International Epidemiological Association 2003

Printed in Great Britain

International Journal of Epidemiology 2003;32:131136 DOI: 10.1093/ije/dyg014

HEALTH PROMOTION

A cluster randomized trial of a sex education programme in Belize, Central America

Alexandra Lynda Conboy Martiniuk,1,2 Kathleen Steel OConnor3 and Will D King4
Accepted 30 September 2002

Background Concerns about adverse consequences of early childbearing and risk of sexually transmitted diseases (STD) have renewed interest in the sexual behaviour of adolescents in developing countries, where they represent a large proportion of the population and are at highest risk. To date, little is known about the sexual knowledge of adolescents in developing countries. This studys primary objective was to evaluate the effectiveness of a responsible sexuality education programme (RSP) in changing knowledge associated with sex and sexuality; secondary objectives were to evaluate changes in attitudes and behavioural intent. Methods Results A cluster randomized design randomizing high school classes in Belize City. Subjects were 1319 years old. Seven schools in Belize City were selected; 8 classrooms were randomized to the intervention arm and 11 classrooms to the control arm (N = 399). The intervention was associated with two more correct answers on the post-test (difference score was 2.22 points, 95% CI = 0.53, 3.91) after adjusting for gender and previous sexual experience. After controlling for gender and previous sexual experience, the intervention was associated with no change in the attitudes (0.06, 95% CI: 2.89, 2.82) or behavioural intent domains (0.84, 95% CI: 1.12, 2.46).

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

Conclusions Greater changes in knowledge were observed in the intervention group than in the control group following the intervention. Changes were not observed for the attitude or behavioural intent domains. These results and the results of similar studies may be used to further improve sex education programmes as it is imperative that students have access to the information necessary to make informed decisions regarding their sexual health. Keywords Sex education, cluster randomized trial, Belize, HIV/AIDS, STD, adolescents

There are several good reasons to study sex education and its use in the adolescent population. Around the world, and particularly in developing countries, the age of marriage is increasing and the number of adolescents is growing. This places increasing numbers of adolescents at risk for premarital pregnancy and sexually transmitted diseases (STD).1 Premarital pregnancies are more likely to be unplanned, and unplanned pregnancies have been shown to increase the risk of maternal and child
1 Department of Community Health and Epidemiology, Queens University. 2 Current affiliation: Department of Epidemiology and Biostatistics at the

University of Western Ontario, Canada.

3 Department of Community Health and Epidemiology, Queens University

and Kingston, Frontenac, Lennox and Addington Health Unit.

4 Department of Community Health and Epidemiology, Queens University.

mortality, and morbidity.2,3 Later age at marriage also contributes to exposure to a larger number of sexual partners which is associated with higher rates of STD including human immunodeficiency virus (HIV).47 The only known protection against these diseases is abstinence or the latex condom, yet it is estimated that less than 10% of unmarried sexually active adolescents in developing countries are using condoms.8 Relatively little is known about the sexual knowledge and experience of adolescents in developing countries, yet they make up a large proportion of the population in these countries.912 This is the case for Belize where it is estimated that 42% of the population are under the age of 15 and 6% are over the age of 60 years.13 Belize is experiencing increasing problems with STD including HIV/AIDS. The number of people infected with HIV is estimated to be 925 infected individuals per 100 000 people;14

131

132

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

in 1996 the AIDS annual incidence rate was estimated to be 161.7 per 1 000 000.15 Mortality rates in Belize from communicable diseases are third highest in Central and South America, after Ecuador and El Salvador (age- and sex-adjusted rates for 19952000).15 The question remains as to the best way to educate individuals about sex and sexuality in developing countries. Unfortunately, there exist little data to support a decision since methodologically sound evaluations of sex education programmes in schools are rare in developing countries.16 The most recent systematic review of randomized controlled trials (RCT) of adolescent pregnancy prevention programmes calls for future research into sex education programmes developed from suggestions made by young people that emphasize negotiation skills in sexual relationships and communication.17 The Responsible Sexuality Education Program (RSP) is one such programme. The programme is based on Banduras Social Learning Theory and was developed with input by high school and university students. It is felt that programmes based on social learning theory are the most effective in influencing behaviour and do so by using modelling, role playing, and shaping techniques.1820 The programme is a 3-hour scripted responsible sexuality education intervention which provides a framework for adolescents for decision making in relationships and provides unbiased information about sex and sexuality. According to behaviour theory, if the aim of sex education is to improve the safe sexual behaviour of individuals, the first aim must be to improve knowledge, and then foster safer attitudes and behavioural intentions.12,21 Therefore this studys objectives were to evaluate changes in knowledge, attitudes, and behavioural intent following the RSP. This study was conducted to gain a better understanding of the effectiveness of the RSP and also to better understand the role of sex education in Belize high schools.

