Teenage pregnancy is a critical public health issue in both the developing and developed world. It has been thought to have an intrinsic effect on the infant and maternal morbidity and mortality statistics worldwide. In its publication, the State of the World’s Children report, UNICEF stated that worldwide over 500,00 women of all age groups die yearly and 70000 females aged 15-19 years would die during child birth [1] . Currently, evidence of ‘causal’ hypothesis is conflicting and inconclusive as to whether adverse outcomes are the result of immaturity of the reproductive system or attributable to other socio-demographic characteristics of adolescents .A study demonstrated that majority of pregnant adolescents had no source of income and lacked health insurance [2] .Teenagers were also found to be more likely to be single, less educated and receive or attend insufficient antenatal care when compared with older mothers [3-5] .

Fraser et al conducted a large population-based study which showed that pregnancy in adolescence was associated with an inherently increased risk for obstetric and neonatal outcomes [6] .However, some other studies demonstrated a lack of association attributing the outcome to social factors.

This article aims to review, critically appraise, and synthesise evidence from original publications of observational studies on the relationship between teenage pregnancy and adverse reproductive outcomes. It focuses mainly on prematurity, low birth weight and route of delivery as there are a myriad of adverse birth outcomes-maternal: preeclampsia, anaemia, premature rupture of membranes (PROM), perineal tears, instrumental delivery, caesarean delivery and infant-related complications: prematurity, low birth weight, intrauterine growth restriction, small for gestational age, perinatal morbidity- attributable to teenage pregnancy and there is strict limitation on the article word count.

Methodology: literature search and selection of studies

A literature search on teenage pregnancy and adverse reproductive outcomes of primary studies published in the last 10 years was carried out .Included studies were journal articles published in the English language-this limitation confers some degree of bias to the review. Epidemiological evidence for this review is defined as observational studies- cross-sectional surveys, case-control studies, retrospective cohort studies and prospective cohort studies.

Database searching of Medline (U.S. National Library of Medicine, Bethesda, Maryland) and Embase (Elsevier) was conducted using the following keyword phrases and related terms as search terms – teenage pregnancy or pregnancies, adolescent pregnancy or pregnancies and pregnancy outcome, adverse reproductive outcomes and related terms(refer appendix 1).

The Medline search yielded 110 articles but reviewing the abstracts showed only few of the studies met the inclusion criteria or were readily available online or in print. A similar search conducted on the Embase database yielded less promising results. Additional journal articles were located by reviewing cited references and citation tracking of some of the selected studies. The related article or similar article feature of some journals was used to identify similar studies and their abstracts were reviewed to check if they met the selection criteria. Case studies, case reports, editorials, and reviews were excluded from the search.

Selection criteria

To be included in this review, the selected studies had to meet the following criteria:

-teenage pregnancy is defined as pregnancy in young women under 20 years

-women above 35 years old were either excluded from or treated as an independent category in the study as they are known to have high obstetric risks

-must demonstrate some statistical description and /or analysis of confounding variables in the association between teenage pregnancy and adverse reproductive outcomes

-should have some comparative element in which teenagers are compared with a suitable reference category

-outcome measures include at least two of the following: prematurity preterm delivery, Caesarean section (CS), low birth weight (LBW), infant mortality, neonatal mortality, perinatal mortality, maternal mortality, severe anaemia, preeclampsia and eclampsia

-a significance assessment can be made either by using p-values or confidence intervals

Table 1:Characteristics of selected studies

Authors Publication year(Study period) Setting Study design Sample size Age of teenage subjects(years) Outcome measures of interest Confounding variables considered Other study characteristics Ekwo and Moawad [3] 2000(1989-1995) U.S.A Hospital based retrospective cohort 6,072 3 groups-=15,16-17,18-19

