 6 h on increased levels of NBSP (Z m = 2.27, p < 0.05). Lack of sleep (<6 h) also caused individuals to drink more coffee per day, which led to more frequent neck and shoulder pain. Figure 1.2 illustrates that SLD < 6 h mediated the effect of drinking MCC per day on increased NBSP (Z m = 2.95, p < 0.01). Overall, Figures 1 and 2 demonstrate that long-term coffee drinking and lack of sleep can further increase the occurrence of neck and shoulder pain. Finally, this study used mediation analysis to determine the existence of a mutual relationship between coffee, MS pain, and sleep. Figure 1.1 demonstrates that drinking MCC per day mediated the effect of SLD < 6 h on increased levels of NBSP (Zm = 2.27, p < 0.05). Lack of sleep (<6 h) also caused individuals to drink more coffee per day, which led to more frequent neck and shoulder pain. Figure 1.2 illustrates that SLD < 6 h mediated the effect of drinking MCC per day on increased NBSP (Zm = 2.95, p < 0.01). Overall, Figures 1 and 2 demonstrate that long-term coffee drinking and lack of sleep can further increase the occurrence of neck and shoulder pain. Figure 3 adopts SLD < 6 h and NBSP as dependent and independent variables for the mediation model, respectively. Based on the findings, drinking MCC per day mediated SLD < 6 h and increased NBSP (Zm = 2.5, p < 0.05). Specifically, individuals who suffer from neck and should pain tend to drink coffee to cope with such pain. However, it eventually decreases their sleep duration per day. Figures 1-3 confirmed Hypothesis 3 and determined that coffee intake (more than two cups per day) really opens the vicious circle between lack of sleep (less than 6 h per day) and MS pain (especially neck and shoulders pain).     Figure 3 adopts SLD < 6 h and NBSP as dependent and independent variables for the mediation model, respectively. Based on the findings, drinking MCC per day mediated SLD < 6 h and increased NBSP (Z m = 2.5, p < 0.05). Specifically, individuals who suffer from neck and should pain tend to drink coffee to cope with such pain. However, it eventually decreases their sleep duration per day. Figures 1-3 confirmed Hypothesis 3 and determined that coffee intake (more than two cups per day) really opens the vicious circle between lack of sleep (less than 6 h per day) and MS pain (especially neck and shoulders pain).

Figure 2.
Mediation effect of SLD < 6 h on MCC per day and NBSP. * p < 0.05; ** p < 0.01; *** p < 0.0001; M, mediation factor; I, independent variable; D, dependent variable; c is the simple linear or logistic regression coefficient for the independent variable against dependent variable in the absence of mediation factor; c' is the binary linear or logistic regression coefficient for the independent variable against dependent variable in the presence of mediation factor; is the simple linear or logistic regression coefficient for the independent variable against the mediation factor; is the regression coefficient for the mediation factor against the dependent variable in the binary linear or logistic regression model; and represent the standard deviations of and .

Figure 3.
Mediation effect of MCC per day on NBSP and SLD < 6 h. * p < 0.05; ** p < 0.01; *** p < 0.0001; M, mediation factor; I, independent variable; D, dependent variable; c is the simple linear or logistic regression coefficient for the independent variable against dependent variable in the absence of mediation factor; c' is the binary linear or logistic regression coefficient for the independent variable against dependent variable in the presence of mediation factor; is the simple linear or logistic regression coefficient for the independent variable against the mediation factor; is the regression coefficient for the mediation factor against the dependent variable in the binary linear or logistic regression model; and represent the standard deviations of and .

Discussion
The present study confirms three hypotheses and determined that long-term heavy coffee intake (two cups per day) and a shorter sleep duration (<6 h per day) are associated with neck and shoulder pain. Notably, long-term heavy coffee intake plays a mediating factor in the vicious circle between shorter sleep duration and neck and shoulder pain. In addition, alcohol use, the lack of regular exercise at least once a week, overtime work in a month, and the presence of chronic diseases were significantly associated with pain in the neck/shoulders or ankles. Figure 3. Mediation effect of MCC per day on NBSP and SLD < 6 h. * p < 0.05; ** p < 0.01; *** p < 0.0001; M, mediation factor; I, independent variable; D, dependent variable; c is the simple linear or logistic regression coefficient for the independent variable against dependent variable in the absence of mediation factor; c' is the binary linear or logistic regression coefficient for the independent variable against dependent variable in the presence of mediation factor; a is the simple linear or logistic regression coefficient for the independent variable against the mediation factor; b is the regression coefficient for the mediation factor against the dependent variable in the binary linear or logistic regression model; s a and s b represent the standard deviations of a and b.

Discussion
The present study confirms three hypotheses and determined that long-term heavy coffee intake (two cups per day) and a shorter sleep duration (<6 h per day) are associated with neck and shoulder pain. Notably, long-term heavy coffee intake plays a mediating factor in the vicious circle between shorter sleep duration and neck and shoulder pain. In addition, alcohol use, the lack of regular exercise at least once a week, overtime work in a month, and the presence of chronic diseases were significantly associated with pain in the neck/shoulders or ankles.
Related studies have illustrated that reduced alcohol use [18][19][20][21], physical activity [21], and fewer work hours [14] could reduce the risk of MS pain. In addition, individuals with chronic diseases [26,27] have a high risk for MS pain. These risk factors are consistent with our findings.
A literature review on healthcare workers demonstrated that MS pain occurred primarily in the lower and upper back, neck, and shoulders [53]. Our study found that the common pain sites were the shoulders (43.09%), neck (36.22%), waist or lower back (27.93%), and upper back (16.90%), which was consistent with the findings of a previous study.
Only 25% of adults achieve the recommended minimum sleep duration of 7 h per night for healthy adults [39]. However, only 16.72% of healthcare workers in the present study satisfy the 7 h sleep condition, as shown in Table 2. Therefore, the lack of sleep could be a common problem among healthcare workers in Taiwan, and this should be noted and further explored.