internal consistency for the questionnaire, Cronbachs alpha was calculated for each of the domains of knowledge, attitudes, and behavioural intent at pre-test. In order to evaluate the RSP, the comparison of interest was the change in knowledge, attitudes, and behavioural intent for the intervention group compared with the control group. Students in the intervention and control arms were compared with respect to age, gender, and previous sexual experience as well as baseline levels of knowledge, attitudes, and behavioural intent.

Statistical analysis
All analyses were conducted for the total sample as well as separately for males and females, since the previous literature has indicated that sex education programmes have different effects by gender.2426 In this study a change in score refers to the average change in score from pre-test to post-test whereas a difference score refers to the difference in change scores between intervention and control groups. In order to control for possible confounding by age, gender, and previous sexual experience, a regression analysis, taking into account random effects using the SAS procedure (PROC MIXED), was used to determine if there were significant differences between the intervention and control groups with respect to knowledge, attitudes, and behavioural intent. PROC MIXED fits random effects models in order to accommodate several sources of variation instead of just one.27 In this study the clustered nature of the data was taken into account by fitting a random effects parameter for classroom.

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

Results
Seven of nine high schools in Belize City were selected for study (Table 1). There were 19 classrooms in total; 11 classrooms were control and eight were intervention. Loss to follow-up was similar in the intervention and the control groups (15% and 14% respectively). One individual from each arm had to be removed because of missing information. The final sample consisted of 399 students, 166 in the intervention group and 233 in the control group. Table 2 outlines baseline characteristics of study subjects in the intervention and control groups. The mean age in the intervention group was 15.6 years and it was 15.5 years in the control group. However, the two groups are not similar with respect to gender and previous sexual experience; this is likely to be due to the clustered nature of the data. Therefore, gender and previous sexual experience could act as confounders and were controlled for in the analysis (Table 3). In order to gain a better understanding of the properties of the measurement tool being used, Cronbachs alphas for each
Table 1 Composition of study population Belize (N) High schools Female students Male students Total students 16 2586 1951 4537 Belize City (n) 9 1996 1357 3353 Study (n) 7 317 151 468

Methods
This studys primary objective was to evaluate the effectiveness of the RSP in changing knowledge associated with sex and sexuality. Secondary objectives were to evaluate changes in attitudes and behavioural intent concerning sex and sexuality. It was hypothesized that students randomized to the intervention would show improved knowledge, as well as safer attitudes and behavioural intent concerning sex and sexuality. A cluster randomized design was used with high school classes (ages 1319) in Belize City serving as a cluster. An a priori sample size requirement was calculated to be 513 students, using 0.05 as the estimated intra-cluster correlation coefficient. This sample size was adjusted for possible loss-tofollow-up and was calculated to provide 80% power, at = 0.05, to detect a change score of 1.8 points in the knowledge domain, a difference estimated to be clinically meaningful.22,23 A sampling frame of all high school classrooms from seven available high schools was obtained and 19 classrooms randomly selected for study (based on classroom N = 27). Allocation into intervention and control arms was achieved by flipping a coin. Data were collected using a 79-item questionnaire. Parts One and Two examined attitudes and behavioural intent rated on a five-point Likert scale. Part Three examined knowledge using 20 true/ false/dont know questions. Information on age, gender, and previous sexual experience was also collected. As a measure of

CLUSTER RANDOMIZED TRIAL OF A SEX EDUCATION PROGRAMME IN BELIZE

133

Table 2 Comparison of intervention and control groups with respect to baseline characteristics Intervention (N = 197) (%) Gender Male Female Age Mean age SD 80 (40.8%) 116 (59.2%) 15.6 1.14 Control (N = 271) (%) 70 (25.9%) 200 (74.1%) 15.5 1.18 73 (28.1%) 122 (47.1%) 64 (24.7%) P 0.001a

