Skip to main content

Table 3 Summary of findings for effect of screening on cervical cancer mortality and incidence

From: Screening for cervical cancer: a systematic review and meta-analysis

Outcome Illustrative comparative risksa    
Assumed risk for no screening Number per million Corresponding risk for screening Number per million (95% CI) Relative effect (95% CI) Number of participants (Number of studies) GRADE quality of evidenceb
Cervical cancer mortality (invited to HPV test or cytology versus no screening) RCT; follow-up: 8 years 2,033c 1,330 (964, 1,834)c RR 0.65 (0.47, 0.90)d 97,672 (1e) Moderatef,g,h,i,j
Incidence of stage II+ cervical cancer (invited to HPV test or cytology versus no screening) RCT; follow-up: 8 years 2,604c 1,466 (1,093, 1,966)c rr 0.56 (0.42, 0.75)d 97,672 (1e) Moderatef,g,h,i,j
Incidence of invasive cervical cancer (invited to HVP test or cytology versus no screening) RCT; follow-up: 8 years 3,747c 4,216 (3,401, 5,226)c rr 1.12 (0.91, 1.39)d 97,672 (1e) Moderatef,g,h,i,j
Incidence of invasive cervical cancer (cytology versus no screening) cohort study; follow-up: 3 years 1,596k 609 (368, 1,006)l rr 0.38 (0.23, 0.63) 116,022 (1m) Low g,i,j,n
Exposure to cytology screening (cases: diagnosed with invasive cervical cancer; controls: no cervical cancer); exposure: in previous 3 years to lifetime 4,781 cases and 17,916 controls OR 0.35 (0.30, 0.41) 22,697 (13o) Very low p,q,r,s
  1. a The assumed risk is the median control group risk. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). b High quality: Further research is very unlikely to change our confidence in the estimate of effect; Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate; Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate; Very low quality: We are very uncertain about the estimate. c Rates were adjusted for age by study authors. d Study authors do not provide a hazard ratio for the HPV testing and cytology groups combined versus the control group. Using sample and event data we computed a risk ratio/relative risk. e Sankaranarayanan et al. [20]. f Random sequence generation unclear and allocation concealment not described, however study limitations were not downgraded for these risks/uncertainties. g Single study, therefore inconsistency not applicable. h Directness downgraded due to concerns regarding generalizability of population characteristics (rural women living in a low-income country) and intervention characteristics (one-time opportunistic screening; short duration (3 months) of training received by laboratory technicians responsible for processing and reading the samples) to Canadian context. i The number of events is small (<300, a threshold rule of thumb value for dichotomous outcomes), however considering the specific outcome the evidence is not downgraded. j Insufficient number of studies to assess publication bias [41]. k Twenty cases of cervical cancer diagnosed in women who had been screened in the 0.5- to 5.5-year interval; six of these cases were screen-detected cancers while 14 cases were symptomatic cancers. l Sixty-three cases of cervical cancer diagnosed in women who had been screened in the 0.5- to 5.5-year interval; 37 of these cases were screen-detected cancers while 26 cases were symptomatic cancers. m Herbert et al. [21]. n Newcastle-Ottawa Scale [39] for cohort studies was completed to assess study limitations; eight out of a possible nine stars were awarded. o There are 12 included case–control studies [1518, 2330]. The number of studies appears as 13 because two different data sets from one study [23] were used as separate entries in the meta-analysis. p Newcastle-Ottawa Scale [39] completed to assess study quality. None of the studies satisfied all of the rating criteria. Despite some uncertainties (for example, lack of information on non-response rates in some studies) and limitations (for example, one-third of the studies used hospital controls rather than community controls), the evidence was not downgraded for study limitations. q The CIs overlap and the directness of the effect is consistent across studies (see Figure 2); all studies favor screening and only two of the 13 CIs marginally intersect the line of no difference. However, statistical heterogeneity is high (Chi2 = 50.98, df = 12 (P <0.00001); I2 = 76%). Sensitivity analyses were conducted and moderate heterogeneity was found when testing for differences between studies conducted in generalizable (to Canadian context) countries versus less generalizable countries (Chi2 = 2.27, df = 1, P = 0.13, I2 = 55.95) but minimal to no heterogeneity (I2 0% to 21.6%) was found when other differences were explored (that is, Canada and US versus other countries, screening approaches, recency of exposure, sources of screening history, diagnosis dates, sources of controls). r Directness downgraded due to concerns regarding inclusion of both organized and opportunistic screening approaches; diversity of study locations which included developed and developing countries (Canada, US, Finland, Sweden, Japan, Italy, South Africa, Columbia, Costa Rica, Panama, Mexico); and the related potential for important differences in participants and screening procedures, particularly given that half of the studies looked at screening that occurred more than 20 years ago and all studies looked at screening that occurred more than 10 years ago. s Publication bias was strongly suspected due to asymmetry in the funnel plot and the recognition that the risk of publication bias may be substantial for observational studies, particularly small studies that utilize data from electronic medical records or disease registries [41, 42]. CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development and Evaluation; HPV, human papillomavirus; OR, odds ratio; RCT, randomized controlled trial; rr, relative risk; RR, risk ratio.