Study design

We conducted a multi-centre, hospital-based crosssectional descriptive and analytic study at the Rheumatology clinics of five tertiary institutions in Nigeria (University of Maiduguri Teaching Hospital, Jos University Teaching Hospital, University of Ilorin Teaching Hospital, Lagos State University Teaching Hospital, and University of Uyo Teaching Hospital) all spread across the northern, middle and southern parts of the country.

Study population

Individuals aged 18 years and above, satisfying the American College of Rheumatology (ACR) clinical classification criteria for knee OA [11], were recruited over a period of 3 months.

Sample size determination

Sample size was obtained using Fisher’s statistical formula for estimating minimum sample size in descriptive health studies for population sizes greater than 10,000 [12]. Based on the reported knee OA prevalence of 11.5% determined by Adebusoye et al. in Ibadan, Southwest Nigeria [13], the calculated sample size for the study was estimated to be 156. However, to increase the power of the study, we recruited a total of 250 patients at a rate of 50 patients per centre.

Sampling method

Patients were recruited using a systematic random sampling method in which every other patient with knee OA was recruited. Prior communication with the co-investigators representing each of the participating hospitals revealed that a median of 100 patients with knee OA were seen over a period of three months. With a recruitment plan of 50 participants per centre, a sampling interval of 2 was chosen. The first patient seen each day meeting the inclusion criteria for the study was chosen followed by the 3^rd patient and so on. If an individual did not satisfy the inclusion criteria, or the person refused to participate, the next individual was recruited. Patients’ folders were labelled and a research register was kept to avoid double sampling. The procedure was repeated each day until the sample size was attained.

Inclusion and exclusion criteria

Consenting patients aged 18 and older who satisfied the ACR clinical classification criteria for knee OA served as inclusion criteria. Patients with a history of previous knee surgery, known mental illness, previous diagnosis of insomnia, inflammatory arthritis, systemic infection, or malignancy and other symptomatic chronic diseases, e.g. diabetes mellitus, chronic liver disease, chronic obstructive lung disease, were excluded from the study.

Data collection

Informed verbal and written consent was obtained from all participants while a semi-structured interviewer-administered questionnaire was used to collect demographic and relevant clinical information. Pain severity was assessed using a 0–10 numerical rating scale (NRS), with 0 representing no pain and 10 being maximum pain experienced. A score of 4 or less was considered mild pain, 5–7 as moderate and 8 and above as severe pain. The WOMAC (Likert) pain subscale was also used to grade pain intensity. We considered cut-off points of less than 5, 6–10 and 11 to 20 as mild, moderate and severe pain respectively.

Individuals were grouped into four functional classes using the Steinbrocker functional classification [14]. Each knee was examined separately for the presence or absence of deformities, effusion, joint line tenderness and crepitus. The range of motion (ROM) and alignment of the knee joint were measured using an International Standard Goniometer. Quadriceps strength was graded using the Medical Research Council grading scale. The weight to the nearest 0.01 kg and height to the nearest 0.01 m were measured using a Seca electronic weighing/ height scale (Seca Deutschland, Hamburg Germany) using standard techniques. Body mass index (BMI) was calculated using the formula, weight (kg)/height² (m). For the purpose of our analysis we grouped patients as not obese (BMI < 30 kg/m²) or obese (BMI ≥ 30.0 kg/m²).

Blood pressure (BP) was measured with a mercury sphygmomanometer (Accosson, England) using standard techniques [15]. Standard antero-posterior (AP) and lateral semi-flexed radiographs of both knees in weight bearing position were taken by qualified radiographers of the radiology department of the hospitals. The radiographs were interpreted and graded by both the radiologists and the principal investigators using the Kellgren and Lawrence (KL) criteria [16]. The higher KL grade of the knee was used for the final analysis.

The sleep quality of respondents was assessed using the Pittsburgh Sleep Quality Index (PSQI) [17]. The 19-item questionnaire generates seven components: sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medications and daytime dysfunction. The global score ranges from 0 to 21, with higher scores indicative of poorer subjective sleep quality. The PSQI has been validated in Nigeria by Aloba et al. [18] with the best cut-off score set at 5. We dichotomized patients into two groups, scores less than 5 (no sleep abnormality) and score ≥ 5 (poor sleep quality).

Family functioning was assessed by the Family APGAR (Adaptation, Partnership, Growth, Affection, and Resolve) questionnaire. The measure consists of five parameters of family function: Adaptability, Partnership, Growth, Affection and Resolve. The instrument has a minimum score of 0 and maximum score of 10. Family APGAR scores are interpreted as follows: 7–10 = highly functional family; 4–6 = moderately dysfunctional family; 0–3 = severely dysfunctional family. The correlation of the instrument with the previously validated instrument (Pless-Satterwhite index) is 0.80 and the correlation with a clinical report is 0.64 [19]. Family APGAR has been validated and used for previous studies in Nigeria [20].

