Ethical approval for this study was granted by the Mbarara University of Science and Technology Research Ethics Committee (MUST-2022-396) and the Uganda National Council for Science and Technology (HS2395ES). Additionally, administrative clearance from the Hospital Director at Hospital Director at Mbarara Regional Referral Hospital was sought before commencing the study. We obtained written informed consent from each participant and/or their legal guardian prior to their involvement in the study. In the case of illiterate participants, written informed consent was acquired from their authorised representatives. To maintain confidentiality, study numbers were used instead of names on all data collection tools, and participant identifiable information was separated during data analysis. Interactions between participants and the research team occurred in a private and comfortable setting, free from disturbances and accessible to only one participant at a time. Women diagnosed with cervical intraepithelial lesions (cases) received the standard care package at the cervical cancer clinic. The nurse at the clinic provided participants with their visual inspection with acetic acid, Pap smear, histology, and lipid profile results. All methods employed in this study adhered to relevant guidelines and regulations.

This study was conducted at the cervical cancer clinic of Mbarara Regional Referral Hospital, between 01 December 2022 and 27 February 2023. Mbarara Regional Referral Hospital is a regional referral hospital for the entire south-western Uganda and it is located in Mbarara city, approximately 260 km from the capital city, Kampala. This clinic offers services to an average of 75 women per week. It receives women from over 13 districts of the entire south-western region of Uganda plus the neighbouring countries including Tanzania, Rwanda and Democratic Republic of Congo.

We conducted this unmatched case-control study including women who visited the Mbarara Regional Referral Hospital cervical cancer clinic. Cases were defined as women with a positive visual inspection with acetic acid result who underwent Pap smears and histology, subsequently confirmed to have CIN in accordance with the clinic’s standard care. Controls consisted of women with a negative visual inspection with acetic acid result and those confirmed to be devoid of cervical lesions based on cytology and histology.

We calculated our sample size using the online OpenEpi, Version 3, open-source calculator-SSCC (OpenEpi – Sample Size for Unmatched Case-Control Studies). For this calculation, we considered a two-sided confidence level of 95, a study power of 80% and a case-to-control ratio of 1. We took into account a proportion of cases with dyslipidaemia of 40.99%,²⁵ along with a proportion of controls with dyslipidaemia of 18.8%,²⁶ together with a least extreme odds ratio of 3.0. This is based on previous studies which have demonstrated a high variability in prevalence of dyslipidaemia in Africa, ranging from 5.2% to 89%.²⁷ These studies offer the most accurate depiction of the cervical cancer study population that we could find.²⁵ This gave a total of 75 cases and 75 controls using the Fleiss module with continuity correction.²⁸ After factoring in the possible 15% attrition rate, the resultant sample size came to at least 86 cases and 86 controls.

We prospectively selected cases and controls by using purposive sampling to recruit cases after cytology or histology specimen collection and following a positive screening test. Then we recruited a control each time a case was identified, that is, the incidence density sampling method as described previously.²⁹

We included women aged 21 years and above, who visited the study site during the time of the study and who consented to participate. We excluded any woman who was undergoing any form of treatment with any lipid-altering drugs such as antihypertensive and cholesterol-lowering drugs.

After collection by research assistants and the Principal Investigator, data were entered into an Excel spreadsheet (Microsoft Office Professional Plus 2013, version 15.0.4675.1003, Microsoft Inc., Redmond, Washington, United States) and then imported into STATA 13 (StataCorp LLC, College Station, Lakeway, Texas, United States) software. We used descriptive statistics including frequencies, means ± standard deviations, median values and interquartile ranges to characterise the population. Differences between respective groups were tested using the standard t-tests, Chi-square tests and the Fisher’s exact tests. Associations between CIN grades and specific dyslipidaemias were derived using logistic regression analysis. Cases with CIN2 or CIN3 were combined into a single group called ‘CIN2+’, because of the small number of cases in each group. All statistical outcomes were reported with 95% confidence intervals and p-values of < 0.05 were considered significant. We performed both bivariate and multivariate logistic regression analysis. Factors which were biologically plausible in cervical cancer and those whose p-values were ≤ 0.2 at bivariate analysis were considered for multivariate logistic regression analysis. These factors included age, smoking, contraceptive use, level of education, history of high blood pressure and history of diabetes.