20-24 as reference group Preterm birth, low birth weight Maternal smoking, drug abuse, insurance status, adequacy of prenatal care, median family income, marital status Primaparous black women , singleton pregnancies Bukulmez et al [7] 2000(1990-1998) Turkey Hospital-based matched case-control study 4,470(2,490 cases,3980 controls) Cases:15-19 controls:20-34 stratified during analysis as =17,18-19,20-34 Low birth weight, preterm delivery, pregnancy induced hypertension(preeclampsia, eclampsia),LBW, Antenatal care, gravidity, parity, Singleton pregnancies, subjects matched on marital status, socioeconomic class and ethnicity-white married women of high social status Jolly et al [4] 2000(1988-1997) United Kingdom Hospital based retrospective cohort 341,708 <18,18-34 as reference group Preeclampsia, anaemia, emergency caesarean delivery, preterm delivery, Body mass index, race ,parity , hypertension at booking, pre-existing diabetes, gestational diabetes, smoking Singleton pregnancy Smith and Pell [8] 2001(1992-1998) Scotland Population based retrospective cohort studies 110,233 15-19,20-29 as reference group Stillbirths, neonatal deaths, preterm delivery,

emergency caesarean section, and small for

gestational age Maternal height, maternal weight ,social deprivation Non-smoking mothers, stratified by parity(first and second births) Chen et al [9] 2007(1995-2000) U.S.A Population-based retrospective cohort 3,886,364 <16,17-18,18-19,20-24 as reference category Very preterm delivery, preterm delivery, very LBW,LBW,SGA, Very low APGAR score, low APGAR score Maternal race, age-appropriate educational level, smoking in pregnancy, alcohol use in pregnancy, marital status, prenatal care utilisation Prematurity, very LBW,LBW,IUGR,5min, APGAR score, early neonatal death ,foetal death Igwegwe and Udigwe [10] 20019June 1995-November 2007) Nigeria Hospital based retrospective cohort 58 >20,20-24 Preterm delivery, low birth weight ,preeclampsia, caesarean delivery, APGAR score parity,

occupation, social class, marital status, booking for prenatal care , pregnancy complications, route of delivery Women of the Igbo ethnic group Kongnyuy [5] 2007(November 2004-April 2005) Cameroon Multi-centre cross-sectional questionnaire survey 1,100 = 19,20-29 Low birth weight, prematurity, neonatal deaths, preeclampsia, eclampsia, perineal tears, caesarean delivery Gravidity, frequency of antenatal visits, marital status ,employment status, educational level Primaparas, singleton pregnancy,

Results

Teenage pregnancy and preterm deliveries

Prematurity is said to be the most significant predictor of perinatal mortality [4] . Some of the reviewed studies found an association between teenage pregnancy and prematurity. Jolly et al found that adolescents were about 40 per cent more likely to have very preterm babies i.e. infants less than 32 completed weeks of gestation (OR= 1.41;955 CI=1.02-1.90) [4] .A large hospital-based study conducted using representative data from all 50 states of the U.S and the District of Columbia demonstrated a significantly increased risk of very preterm and preterm infants(infants born at less than 37 completed weeks of gestation) in teenagers when compared with the older women [9] . One cross-sectional evidence also showed the teenage pregnancy was significantly associated with premature births (adjusted OR= 1.77; CI= 1.24-2.52)[5].One study however found that 35ð7% teenagers in the study had preterm birth while the control group had none at all[10]. This may however be due to lack of statistical power to detect any in the control as the small sample size was small, due to inadequately kept hospital records, and there was no adjustment for confounding by other risk factors measured.

Some studies however did not report an increased risk in teenage mothers. One study found no significant difference in preterm births between teenage and older mothers after adjusting for covariates. A reviewed hospital-based study set in Turkey also had similar findings(OR=0.72;95% CI=0.23-2.33;p-value=0.22) [7] . The population-based Scottish study furthermore demonstrated that teenagers having their first babies were significantly at no increased risk for having very preterm infants(OR=1.1; CI=0.9-1.4) [8] .