Although the sample size for this study was calculated a priori, since only estimates for the intra-cluster correlation coefficient and the variance were available before the study, actual values obtained are presented in Table 6. In the knowledge domain a value of 0.025 was obtained for the intra-cluster correlation coefficient; this falls within the range of expected intra-cluster correlation coefficients.22

0.629b 0.003a

Previous sexual experiencec None 51 (26.4%) Some 67 (34.7%) Sexual intercourse 75 (38.9%)
a 2 test. b T test.

Discussion
This study has examined the impact of the RSP on high school students knowledge, attitudes, and intention to practice safer behaviours regarding sex and sexuality. Greater changes in knowledge were observed in the intervention group than in the control group. Changes were not observed in either the attitude or the behavioural intent domains following the intervention. One weakness of this study includes the questionnaire tool used to assess change. After data collection, Cronbachs alphas were calculated for each of the three domains of knowledge, attitudes, and behavioural intent. The knowledge domain is consistent in the measurement of a knowledge concept with a Cronbachs alpha of 0.70 which reaches the recommended range and also the range of alphas for knowledge domains from other studies.12,18 On the other hand, Cronbach alphas for the other two domains, attitudes and behavioural intent, were low, which may be a reflection of the difficulty of measuring these concepts. Due to this difficulty with measurement, results for attitudes and behavioural intent should be interpreted with caution. Two remaining weaknesses in this study relate to the randomization process. Randomization was performed according to classroom rather than school to provide variability in the assignment of intervention and control according to socio-demographic characteristics of the seven high schools included, even though one previous study has shown that similar results are obtained regardless of which entity is randomized.30 Although control arm contamination within schools is a potential weakness, the effect, if any, is likely to be minimal because interaction between classrooms within schools was limited. The second weakness of the randomization process is the imbalance between the intervention and control arms with respect to gender and previous sexual experience. This imbalance in both gender and previous sexual experience may be explained wholly due to the randomization imbalance by gender-specific classes, as gender and previous sexual experience were associated with one another in the population studied, with

c none = no previous sexual experience, some = some previous sexual

experience (e.g. kissing, petting), sexual intercourse = vaginal or anal intercourse.

Table 3 Association between gender and previous sexual experience at baseline Previous sexual experience None Some Sexual intercourse Total Females (%) 109 (34) 153 (48) 55 (18) 317 (100) Males (%) 32 (21) 35 (23) 84 (56) 151 (100)

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

of the three domains (knowledge, attitudes, and behavioural intent) were calculated at pre-test. The alpha for the knowledge domain was 0.70. This falls within the range recommended by Streiner.28,29 Alphas for the attitude and behavioural intent domains were 0.65 and 0.49 respectively. The results of the main analysis are presented in Table 4. A more detailed table presents the percentage correct by question for both the intervention and control groups for the knowledge domain (Table 5). The comparison of interest was the change in knowledge, attitudes, and behavioural intent for the intervention group compared to that of the control group. For the knowledge domain the intervention was associated with an average of two more correct answers on the post-test (difference score was 2.22 points, 95% CI: 0.53, 3.91) after adjusting for gender and previous sexual experience. The intervention was associated with no change in the attitudes or behavioural intent domains. Both crude and adjusted analyses are presented. The adjusted analyses took into account the influence of previous sexual experience.

Table 4 Difference in knowledge, attitudes and behavioural intent attributable to the intervention Difference scorea (95% CI) Knowledge Crude Total Male Female 2.25 (0.60, 3.90) 2.76 (0.72, 4.80) 2.11 (0.23, 3.99) Adjustedb 2.22 (0.53, 3.91) 2.59 (0.53, 4.65) 1.97 (0.05, 3.89) Attitudes Crude 0.11 (2.70, 3.01) 0.19 (2.83, 3.21) 0.05 (2.60, 2.70) Adjustedb (2.89, 2.82) 0.02 (3.18, 3.14) 0.13 (2.50, 2.76) Behavioural intent Crude 0.52 (1.23, 2.14) 0.51 (1.78, 2.80) 0.55 (0.61, 1.71) Adjustedb 0.84 (1.12, 2.46) 0.99 (1.40, 3.38) 0.73 (0.45, 1.91)

a Difference in change scores (post-test minus pre-test) between the intervention and control arm. b Total scores are adjusted for gender and previous sexual experience. Difference scores stratified by gender adjusted for previous sexual experience only.