In this case-control study, we enrolled 93 cases and 93 controls (Table 1). There was a statistically significant variation in age between the two groups (p < 0.001); the mean age of participants control group was 38.87 years and in the case group was 34.57 years. Additionally, more than half of the participants fell within the age group of 30 years – 49 years and this observation was statistically significant (p = 0.044). There were notable statistically significant differences in the highest level of education attained among study groups, with 13 out of 93 (13.9%) controls and 3 out of 93 cases (3.2%) reporting never having studied. Contraceptive use was reported by 38 out of 93 (40.9%) controls and 56 out of 93 (60.2%) cases. Among the cases, 7 out of 93 reported being smokers, resulting in a smoking prevalence of 7.6% among cases.

Out of the 93 cases, 81 had CIN1 while 12 had CIN2+. The mean triglyceride levels among cases was 95.9 mg/dL and among controls was 123.7 mg/dL, demonstrating a statistically significant difference (p = 0.001). High triglycerides showed a prevalence of 13.9% (12/93) among the controls and a lower prevalence of 3.7% (3/81) among cases, though this difference in distribution was not statistically significant (p = 0.08). High LDL showed a prevalence of 12.9% (12/93) among controls and a prevalence of 3.7% (3/81) among cases, and this distribution was also not statistically significant (p = 0.08). The prevalence of low HDL was higher in CIN2+ cases (66.7%; 8/12) although not statistically significant when compared to controls (p = 0.086) (Table 2).

From univariate logistic regression analysis, both high serum triglycerides and high serum LDL were found to be associated with CIN (Table 3). However, in multivariate logistic regression analysis, with adjustments made for age, smoking, contraceptive use, level of education, history of high blood pressure and diabetes, only high serum triglycerides was significantly associated with CIN (odds ratio: 1.395, 95% confidence interval: 0.084–1.851, p = 0.007). On the other hand, high serum LDL showed borderline association with CIN (odds ratio: 0.251, 95% confidence interval: 0.061–1.047, p = 0.058).

This study demonstrates the association between triglyceride dyslipidaemia and CIN. With increasing prevalence of non-communicable diseases, patients would benefit from combined screening for cervical cancer as well as dyslipidaemia, preferably on a single visit. Just as for cervical cancer screening, a test and treat approach would be more impactful if coupled with screening for dyslipidaemia.

A major limitation of this study was the fact that controls were not matched with cases. Some confounders may still have had an effect on statistics. This further presents a shortcoming for case-control studies, in addition to their inability to ascertain causation. However, to the best of our knowledge, we gathered information on all factors known to be associated with cervical lesions and performed logistic regression analysis which generated a logistic regression model with all the factors included. We also expect our study to have suffered from selection bias as well as recall bias during participant enrollment. During collection of data, participants were required to remember information on a number of variables including history of diabetes and hypertension. These could have led to false conclusions about the study population since a lot of data may have been left out or improperly reported during data collection. However, for most of the variables that required participants to recall, we did our best to compare the provided information with participants’ charts. We also observe that the sample size used for this study was smaller than what is observed in other studies, despite the adequate power. This therefore might have a negative impact on the generalisability of results to the general population.

We observed a significant association between dyslipidaemia, especially triglyceride dyslipidaemia, and CIN. This underscores the importance of expanding screening efforts for dyslipidaemia to effectively manage potential complications and other non-communicable diseases associated with dyslipidaemia. Furthermore, we advocate for prospective cohort studies, with larger sample sizes, to deepen our understanding of the relationship between dyslipidaemia and CIN, as well as its broader implications for cervical cancer. Adequate management of dyslipidaemia could stop progression of CIN to cervical cancer. Molecular studies could be carried out to understand the biochemical processes and interactions behind lipids and cervical carcinogenesis.