Sampling constraints, sparse-cell estimates, and cross-sectional inference: considerations on hr-HPV prevalence and associated factors among HIV-positive women in Mekelle, Ethiopia

Abstract

Shfare et al. report a high-risk human papillomavirus (hr-HPV) prevalence of 39.7% among HIV-positive women attending public health facilities in Mekelle, Ethiopia, and identify several clinical and behavioural predictors using multivariable logistic regression. The study addresses an important evidence gap in a high-burden setting and employs rigorous molecular detection methods. Nevertheless, three methodological considerations affect the interpretation of the findings. First, facility-based recruitment from selected antiretroviral therapy (ART) clinics may limit the generalizability of the reported prevalence estimate and observed associations to the broader population of women living with HIV. Second, several adjusted odds ratios are accompanied by wide confidence intervals, indicating limited precision and warranting cautious interpretation. Finally, the cross-sectional design precludes causal inference regarding the association between oral contraceptive use and hr-HPV infection because residual confounding and reverse causation cannot be excluded. Collectively, these issues suggest that the findings should be interpreted cautiously and highlight the need for prospective population-based studies to validate the reported risk factors before they inform screening or prevention strategies.

Shfare et al. [1] enrolled 390 HIV-positive women consecutively from five conveniently selected ART clinics in Mekelle and reported an hr-HPV DNA prevalence of 39.7% (155/390; 95% CI: 34.9–44.9%). The authors note that this estimate is lower than the pooled prevalence reported for sub-Saharan Africa and interpret this difference as epidemiologically meaningful. However, because participants were recruited from selected ART clinics, the reported prevalence and observed associations may not fully represent the broader population of women living with HIV in northern Ethiopia. Women attending public ART facilities may differ systematically from those receiving care elsewhere or not actively engaged in care. Consequently, the reported prevalence and regression estimates should be interpreted within the context of the sampled population rather than assumed to represent the wider regional burden [2].

Firstly, Three of the independently associated variables reported in Table 1 are accompanied by particularly wide confidence intervals, including ART interruptions (AOR 4.37; 95% CI: 1.06–17.99), oral contraceptive use (AOR 4.47; 95% CI: 1.39–14.28), and genital ulcer disease (AOR 4.09; 95% CI: 1.12–14.59). These wide intervals indicate limited statistical precision. One possible explanation is sparse data within some exposure categories, although the underlying cell counts were not reported and therefore this cannot be confirmed. If sparse-data conditions were present, alternative analytical approaches such as Firth’s penalized likelihood regression could help reduce small-sample bias and improve estimate stability [3]. Accordingly, these associations should be interpreted cautiously until confirmed in larger studies.

Table 1.

Multivariate logistic regression analysis of factors associated with the prevalence of high