134

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

Table 5 Percentage correct on knowledge questions for intervention and control groups Intervention Question (correct answer in brackets) During sexual intercourse, using a condom gives some protection against STDa. (T) Most sexual assaults are committed by strangers. (F) Chlamydia (the most common STD amongst teenagers) does not lead to serious problems. (F) You can have HIV for more than ten years without showing signs of illness. (T) Pulling the penis out during sexual intercourse gives protection against all STD. (F) Many people with STD have no signs of illness. (T) The most effective way to avoid pregnancy and prevent disease is to abstain from sexual intercourse. (T) Many people who have an STD do not know they are infected. (T) Legally, only drugstores can sell condoms. (F) Anyone can get AIDS. (T) You can get the same sexually transmitted disease more than once. (T) You can be infected with HIV for up to 6 months before the virus is detected. (T) Sexually transmitted diseases are rare among teenagers. (F) AIDS can be cured if treated early. (F) When used properly, a condom gives one hundred per cent protection against STD and pregnancy. (F) Vaseline is a good lubricant to use with a condom. (F) Untreated sexually transmitted diseases can lead to both men and women being unable to have children. (T) You can tell a person is infected with HIVb by the way he or she looks. (F)
a Sexually transmitted diseases. b Human immunodeficiency virus.

Control Change score (PostPre) Pre-test 2.0 21.1 8.6 20.5 16.5 11.9 0.2 1.4 2.5 1.8 5.2 11.9 1.7 13.3 8.3 52.3 21.7 15.0 91.5 54.6 43.9 54.2 65.3 39.1 90.0 72.0 61.6 91.9 46.1 58.3 70.1 63.1 66.8 26.2 49.2 64.2 Post-test 85.0 56.0 45.3 54.7 73.5 50.9 88.9 69.7 52.6 90.2 50.0 57.3 67.1 67.5 68.8 26.1 61.1 64.1 Change score (PostPre) 6.5 1.4 1.4 0.5 8.2 11.8 1.1 2.3 9.0 1.7 3.9 1.0 3.0 4.4 2.0 0.1 11.9 0.1

Pretest Post-test 87.2 56.6 48.0 50.0 64.8 42.3 84.7 71.9 66.8 93.4 45.4 52.0 73.0 60.2 59.2 19.4 46.4 59.7 89.2 77.7 56.6 70.5 81.3 54.2 84.9 70.5 69.3 91.6 50.6 63.9 74.7 73.5 67.5 71.7 68.1 74.7

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

Table 6 Smallest detectable difference in knowledge, attitudes, and behavioural intent rc Knowledge Attitudes Behavioural intent 0.025 0.0002 0.0005 Pre-test variance s1 3.800 6.300 5.100 Post-test variance s2 4.900 7.600 5.600 Inflation factora 1.500 1.004 1.010 Detectable differenceb 1.390 1.820 1.401

a Inflation factor F = 1 + (n 1) p. b 80% power to detect the difference score indicated with the average cluster size N = 21, k = 1.4 and N1 (knowledge) = 111, N1 (attitudes) = 167 and N1

(behavioural intent) = 165.

c The intra-cluster correlation coefficient rho.

more males indicating they had engaged in sexual intercourse (56%) than females (18%). Because this study used a cluster randomized design the imbalance of classes amplified the imbalance of subjects. After randomization there were five co-ed, one all male and two all female classrooms on the intervention arm. On the control arm there were four co-ed, one all male and six all female classrooms. Unfortunately information on the constellation of gender-specific classrooms was not available at the time of randomization therefore stratified randomization could not be performed. Stratified randomization would have increased study efficiency by balancing gender, and therefore previous sexual experience, between study arms and prevented

them from becoming potential confounders. Overall validity of study results were not affected, however, since both gender and previous sexual experience were adjusted for in the multivariable analyses. Mean age was not significantly different between the intervention and the control group, and therefore could not be a confounder. It is a concern that the study arms may also be unbalanced on unknown confounders, such as socioeconomic status, ethnicity, substance use, and involvement in trading sex for money. Information on factors that may be of concern in this regard could not be collected. Many studies have evaluated sex education programmes and several reviews, as well as one meta-analysis, have summarized

