Methodological suggestions for evaluating cervical cancer screening in India: A critical comment on a recent NFHS-5-based study

We read with great interest the article by Maheshwari et al.,[1] titled “Prevalence and predictors of cervical cancer screening among married women in India: A cross-sectional analysis of National Family Health Survey-5 data.” Cervical cancer, despite being preventable, continues to rank as the second most common cancer and cause of cancer-related deaths among Indian women.[2] Therefore, analyzing the prevalence and predictors of cervical cancer screening, among married women using nationally representative data is both critical and timely from a public health and policy perspective. Previously, Yadav et al.,[3] also evaluated the prevalence and determinants of cervical cancer screening among Indian women using NFHS-5 data. The study population in Maheshwari et al.,[1] was married women aged 30-49 years, whereas Yadav et al.,[3] included women aged 15-49 years. The Government of India’s (GOI) operational guidelines recommend screening for women aged 30-65 years.[4] Furthermore, the National Cancer Registry Program has reported that the peak age for cervical cancer is in the 50-75 + age group.[5] However, NFHS includes women aged only 15-49, thereby excluding a significant portion of the target age group, especially women aged 50 and above. This age limit may significantly underestimate accurate screening coverage and programmatic reach. This should have been acknowledged as a limitation. It is alarming that cervical screening rates are as low as 0.2% in some states like West Bengal and Assam, compared to the national average of 1.9%. These rates are far below the 70% coverage needed to significantly reduce cancer incidence and mortality.[6] Geographical variation plays a substantial role in healthcare accessibility and utilization in India. Including region as a stratifying variable or predictor in future models could help identify priority areas for targeted interventions. The authors highlighted various socio-demographic factors (age, education, place of residence, religion, wealth index), health (body mass index and co-morbidities), and behavioral (contraception, tobacco, and alcohol use) determinants of screening coverage. A previous study also considered other factors such as barriers to accessing health facilities, autonomy in healthcare decisions, social group affiliations, number of sexual partners in the past year, and awareness of sexually transmitted infections, which may impact on the outcome variable.[3] The authors used the negative binomial regression (NBR) model to identify predictors of cervical cancer screening. However, the outcome variable “Have you ever been screened for cervical cancer?” is a binary nominal variable (“Yes” or “No”). While NBR is appropriate for over-dispersed count data, it may not be suitable here, as the outcome is not a count but a binary response.[7] The authors could have considered using multivariable logistic regression or another analytical model more appropriate for binary outcomes. The authors noted that the cervical cancer screening questions differed between NFHS-4 and NFHS-5, affecting coverage comparisons. However, it is also important to recognize that NFHS-5 uses medical terminology (e.g., “cervical”) in vernacular translations.[8] Given the limited awareness of cervical cancer, especially in rural populations, this may have led to ascertainment bias. Furthermore, NFHS-5 does not report the time-frame of the last screening, which ideally should align with the GOI’s recommended intervals of 3-5 years.[4,8] We noted these limitations in our own study on cancer screening uptake among women in Uttar Pradesh.[8] The efforts of the GOI in capturing data on cancer screening uptake through NFHS-5 are commendable. Future surveys could benefit from including additional questions on topics such as screening intention, timing of the most recent screening, prior screening experiences or symptoms, and perceived barriers to or reasons for non-participation. Furthermore, to ensure inclusivity, survey questions should be understandable regardless of the respondent’s health literacy. Aligning data collection with national screening age guidelines, using locally understood terminology, and accounting for regional disparities will provide more robust and actionable insights for policymakers.[9] Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.