Variables	High-risk HPV	DNA result		COR (95% CI)	P value	AOR (95% CI)	P
	Negative	Positive	Total				value
Residence
Urban	201(58.3%)	144(41.7%)	345(100%)	2.21(1.08–4.53)	0.029	1.39(0.44–4.34)	0.57
Rural	34(75.6%)	11(24.4%)	45(100%)	1		1
Marital status
Married	100(73.0%)	37(27.0%)	137(100%)	1		1
Not married	14(29.2%)	34(70.8%)	48(100%)	6.56(3.17–13.59)	< 0.001	1.62(0.49–5.28)	0.42
Divorced	64(53.8%)	55(46.2%)	119(100%)	2.32(1.37–3.91)	0.002	1.56(0.72–3.42)	0.259
Widowed	57(66.3%)	29(33.7%)	86(100%)	1.37(0.76–2.46)	0.286	1.43(0.62–3.31)	0.395
Employment status
Housewife	110(65.5%)	58(34.5%)	168(100%)	1.05(0.51–2.15)	0.884	1.11(0.33–3.73)	0.864
Farmer	14(70.0%)	6(30.0%)	20(100%)	0.85(0.27–2.71)	0.793	1.38(0.21–9.17)	0.733
Merchant	49(53.3%)	43(46.7%)	92(100%)	1.75(0.82–3.75)	0.147	1.10(0.31–3.90)	0.874
Governmental employee	28(66.7%)	14(33.3%)	42(100%)	1		1
Non-governmental employee	11(68.8%)	5(31.3%)	16(100%)	0.90(0.26–3.13)	0.88	1.03(0.19–5.44)	0.968
Daily laborer	20(43.5%)	26(56.5%)	46(100%)	2.60(1.09–6.18)	0.031	3.76(0.96–14.78)	0.057
Self-employed	3(50.0%)	3(50.0%)	6(100%)	2.0(0.35–11.21)	0.431	1.57(0.12–20.12)	0.728
Educational level
No formal education	63(68.5%)	29(31.5%)	92(100%)	1.19(0.58–2.42)	0.63	1.32(0.47–3.70	0.593
Read and write	22(44.9%)	27(55.1%)	49(100%)	3.17(1.43–7.02)	0.004	2.54(0.81–7.95)	0.108
Elementary	72(52.9%)	64(47.1%)	136(100%)	2.30(1.19–4.42)	0.012	1.26(0.52–3.04)	0.608
Secondary school	44(72.1%)	17(27.9%)	61(100%)	1		1
College and above	34(65.4%)	18(34.6%)	52(100%	1.37(0.61–3.04)	0.44	1.92(0.60–6.14)	0.269
Current type of contraceptive use
None	172(68.5%)	79(31.5%)	251(100%)	1		1
Injectable	20(46.5%)	23(53.5%)	43(100%)	2.50(1.30–4.82)	0.006	1.42(0.55–3.6)	0.465
Condom	21(60%)	14(40%)	35(100%)	1.45(0.70 − 3.00)	0.315	0.23(0.06-0.89)*	0.034*
IUCD	6(54.5%)	5(45.5%)	11(100%)	1.81(0.53-6.13)	0.337	1.87(0.34–10.29)	0.468
Oral contraceptive	8(25.8%)	23(74.2%)	31(100%)	6.25(2.68–14.60)	< 0.001	4.47(1.39–14.28)*	0.012*
Implant	8(42.1%)	11(57.9%)	19(100%)	2.99(1.15–7.73)	0.024	1.03(0.24–4.33)	0.968
Age at first sexual intercourse
11–18	112(56.6%)	86(43.4%)	198(100%)	2.36(1.16–4.81)	0.059	0.81(0.32–2.06)	0.664
19–24	86(60.1%)	57(39.9%)	143(100%)	2.04(0.98-4.25)	0.017	0.89(0.34–2.28)	0.81
25–35	37(75.5%)	12(24.5%)	49(100%)	1		1
History of STI syndrome
None	192(71.9%)	75(28.1%)	267(100%)	1		1
Vaginal discharge	16(36.4%)	28(63.6%)	44(100%)	4.48(2.29–8.75)	< 0.001	2.61(1.01–6.70)*	0.046*
Genital ulcer disease	6(19.4%)	25(80.6%)	31(100%)	10.66(4.20-27.03)	< 0.001	4.09(1.12–14.59)*	0.030*
PID	5(22.7%)	17(77.3%)	22(100%)	8.70(3.10-24.43)	< 0.001	3.15(0.75–13.13)	0.115
Inguinal Bubo	9(64.3%)	5(35.7%)	14(100%)	1.42(0.46-4.38)	0.54	1.84(0.44–7.60)	0.402
A mix of two or more types	7(58.3%)	5(41.7%)	12(100%)	1.82(0.56-5.94)	0.315	1.70(0.31–9.43)	0.542
Partner’s history of STIs
No	149(73.0%)	55(27.0%)	204(100%)	1		1
Yes	17(34.7%)	32(65.3%)	49(100%)	5.09(2.62–9.91)	< 0.001	2.23(0.79–6.34)	0.129
I do not know	69(50.4%)	68(49.6%)	137(100%)	2.67(1.69–4.21)	< 0.001	1.31(0.687–2.49)	0.412
History of unprotected sex before marriage
Yes	79(42.5%)	107(57.5%)	186(100%)	4.40(2.84–6.80)	< 0.001	2.20(1.14–4.24)*	0.017*
No	156(76.5%)	48(23.5%)	204(100%)	1		1
Lifetime number of sexual partners
One	135(76.3%)	42(23.7%)	177(100%)	1		1
Two and above	100(46.9%)	113(53.1%)	213(100%)	3.63(2.34–5.630	< 0.001	1.41(0.73–2.72)	0.302
Awareness of STI causes
No	55(42.6%)	74(57.4%)	129(100%)	1		1
Knows one cause of STIs	73(64.0%)	41(36.0%)	114(100%)	0.41(0.24-0.70)	0.001	0.53(0.25–1.14)	0.107
Knows two or more causes of STIs	107(72.8%)	40(27.2%)	147(100%)	0.27(0.16-0.46)	< 0.001	0.56(0.27–1.17)	0.125
Awareness about HPV-linked cervical cancer
No	113(49.1%)	117(50.9%)	230(100%)	3.32(2.12–5.19)	< 0.001	2.23(1.18–4.24)*	0.013*
Yes	22(76.3%)	38(23.8%)	160(100%)	1		1
Years since ART initiation
4/12 to 11 years	96(48.0%)	104(53.0%)	200(100%)	2.95(1.93–4.51)	< 0.001	2.05(1.13–3.74)*	0.018*
11&1/12 years to 35 years	139(73.2%)	51(26.8%)	190(100%)	1		1
History of ART interruption events
No	161(73.9%	57(26.1%)	218(100%)	1		1
Once	61(49.2%)	63(50.8%)	124(100%)	2.91(1.83–4.63)	< 0.001	1.17(0.60–2.29)	0.637
Twice	9(36.0%)	16(64.0%)	25(100%)	5.02(2.10-11.99)	< 0.001	0.83(0.20–3.44)	0.803
Three times or more	4(17.4%)	19(82.6%)	23(100%)	13.4(4.37–41.10)	< 0.001	4.37(1.06–17.99)*	0.041*
Recent HIV viral load
Undetectable	225(71.7%)	89(28.3%)	314(100%)	1		1
≥ 40 copies/mL	10(13.2%)	66(86.8%)	76(100%)	16.68(8.21–33.90)	< 0.001	6.89(2.75–17.28)*	< 0.01*