CLUSTER RANDOMIZED TRIAL OF A SEX EDUCATION PROGRAMME IN BELIZE

135

the literature regarding the effectiveness of pregnancy prevention programmes for adolescents.3036 The most recent review is of 26 published and un-published RCT evaluating adolescent pregnancy prevention programmes (including sex education classes, abstinence programmes, family planning clinics, and community-based programmes).17 In this review the authors conclude that these programmes do not delay sexual intercourse or increase birth control use. However, large heterogeneity of results existed across studies and both positive and negative studies were included in those rated highest for methodological rigour. This heterogeneity may be due to attempts to summarize overall effects from studies that evaluate different educational programmes, in different populations and with variable lengths of follow-up. Most importantly, this systematic review included only RCT in developed countries (North America, Western Europe, Australia, and New Zealand) where the intervention being evaluated was consistently being compared to a conventional intervention.17 Therefore this overall lack of effect observed in the systematic review, as the authors indicate, could be due to the inability of the new intervention to exceed the effect of the conventional intervention and not due to the fact that these programmes have no positive effect on knowledge, attitudes or behaviour. Unfortunately knowledge from a systematic review such as this one does not help in the evaluation of the effectiveness of sex education programmes in developing countries, where there may be no other formal sex education in place such as in segments of Belize. Studies conducted to determine which types of programmes are most effective as well as methods to improve these programmes are still needed. By evaluating programmes in populations that differ by ethnicity, socio-demographics, and other factors, we may be better able to deliver tailored and therefore more effective sex education. Five RCT evaluating sex education programmes have been conducted in developing countries; these were not included in the most recently published systematic review. All five RCT demonstrated an increase in knowledge following the educational programme.16,22,24,25,37 Of the four studies measuring change in attitudes, two showed increases in positive attitudes.24,37 Of the three studies measuring behavioural intent only one showed a positive change.25 Four of these studies were randomized by cluster;16,24,25,37 yet only one took into account the clustered nature of the data upon analysis.37 This study, conducted in the Philippines, found improved knowledge and attitudes about AIDS following the intervention. Only one of the five RCT was conducted in a region that may share some ethnic and sociodemographic similarities with Belize. This study was conducted in Nicaragua, Central America and also found improved knowledge following a sex education programme. However, this was not an evaluation of a school-based programme but instead a programme for adults accrued from their homes.16 Also this study randomized only two clusters to each arm and significant differences existed between the two arms on level of education at baseline. The current evaluation of the RSP expands upon a previous evaluation of the RSP conducted in Canada using a quasiexperimental design and enrolling 64 students.38 This previous study also showed a significant improvement in knowledge after the programme (P 0.01), but only for females. The Canadian study also found the RSP improved attitudes for females (P 0.05) and fostered safer behavioural intent for both males

and females (P 0.05). However, this study was quasiexperimental and did not use appropriate statistical techniques for its clustered design. The most recent systematic review of RCT of adolescent sex education programmes calls for future research into programmes developed from suggestions made by young people that emphasize negotiation skills in sexual relationships and communication, as the RSP does.17 Therefore the current evaluation of the RSP is an important one. It is contributing sequentially to a large previous literature on sex education programmes by providing an evaluation of a programme which emphasizes components believed to be important and by conducting the evaluation in Belize, a developing country where no previous literature exists on effectiveness of sex education programmes in area schools. The RSP received positive feedback from students, teachers, principals, and government officials and increased knowledge after the programme. Results from this study and other similar studies may be used to further improve the RSP and other sex education programmes. Future research in this area should include studies that aim to improve instruments to measure knowledge, attitudes, and behavioural intent. Research should also focus on developing an appropriate causal model for responsible sexuality and to examine within, the relative effects of knowledge, attitudes, and behavioural intent on behaviour change. Inclusion of covariates into the causal model would also supplement our understanding of the role of sex education and may increase our understanding of the effectiveness of sex education for particular sub-groups.