Teenage pregnancy and birth weight

In one study,31ð4% babies of teenagers and 6ð7% (2) of the older age group had low birth weight (p-value= <0.05) .Another evidence found an independent relationship of teenagers having LBW infants (OR= 1.71; 95%CI=1.15-2.50)[5].A large study demonstrated similar results with adolescent mothers being at a significantly increased risk for both very LBW infants-overall result for 10-19 years group (RR=1.17; 95 % CI=1.14-1.20) and LBW infants [10-15 years (Relative risk(RR)=1.61;95% CI=1,41-1.84),16-17years(RR=1.42;95 % CI=1.35-1.50),18-19(RR=1.17;95% CI=1.13-1.21) and overall for teenage mothers (RR=1.24; 95% CI=1.20-1.27)- when compared with mothers aged 20-29 years[9].

Bukulmez et al did not find an independent association(adjusted OR=2.09;95% CI=0.68-1.64) [7] .The study conducted at a Cameroonian hospital similarly did not demonstrate any significantly increased trend in having LBW infants with decreasing maternal age but mothers in the 16-17 group in the study were more likely to have lighter infants [3] .

Teenage pregnancy and pregnancy induced hypertension

One reviewed evidence found a highly significant increased association between pre-eclampsia (OR=1.99; 95 %CI, 1.24-3.15; p-value=0.004) eclampsia (OR, 3.18; CI, 1.21-8.32; p-value=0.016) [5].The study conducted in Nigeria failed to detect any significantly difference significant in the development of preeclampsia and eclampsia in teenage mothers and older controls. However, another study showed that teenagers were not significantly at risk after adjusting for significant confounders.

Teenage pregnancy and risk of having a caesarean section

Regarding the route of delivery, most studies have shown that there was no significant difference between teenagers and older women[10]. Evidence from the teaching hospital in Nigeria found that there was no statistically significant difference between teenagers and the older age group groups with most teenagers having normal delivery. Caesarean section rates not significantly dif from older women (OR=1.29; 95 % CI=0.80-2.04; p-value=0.295).The Scottish study also reported similar findings showing that teenagers were significantly not at increased risk of having an emergency caesarean section at first delivery adolescents being less likely to have a caesarean (OR= 0.5;CI=0.5-0.6)[8].

Discussion

Study design issues

Cross-sectional studies lack the ability to establish causal relations and also rare outcomes of teenage pregnancy such as intrauterine foetal deaths cannot be studied using this design. In the case-control studies, recall bias pose a validity problem. Longitudinal studies represent a better level of evidence for establishing temporal relationships. The retrospective design of most of the studies allows for data on many outcomes to be readily available and eliminates attrition from loss to follow-up. In reality, the final statistical analysis in these studies is usually cross-sectional using appropriate tests of significance depending on outcome.

However, the prospective design in which carefully planned data can be collected in other to account for all possible confounders of interest in analysis is expensive, time-consuming, and recruiting teenage participants may prove difficult. Ethical consideration of expectation of adverse outcomes is also an issue. Randomised trials, though having a higher level of evidence, would be most suitable in assessing interventions thought to reduce the risk of adverse outcomes such as prenatal care.

Sample size issues and the role of chance

Though most of the studies used a large sample size [4, 9] by increasing their overall statistical power through matching with more controls, the real determinant of an effect is the number of subjects having a particular outcome [7] .A small sample size may yield imprecise results. Only the study carried out in Cameroon included a power calculation in justifying the sample size used [5] .

Data collection issues and the role of chance (random error)

The reviewed cross-sectional study was carried out using questionnaire-based survey with demographic data obtained from obstetric case record and outcome measures recorded after delivery for which the mode of administration was not well explained [5] .The mode of administration, wording of questionnaires and the level of expertise of the interviewer all have an impact of the results. The response rate also affects the validity of the study.

