Literature review > Issue_5 > Review on Anderson et al. 

Self-collected specimens and second-generation nucleic acid amplification tests (NAATs) have meant that population-level sexually transmitted infection screening can occur untethered to the requirement of having individuals attend a clinical setting. Advances in statistical analysis likewise facilitate the identification of selective screening criteria. This is a worthy goal in an era of constrained resources in which the cost of screening must be weighed against anticipated yield and benefits. Anderson and colleagues set out to determine whether self-reported sexual behavior, lifestyle, and sociodemographic characteristics could be used to limit the number of individuals requiring testing for C. trachomatis (CT) infection in a population-based screening program for young Danish men and women.

Using information obtained from a 37-item questionnaire and transcription-mediated amplification assay (TMA, Gen-Probe) laboratory diagnosis of C. trachomatis, a set of self-reportable screening criteria was developed from a sample of asymptomatic individuals randomly selected from among all 21 to 23 year olds living in Aarhus County, Denmark. Specimens were obtained at home using vaginal flush sampling for women and first void urine sampling for men. Response rates were 29.4% among women (1175 respondents tested) and 20.1% among men (1033 respondents).

Stepwise multivariate logistic regression was used to develop a model on 60% of the data, which was then validated on the remaining 40% of the sample. The area under the receiver operator characteristic curve (AUC) of the validation model was 0.68 (95% CI 0.60-0.75); using any of the model's criteria, 95% of the subjects would have to have been screened in order to identify all infections. Generally, AUCs of 0.7 - 0.8 are considered acceptable [1]. The authors therefore concluded that the predictive value was insufficiently strong to justify using this selective screening approach over universal screening in this population. This finding is perhaps unsurprising given the narrow age range, and likely demographic and socioeconomic homogeneity, of this study population. However, it echoes the negative findings of others who have tried without success to identify cost-efficient population-level selective screening criteria for CT [2].

The authors also showed that previously published CT screening criteria applied to this population performed less well than in the original studies, and less well than the criteria established in their study. However, the authors selectively used only those elements of the previously published criteria for which they had analogous variables in their questionnaire, which may have weakened the predictive power.

Many authors historically have relied on automated selection regression procedures to develop selective screening criteria or clinical prediction rules. It may be preferable to use a best subsets algorithm approach. This can be done, for example, by applying Akaike's information criterion, which utilizes a penalty factor to reduce overfitting of models to the original dataset [3,4]. This will not obviate the need to validate screening rules in other populations, however.

It is a truism that clinical prediction rules tend to perform less well when applied to populations other than the one on which the rule was developed, and thus should be externally validated, ideally assessing for impact on actual clinical practice (though this is rarely done) [5]. Does this justify abandoning the search for the holy grail of efficient screening criteria? Arguably no - as long as we acknowledge that such criteria are population-dependent, as Anderson et al. concluded. This necessitates local, and likely, iterative, determination of screening criteria.

As NAATs become increasingly used and more individuals and their sex partners get treated, transmission patterns - and thus relevant screening criteria - are likely to shift [6,7]. La Montagne and colleagues analyzed CT screening criteria used in the northwest United States for the past 12 years and found that using additional factors would increase screening criteria efficiency by around 25% while identifying over 90% of infections [8]. Gotz and colleagues were able to identify a set of CT screening criteria in a population-based study of 6303 Dutch citizens that had a bootstrap-corrected, though not externally validated, AUC of 0.78. They concluded that universal screening may be best for regions with high CT prevalence, whereas selective screening may be useful in areas of low prevalence [9]. This is likely to hold true in a variety of settings, and should be kept in mind as researchers search for locally relevant criteria that facilitate asymptomatic screening in general populations.

References:

1. Hosmer D LS. Assessing the fit of the model. Applied logistic regression. NY: Wiley & Sons; 1999:135-175.

2. van Valkengoed IG, Morre SA, van den Brule AJ, et al. Low diagnostic accuracy of selective screening criteria for asymptomatic Chlamydia trachomatis infections in the general population. Sex Transm Infect. Oct 2000;76(5):375-380.

3. Ambler G BA, Royston P. Simplifying a prognostic model: a simulation study based on clinical data. Statis Med 2002:3803-3822.

4. All the elements needed to calculate AIC scores can be obtained from most statistical analysis software such as SAS, R, S-PLUS, SPSS, or can be calculated automatically in STATA using an upload.

5. Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A review and suggested modifications of methodological standards. Jama. Feb 12 1997;277(6):488-494.

6. Kohl KS, Markowitz LE, Koumans EH. Developments in the screening for Chlamydia trachomatis: a review. Obstet Gynecol Clin North Am. Dec 2003;30(4):637-658.

7. Marrazzo JM, Celum CL, Hillis SD, Fine D, DeLisle S, Handsfield HH. Performance and cost-effectiveness of selective screening criteria for Chlamydia trachomatis infection in women. Implications for a national Chlamydia control strategy. Sex Transm Dis. Mar 1997;24(3):131-141.

8. La Montagne DS, Patrick LE, Fine DN, Marrazzo JM. Re-evaluating selective screening criteria for Chlamydial infection among women in the U S Pacific Northwest. Sex Transm Dis. May 2004;31(5):283-289.

9. Gotz HM, van Bergen JE, Veldhuijzen IK, Broer J, Hoebe CJ, Richardus JH. A prediction rule for selective screening of Chlamydia trachomatis infection. Sex Transm Infect. Feb 2005;81(1):24-30.

about SDI | newsletters | grants | publications | literature reviews

WHO Home - WHO Search - TDR Home - SDI Home - SDI Contact us
(c) WHO/OMS 2001