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

References
1 DCruz-Grote D. Prevention of HIV infection in developing countries.

Lancet 1996;348:107174.
2 Mbizvo MT, Bonduelle MM, Chadzuka S et al. Unplanned pregnancies

in Harare: what are the social and sexual determinants? Soc Sci Med 1997;45:93742.
3 Anonymous. Unintended pregnancyNew York, 19881989. MMWR

1991;40:72325.
4 Adimora AA, Schoenbach VJ, Bonas DM et al. Concurrent sexual

partnerships among women in the United States. Epidemiology 2002; 13:32027.

5 Aseffa A, Ishak A, Stevens R et al. Prevalence of HIV, syphilis and

genital chlamydial infection among women in north-west Ethiopia. Epidemiol Infect 1998;120:17177.
6 Stigum H, Magnus P, Harris JR et al. Frequency of sexual partner

change in a Norwegian population, data distribution and covariates. Am J Epidemiol 1997;145:63643.

7 Warszawski J, Meyer L. Gender difference in persistent at-risk sexual

behavior after a diagnosed sexually transmitted disease. Sex Transm Dis 1998;25:43742.
8 Blanc A, Way A. Sexual behavior and contraceptive knowledge and

use among adolescents in developing countries. Stud Fam Plann 1998; 29:10616.
9 Frost JJ, Forrest JD. Understanding the impact of effective teenage

pregnancy prevention programs. Fam Plann Perspec 1995;27: 18895.

10 Hughes J, McCauley AP. Improving the fit: adolescents needs

and future programs for sexual reproductive health in developing countries. Stud Fam Plann 1998;29:23345.
11 Oakley A, Fullerton D, Holland J. Behavioral interventions for

HIV/AIDS prevention. AIDS 1995;9:47986.

136

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

12 National Institutes of Mental Health (NIMH). Methodological over-

25 Klepp K, Ndeki S, Leshabari M et al. AIDS education in Tanzania:

view of a multi-site HIV prevention trial for populations at risk for HIV. AIDS 1997;11(Suppl.1):S1S11.
13 Inter-American Development Bank. Retrieved Feb 26, 2002 from:

Promoting risk reduction among primary school children. Am J Public Health 1997;87:193136.
26 Danielson R, Marcy S, Plunkett A et al. Reproductive health counsel-

http://www.iadb.org/
14 UNAIDS (1997) Retrieved Feb 26, 2002 from: http://www.unaids.org/ 15 World Health Organization and Pan American Health Organization

ling for young men: what does it do? Fam Plann Perspec 1990;22: 11521.
27 Donner A. An empirical study of cluster randomization. Int J Epidemiol

Special Program for Health Analysis. Health situation in the Americas: Basic Indicators (2000) Retrieved Nov 20, 2001 from: http://www. paho.org/sha
16 Pauw J, Ferrie J, Villegas R et al. A controlled HIV/AIDS-related health

1982;11:28386.
28 Streiner DL, Norman RA. Health Measurement Scales: A Practical Guide to

Their Development and Use. New York: Oxford University Press, 1995.
29 Pocock SJ. Clinical Trials: A Practical Approach, London: John Wiley and

education program in Managua, Nicaragua. AIDS 1996;10:53744.

17 DiCenso A, Guyatt G, Willan A, Griffith L. Interventions to reduce

Sons, 1983.
30 Kirby D, Coyle K. School-based programs to reduce sexual risk-taking

unintended pregnancies among adolescents: systematic review of randomised controlled trials. BMJ 2002;324:142030.
18 Kirby D. School based programs to reduce sexual risk-taking behaviours.

behavior. Children and Youth Services Review 1997;19:41536.

31 Franklin C, Grant D, Corcoran J et al. Effectiveness of prevention

J Sch Health 1992;62:55963.

19 Hovell M, Blumberg E, Sipan C et al. Skills training for pregnancy and

programs for adolescent pregnancy: a meta-analysis. J Marriage Fam 1997;59:55167.

32 Kirby D, Short L, Collins J et al. School-based programs to reduce

AIDS prevention in Anglo and Latino youth. J Adolesc Health 1998; 23:13949.
20 Kirby D, Korpi M, Adivi C, Weissman J. An impact evaluation of

sexual risk behaviors: a review of effectiveness. Public Health Rep 1994;109:33960.