First Hypothesis: Coffee Intake Is Significantly Associated with an Increased Risk of MS Pain
Previous studies have found that individuals with a high caffeine intake (4-12 cups/day) had more severe pain than those with a low (0.25-1.5 cups/day) or moderate (2-3.5 cups/day) caffeine intake [54]. In addition, men who drink more than seven cups of coffee per day have an increased risk of knee osteoarthritis [7]. Our study determined that among healthy individuals, long-term drinking of more than two cups of coffee per day was associated with frequent neck and shoulder pain ( Table 3, B = 0.23, p = 0.016). Since the half-life of caffeine is approximately 4 h [55], drinking two or more cups of coffee per day can reach the threshold of caffeine's effect on MS pain, depending on one's genetics [56]. Based on these results, we can confirm our first hypothesis.

Second Hypothesis: Individuals with Shorter Sleep Durations Are More Susceptible to MS Pain
Evidence suggests a close link between short sleep durations and impairments in several physiological responses, including pain [57]. A study on middle-aged adults in the U.S. demonstrated that a sleep duration of <6 h was associated with greater next-day pain [58]. In addition, individuals who reported >6 h of sleep were more likely to have improved pain conditions [59]. Our study found important evidence that sleep duration was associated with MS pain at specific sites. Individuals with sleep durations of <5 or 6 h tend to experience more neck and shoulder pain than others ( Table 2; mean = 0.26 ± 1.04, 0.12 ± 1.00). Table 3 shows that a sleep duration of <6 h per day was significantly associated with increased neck and shoulder pain in the multiple regression model (B = 0.15, p = 0.001). These results confirm our second hypothesis.

Third Hypothesis: Coffee Intake Could Lead to a Vicious Circle between Lack of Sleep and MS Pain
A previous study showed that individuals who reported a sleep duration of <6 h consumed 3.6 times more caffeine per day than those who reported a sleep duration of >8 h [40]. Our study of healthcare workers found that individuals who reported a sleep duration of <6 h consumed 2.69 times (Figure 1.1, β = 0.99, odds ratio = e0.99 = 2.69, p < 0.0001) more caffeine per day than those who reported a sleep duration of >6 h. This close relationship between sleep duration and coffee intake indicates a causal relationship between sleep duration, NBSP, and coffee intake.
The mediation model in Figure 1.1 demonstrates that individuals who had shorter sleep durations tended to drink multiple cups of coffee, which can lead to increased MS pain (Zm = 2.27, p < 0.05). In addition, the mediation model in Figure 1.2 shows that individuals who chronically drink multiple cups of coffee generally experience shorter sleep durations and increased MS pain (Zm = 2.95, p < 0.01). These mediation models regarding coffee intake, MS pain, and sleep duration show that long-term heavy coffee intake (more than two cups per day) plays a mediating role in the two-way association of sleep duration <6 h and NBSP. Specifically, long-term heavy coffee intake will induce a vicious circle of sleep and neck and shoulder pain. These results are consistent with our third hypothesis.
This study has several limitations. First, we used the number of cups to measure the degree of caffeine intake per day. However, this is not an exact measurement method because cups have different volumes. Second, different coffee-brewing methods can lead to varied caffeine concentrations and errors in the dose-response of caffeine on MS pain. However, we believe that the differences in volume and caffeine concentration can be overcome. In addition, caffeine's effect on MS pain reaches the threshold depending on one's genetics [56]. Therefore, the threshold of more than two cups of coffee per day might not be suitable for other countries or races. Additionally, since sleep duration and sleep quality are subjective, future research should adopt other scales to measure sleeprelated issues. Third, MS pain can be the result of workloads, work styles, or posture. Unfortunately, our study did not collect such data in the regression models. Fourth, the effects of caffeine on individuals can be associated with genetics [56] and nationality. For example, the effects of coffee intake on MS pain in Europeans or Americans may differ from our results because our participants were Taiwanese.
Despite the adjustment for sex in the multiple linear regression, the results of the present study could be better suited to women because female participants accounted for >80% of the study population. Regarding the sex difference in MS pain, it could be caused by estrogen and progesterone. For instance, testosterone, the major male sex hormone, protects men from chronic MS pain [60]. Because the study population only included physicians, nurses, professional and technical personnel, and administrative staff, we added "healthcare workers" in the title to limit the applicability to occupational groups.
Notably, we could determine whether high work stress or emotional exhaustion caused by the pandemic affected the findings; thus, a similar study during the nonpandemic period should be replicated, and its results compared with those from the pandemic period. Finally, the mediation models in our study could be biased [61] because the relationship was based on a higher risk of judgment. Therefore, we excluded the phrase "causal relationship" to avoid confusion.

Conclusions
The present study determines that keeping good living habits (such as decreased alcohol use, regular exercise a week, and sufficient sleep), maintaining physical health (such as staying away from chronic diseases), and avoiding overtime work are ways to lower the risk of MS pain. We further examined the effects of frequent coffee drinking on individuals experiencing MS pain and lack of sleep. Based on the results, neck and shoulder pain was the most common among the healthcare workers. In addition, a sleep duration of less than 6 h and drinking more than two cups of coffee per day increased the occurrence of such pain, while controlling for other risk factors. Notably, long-term heavy coffee drinking created a vicious cycle between neck and shoulder pain and sleep duration of less than 6 h. The implication of the findings is that individuals who sleep less than 6 h, or who suffer from neck and shoulder pain, should limit their coffee intake to two cups per day.