Abbreviations/note: ART: Antiretroviral Therapy; AOR: Adjusted odds ratio; CI: Confidence Interval; COR: Crude odds ratio; IUCD: Intrauterine Contraceptive Device; HIV: Human immunodeficiency virus; HPV: Human papillomavirus; PID: Pelvic Inflammatory Disease; STI: Sexually Transmitted Infections; “1”: represents the reference group; p values indicated by *show statistically significant variables (P < 0.05). Hosmer and Lemeshow Test (P = 0.495); multicollinearity assessment (Pearson correlation coefficient (r) < 0.3; Variance Inflation Factor (VIF) < 2 (Minimum = 1.106; Maximum = 1.426)

Secondly, the association between oral contraceptive use and hr-HPV infection deserves particular consideration because several plausible explanations exist. The biological mechanisms discussed by the authors represent one possible interpretation; however, the cross-sectional design precludes determining whether oral contraceptive use itself contributes to hr-HPV infection or whether the association reflects residual confounding, behavioural differences, or reverse causation. Consequently, the observed association should not be interpreted as evidence of a causal relationship but rather as a finding requiring confirmation in prospective longitudinal studies [4].

These methodological considerations do not diminish the strengths of the study. The use of the Abbott m2000 RT-PCR platform to detect all 14 WHO-recognized high-risk HPV genotypes represents an important methodological advantage over screening approaches relying solely on cytology or visual inspection with acetic acid. The reported predominance of non-16/18 high-risk HPV genotypes may have implications for future HPV vaccination strategies in Ethiopia. However, vaccine effectiveness and policy recommendations cannot be inferred from this single cross-sectional investigation and require confirmation through broader epidemiological studies [4].

Future research would benefit from prospective cohort studies recruiting women from both health facilities and community settings while incorporating detailed measures of sexual behaviour, ART adherence, viral suppression, and immunological status. Longitudinal follow-up would allow investigators to distinguish incident from prevalent hr-HPV infections and better evaluate temporal relationships between potential risk factors and infection. Given the strong observed association between viral load and hr-HPV infection, viral load may represent an important candidate variable for future risk-stratified screening research; however, this hypothesis requires validation before informing clinical practice or screening policy.

Abbreviations

AOR

Adjusted odds ratio

ART

Antiretroviral therapy

CI

Confidence interval

hr-HPV

high-risk human papillomavirus

HIV

Human immunodeficiency virus

OC

Oral contraceptive

RT-PCR

Real-time polymerase chain reaction

VIA

Visual inspection with acetic acid

Author contributions

S .T. conceived, drafted, critically revised, and approved the final version of this Matters Arising article.

Funding

No funding was received for this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable. This Matters Arising article is a methodological commentary on a published study and did not involve collection of any human or animal data.

Consent for publication

Not applicable.

Authors'' information

Shweta Thusoo is an Assistant Professor at Dr. D.Y. Patil School of Allied Health Sciences, Dr. D.Y. Patil Vidyapeeth (Deemed to be University), Pune, India, with research interests in reproductive health, infectious diseases, and academic medical writing.

Competing interests

The authors declare no competing interests.