33 NHS. Preventing and Reducing the Adverse Effects of Unintended Teenage

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

project SNAPP: an AIDS and pregnancy prevention middle school program. AIDS Educ Prev 1997;9(Suppl.A):4461.
21 Fishbein M, Ajzen I. Belief, Attitude, Intention and Behavior: An Intro-

Pregnancies. York: NHS Centre for Reviews and Dissemination, 1997.

34 Oakley A, Fullerton D, Holland J et al. Sexual health education

duction to Theory and Research. CA, USA: Addison-Wesley Publishing Company, 1975.
22 Mbizvo MT, Kasule J, Gupta V et al. Effects of a randomized health

interventions for young people: a methodological review. BMJ 1995; 310:15862.

35 Visser AP, van Bilsen P. Effectiveness of sex education provided to

adolescents. Patient Educ Couns 1994;23:14760.

36 Guyatt GH, DiCenso A, Farewell V et al. Randomized trials versus

education intervention on aspects of reproductive health knowledge and reported behavior among adolescents in Zimbabwe. Soc Sci Med 1997;44:57377.
23 Eisen M, Zellman GL, McAlister AL. A health belief model-social

observational studies in adolescent pregnancy prevention. J Clin Epidemiol 2000;53:16774.

37 Aplasca M, Siegel D, Mandel J et al. Results of a model AIDS

learning theory approach to adolescents fertility control: findings from a controlled field trial. Health Educ Q 1992;19:24962.
24 Fitzgerald AM, Stanton BF, Terreri N et al. Use of western-based HIV

prevention program for high school students in the Philippines. AIDS 1995;9(Suppl.1):S7S13.
38 Kassirer A, Griffiths J. The effectiveness of the Responsible Sexuality

risk-reduction interventions targeting adolescents in an African setting. J Adolesc Health 1999;25:5261.

Education Program: a brief high school sexual education intervention. J Sex Educ Ther 1997;22:511.

International Epidemiological Association 2003

Printed in Great Britain

International Journal of Epidemiology 2003;32:136137 DOI: 10.1093/ije/dyg042

Commentary: Sex education interventions: increasing knowledge is only a first step

Anna Graham

The aim of sex education interventions should be to reduce the adverse outcomes of sexual intercourse. This includes unplanned pregnancies and sexually transmitted infection rates. The findings of the most recent systematic review of interventions to reduce unintended pregnancies amongst adolescents, using the

Division of Primary Health Care, University of Bristol, Cotham House, Cotham Hill, Bristol BS6 6JL, UK. E-mail: a.graham@bristol.ac.uk

results exclusively from randomized trials in developed countries, are disappointing.1 In the past, reviews of evidence have been critical of the methodologies employed to evaluate the impact of sex education interventions. The findings of observational studies in this area are more likely to have positive findings, and are undoubtedly biased, compared with randomized trials.2 More recently a significant number of such interventions have been evaluated using rigorous trial methodologies.