Birth weight measurement using infant weighing scales is also prone to some degree of random error. Sphygmomanometers used in blood pressure measurements also have similar issues.

Definition of study subjects

Most of the reviewed studies considered teenagers as women below 20 years of age; however the UK study [4] defined teenagers as women less than 18 years. A difference in definition across studies makes it difficult to generalise findings.

Selection of study subjects

Selection bias is a particular important issue in longitudinal studies. Most of the studies assessing this hypothesis are retrospective cohort studies and selection of subjects is crucial to validity. A number of studies are hospital-based and unless a well-defined geographical area is served, selection of suitable controls would be a problem and the findings may not be generalisable to the wider population of teenagers.

Definition and measurement of covariates and outcomes

Some of the studies used pre-existing hospital record which reduce the time and expensive required to conduct the study but is highly prone to missing data .Some studies had validated obstetric databases [4] to ensure some degree of data quality while this is uncertain in others. This can introduction selection bias if the data is excluded from the analysis as there may be some systematic difference in outcome and confounders between those included in the final analysis and those accounted for.

The differing birth weight results from the studies might just be inaccurate findings as weighing instruments are highly prone damage from shock and measurement bias. None of the studies described how the birth weight measurements were taken and it was not ascertained if steps were taken to ensure that the infant weighing scales were properly calibrated before use.

A variety of methods was used in assessing the gestational age of infants. One study [3] used either or all of last normal menstrual period, first trimester ultrasound and abdominal assessment for estimating gestational age. Physical examination methods are highly prone to error and may lead to misclassification bias of gestational age in some studies. It is however the most commonly used method in low-resource settings and can miss multiple gestations especially in overweight or obese women. Early ultrasound estimates have however been found to consistently biased towards lower gestational age estimates when compared with dating by last normal menstrual period [11] .

Adequacy of prenatal natal is defined in one study as defined by as more the four clinic visits [5] .This was a subjective definition. Referral filter bias can be a problem in hospital-based studies in which women seen in tertiary or specialist centres may be systematically different from those seen at community health centres and smaller clinics [12] .Studies in which specialist centre records were used are prone to this .

Confounding

Most studies testing the hypothesis between pregnancy in adolescence and its inherent risk for adverse outcomes are longitudinal studies allowing for assessing multiple exposures and multiple outcomes and hence adjusting for a priori and potential confounders. The studies vary greatly in the number and range of confounders considered and some confounders were found to be significant in some studies but not in others.

Bukulmez et al controlled for confounders by matching and restricting the study to white married women of higher social class and stratifying the study population by age in the final analyses but educational status an important potential confounder was not adjusted for [7] .The study conducted in the United Kingdom considered the effect of sexually transmitted infections -bacterial vaginosis specifically- on prematurity [4] but did not control for social class ,an a priori confounder, in the final adjusted model [4] .Complete elimination of the all cofounders is virtually impossible in all studies and results are subject to the effect of residual confounding.

Statistical analyses

Most of the reviewed studies used appropriate analysis methods-Chi square tests, Fischer’s exact test, student’s t-test depending on the outcome of interest and the sample size. Stratification and logistic regression modelling was used for adjusting for confounders but the studies differed in what confounders were just as being most appropriate in the final adjusted model. Studies however differed greatly in their choice of analysis and the reporting of results hence limiting direct comparison.

Conclusion

Evidence linking the biological aspect of teenage pregnancy with adverse birth outcomes is inconsistent and contradictory. This inconsistency in the direction, strength and significance of associations and may be due to the degree of adjustment for confounding [13] ,the sample size and statistical power of the studies, different inclusion and exclusion criteria and to uneven distribution of the socio-demographic characteristics of the studied population. Irrespective of whether the adverse outcomes are the result of reproductive or poor social conditions, teenage pregnancy is a global public health problem at which continued interventions are required. Policy makers worldwide should therefore continue to direct attentions to effective intervention that have impact on reducing teenage pregnancy rates.