Depression was assessed using the Patient Health Questionnaire (PHQ-9) [21]. It has a minimum score of 0 and maximum score of 27. Component scores are rated 0 to 3. The interpretation of the total scores is as follows: 0–4 = no depression; 5–9 = mild depression; 10–14 = moderate depression; 15–19 = moderately severe depression; 20–27 = severe depression. It has been validated for use in Nigeria, having a good concurrent validity with the Beck Depression Inventory (BDI) (r = 0.84, p < 0.001) and good one month test-retest reliability (r = 0.84, p < 0.001) [22]. The sensitivity and specificity with 5 as the cut-off score for minor depressive disorder is 0.989 and 0.973 respectively, and 0.846 and 0.994 respectively with 10 as the cut-off for major depression [22]. PHQ-9 has also been shown to be a valid and reliable tool for assessing depressive disorders and depression severity among patients with knee OA [23]. We dichotomized patients into two groups: PHQ-9 < 5 (no depression) and PHQ-9 ≥ 5 (has depression).

Data analysis

All data obtained were analyzed using SPSS version 21.0 by IBM USA, Armonk, NY 10504. Demographic and clinical data were summarized using frequencies, percentages and proportions. Normality testing was performed using Shapiro-Wilk statistics. Means of normally distributed continuous variables of participants with poor sleep quality and those without poor sleep quality were compared using the t-test. Tests of association between poor sleep quality and each of the determinants under consideration (pain scores, functional status, family functioning, KL grades and depression) were done using the chi-square test with Fisher’s exact test used where frequencies were < 5. Pearson’s and Spearman’s correlation was used to determine correlations between sleep quality and the determinants under consideration. Predictors of poor sleep in patients with knee OA were determined using multiple logistic regression analysis with PSQI score <5 and ≥ 5 as a binary outcome. A p-value < 0.05 was considered statistically significant for all tests.

The study was approved by the Ethical Review Committee of each of the study centres.

Demographic characteristics

The 250 participants comprised 209 (83.6%) females and 41 (16.4%) males. One hundred and forty-one (56.4%) had PSQI scores ≥ 5, indicative of poor sleep quality. Their mean age was 59.90±10.62 years, and 97 (38.8%) were 65 years or older. There was no significant difference in mean ages of those with and without poor sleep (p = 0.353). The majority (74.4%) had a formal education, 185 (74%) were married, 66 (26.4%) were unemployed. The mean BMI was 30.82 ±5.48. Patients with poor sleep had higher BMI compared to those without poor sleep (p = 0.04). Poor sleep was significantly associated with level of education and smoking (p = 0.001 and p = 0.019 respectively). Table I shows the socio-demographic distribution of the patients.

Relationship between sleep quality and pain, functional class, APGAR score, depression and Kellgren-Lawrence grade

The mean ±SD WOMAC pain score was 8.83 ±3.84. Participants with poor sleep had higher mean pain scores compared to those with good sleep quality (9.60 vs. 7.83, p < 0.001). One hundred and eighteen (83.7%) patients with poor sleep quality had moderate to severe WOMAC pain scores, compared to 66 (63.1%) patients without poor sleep (p < 0.001). The proportion of participants with poor sleep increased as pain severity increased (p for trend < 0.0001). Eighty-four (59.6%) patients with poor sleep had depression (PHQ-9 score ≥ 5). Poor sleep was strongly associated with depression (p < 0.0001) and the proportion of participants with poor sleep quality and depression was higher than those with good sleep quality and depression (59.6% vs. 19.3%, p < 0.0001). Poor sleep quality was not associated with functional class (p = 0.854) or APGAR score (p = 0.137) of the participants (Table II).

Radiographic findings

Most participants (97.2%) had KL grade II and above. One hundred and eighty-six (74.6%) had medial knee OA, 156 (62.4%) had lateral knee OA, while 149 (59.6%) had patello-femoral knee OA. The proportion of participants with a PSQI score ≥ 5 (poor sleep quality) was higher among individuals with medial and patello-femoral knee OA compared to those with a PSQI score ≤ 4 (p = 0.024 and p = 0.002 respectively). There was no significant association between lateral knee OA, KL grade and poor sleep (p = 0.103 and p = 0.333 respectively), as shown in Table III.

Predictors of poor sleep quality

After analyzing variables significantly associated with PSQI scores using multiple logistic regression, higher PHQ-9 scores, WOMAC pain scores and higher level of education were found to be the best predictors of poor sleep quality among the participants (p < 0.001, p = 0.035 and p = 0.013 respectively). Participants with depression and knee OA were 1.27 times more likely to have poor sleep than those without depression, while those with higher WOMAC pain scores were 1.11 times more likely to have poor sleep compared to those with lower pain scores (Table IV).