CLUSTER RANDOMIZED TRIAL OF A SEX EDUCATION PROGRAMME IN BELIZE

137

The most recent trial to be published in the UK (SHARE: Sexual Health And Relationships: Safe, Happy and Responsible) included over 8400 pupils aged 1315 years in 25 secondary schools in east Scotland.3 Questionnaires were completed at baseline and at follow-up 2 years later. The intervention was a new 5-day teacher training programme plus a 20-session pack: 10 sessions were delivered in the third year (at 1314 years) of secondary school and 10 in the fourth year (at 1415 years). The primary outcome for the study was use of condoms at first intercourse. Similar proportions of both the intervention and control groups used condoms at first intercourse with less than 10% of pupils reporting first intercourse without a condom. For all other behavioural outcomes (condom use after first intercourse, oral contraceptive use, and unplanned pregnancy) there were no differences between the groups. However, as with Martiniuks study in Belize, published in this issue of the International Journal of Epidemiology, pupils in the intervention group were more knowledgeable than those in the control group.4 The Belize study was well designed in allowing for the clustered nature of the sample both when calculating the sample size and analysing the data. Publishing the intra-cluster correlation coefficient calculated from their data will be useful for planning future research. However, there were a number of weaknesses with the randomization procedures discussed by the authors in their paper. The imbalance between groups in the number of classrooms could have been overcome by using a block method rather than the simple coin toss employed here.5 There were considerable differences between groups at baseline in terms of gender and sexual experience. These data were not available to the researchers prior to the study starting. It may have been appropriate to allocate classes to intervention and control groups when the results from the pre-test questionnaire were available. At this time an alternative randomization procedure such as stratification or minimization may have reduced the chances of imbalance between groups in the study.6 There is evidence, from cross-sectional surveys in the UK, that when school is the main source of information about sexual matters, early and unprotected sexual intercourse is less likely, compared with when other sources such as friends and the media dominate.7 School-based lessons are now the main source of information about sexual matters for young people in the UK.8 Sex education has not been found to increase sexual activity, an accusation frequently made by some. However, it is only one way of addressing the issue of poor sexual health. So far the evidence for its effectiveness is limited. It is likely that other factors are far more important. A review of the European evidence9 suggested that where preventative interventions work in unison, for example when: contraception is easily accessible; sex education is augmented by open attitudes and a positive approach to the sexual health of young people; and law reforms facilitate good sexual health, the combined effects may be greater than the sum of their individual parts.

The factor with the strongest influence preventing teenage pregnancy is educational opportunity. It is well-educated women who tend to delay childbearing. For women aged 2024 years the longer a woman remains in school the less likely she is to have a child before the age of 20. Adolescents with little schooling are often twice as likely as those with more education to have a baby before their 20th birthday. For example, 46% of young Colombian women with less than 7 years schooling have their first child by the age of 20, compared with 19% of those with more education. The contrast is even greater in Egypt, where 51% of less educated women have their first birth before the age of 20 compared with 9% of better educated women. The link between lack of education and early childbearing is also strong among adolescents in the US. Some 58% of young American women who receive less than a high school education give birth by their 20th birthday, compared with 13% of young women who complete at least 12 years of schooling. The report from the Alan Guttmacher Institute, from which these data came, suggested that a low level of education is not necessary a direct cause of early childbearing, however, the two factors are often characteristic of living in impoverished and rural environments.10 The greatest impact to be made in reducing unplanned pregnancies, and sexually transmitted infections, is to increase the time spent in education by young women worldwide. This intervention is likely to change the role of women in society empowering them to avoid the adverse consequences of sexual activity.

Downloaded from http://ije.oxfordjournals.org/ by guest on July 8, 2012

References
1 DiCenso A, Guyatt GH, Willan A, Griffith L. Interventions to reduce

unintended pregnancies among adolescents: systematic review of randomised controlled trials. BMJ 2002;324:142630.
2 Guyatt GH, DiCenso A, Farewell V, Willan A, Griffith L. Randomised

trials versus observational studies in adolescent pregnancy prevention. J Clin Epidemiol 2000;53:16774.
3 Wight D, Raab G, Henderson M et al. Limits of teacher delivered sex

education: interim behavioural outcomes from randomised trial. BMJ 2002;324:143033.

4 Martiniuk ALC, OConnor KS, King WD. A cluster randomized trial

of a sex education program in Belize, Central America. Int J Epidemiol 2003;32:13136.

5 Schulz KF, Grimes DA. Generation of allocation sequences in random-

ised trials: chance, not choice. Lancet 2002;359:51519.

6 Pocock SJ. Clinical Trials. First Edn. Chichester: John Wiley & Sons,

1984.
7 Wellings K, Wadsworth J, Johnson AM, Field J, Whitaker L, Field B.

Provision of sex education and early sexual experience: the relation examined. BMJ 1995;311:41720.
8 Wellings K, Nanchalal K, Macdowall W et al. Sexual behaviour in

Britain: early heterosexual experience. Lancet 2001;358:184350.

9 Kane R, Wellings K. Reducing the Rate of Teenage Conception. An

International Review of the Evidence. Data from Europe. London: Health Education Authority, 1999.
10 The Alan Guttmacher Institute. Risks and Realities of Early Childbearing

Worldwide. New York, Washington: The Alan Guttmacher Institute, 1996.