| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
 |
ABSTRACT |
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Recent results from two studies within areas of intense but highly seasonal malaria transmission have described the impact of long-term use of ITNs/ITCs on all-cause child mortality.17,18 Neither showed any evidence of increased mortality in older children following sustained use of ITNs and ITCs for up to seven-and-a-half years.
Due to ethical concerns regarding withholding a proven life-saving intervention, none of these studies were able to compare mortality rates between sustained ITN users and non-users. Rather, they measured mortality rates during times when both intervention and control groups were under widespread ITN/ITC coverage, comparing children who had used ITNs/ITCs for differing lengths of time. However, it is possible that a general decreasing trend in all-cause child mortality during these studies within the general population in the area, with or without ITNs/ITCs, could have occurred. In such a situation, a rebound in mortality in the children who had used ITNs from birth might have been masked.
We recently presented results from an area of intense, perennial transmission in western Kenya and failed to find any change in the mortality rate of children 12–59 old months 4–6 years after introduction of ITNs.19 Here we present additional data and analyses from the western Kenya ITN trial. These data extend the previous analysis in several important ways. First, although these results primarily focus on the four-year ITN study (1997–2001) during which children were under continuous demographic monitoring, we include here an additional year of follow-up from 2002 with a health and demographic surveillance system (HDSS). Second, in addition to the comparison of age specific all-cause mortality rates, we present an analysis of the proportion of deaths occurring at older ages. Third, to address the issue of not having a contemporaneous control group of children not under ITNs, mortality rates and proportional mortality are compared historically to the original control group without ITNs. In doing so, an attempt is made to control for potential confounders of seasonality and underlying changes in all-cause child mortality across calendar years. Finally, we present data on all-cause child mortality from areas contiguous to the study sites without widespread ITN use to assess the overall trends in child mortality in the area over the course of the study.
|
METHODS |
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Prior to the start of the study, malaria transmission was intense and perennial3,22,23 and ITN use was low (< 5%).24 Historically, all-cause child mortality has been extremely high in this region, since neonatal, infant, and cumulative mortality under five years old was estimated to be 32, 176, and 257 per 1,000 live births, respectively.25 Human immunodeficiency virus/acquired immunodeficiency syndrome is a growing threat to child survival, directly and indirectly, in this area.
Study design and intervention. The ITN study was split into two distinct but continuous study phases: the initial community-randomized controlled trial (henceforth referred to as phase 1), followed by a two-year extended evaluation (henceforth referred to as phase 2). Details concerning rationale, study design, ITN allocation, and timeline have been presented elsewhere for phase 19,26 and phase 2.19 Briefly, half of the 79 and 142 villages within Asembo and Gem, respectively, were randomized via public lottery to receive ITNs at the start of phase 1, which began in Asembo in 1997 and in Gem in 1998. A total of 45,667 ITNs were distributed to 60,000 people in phase 1. All remaining households within control villages (approximately 70,000 people) were allocated 50,974 ITNs two years later at the start of phase 2, which concluded in 2001 in Asembo and 2002 in Gem. The ITNs were pre-treated before distribution with permethrin (Peripel, Hoechst Schering, AgrEvo, Frankfurt, Germany) and then re-treated every 6–11 months during both phases 1 and 2 and in 2002 to a target dose of 500 mg/meter2 of netting.
Estimates of ITN coverage remained at approximately 1.5 people per net over both study phases within villages allocated ITNs. The ITN adherence (percentage sleeping under a properly deployed net) among children less than five years old within villages originally allocated ITNs increased from 66% in phase 1 to 83% in phase 2.19,24 The ITN adherence during phase 2 among children less than five years old within former control villages was 77%.
Data collection. Continuous monitoring of child-years at risk and deaths was performed during both study phases with biannual population censuses of the study population, as described in detail elsewhere.9,26 In summary, all children within the study area were enumerated via a baseline census conducted at the beginning of phase 1. Thereafter, biannual updates were conducted by project-trained traditional birth attendants who visited all households within their villages using pre-printed roll-call forms to record all births, deaths, and migrations.
Following the completion of the four-year ITN study, a new baseline survey was conducted to establish a new HDSS in the study area.19 Due to incompatibilities in database structure between the demographic surveillance under the ITN study and the new HDSS, linking records and subsequent person-time was not possible.
Data from government burial records of children less than 10 years old who died between 1997 and 2002 in divisions contiguous to the study area were collected from divisional registrar offices to ascertain trends in child mortality in areas without widespread ITN use, as described elsewhere.27 In these areas, age-specific mortality rates for each month and year of the study were estimated from reported deaths and the population figures from the 1999 Kenya National Census, and extrapolated to other time points using the reported annual growth rate of 0.23%. The number of deaths reported from governmental burial records was incomplete for two of the months during the study (3%). The corresponding monthly mortality rates were inferred from the mean of monthly child mortality rates for these same months over the previous five years.
Data management and analysis. Data collection forms were checked at central field offices prior to data entry at the Centers for Disease Control and Prevention/Kenya Medical Research Institute field station using Visual FoxPro version 6.0 (Microsoft, Redmond, WA). SAS version 8.01 (SAS Institute, Cary, NC) was used for all data cleaning and analyses. Data cleaning was performed using logic checks, with subsequent error listings verified against the original data forms.
The eligibility criteria used to standardize deaths and person-time for assessing child mortality were consistent between phase 1 and phase 2.9 A counting procedure was used to account for study subjects moving in and out of the survival risk set. The analysis was restricted to children 28–1,824 days old from birth (1–59 months) who resided in the study area for at least one month during the study period. Twenty-nine (< 1%) records with one or more conflicting critical variables that could not be resolved were excluded from analysis.
Person-time was compared using the following four groups to assess the impact of sustained ITN use on all-cause child mortality during the four-year ITN study: control villages during phase 1, henceforth referred to as the baseline (no ITNs during the first two years of the study); intervention villages during phase 1, henceforth referred to as ITN group 1 (under ITNs 0–2 years); former control villages during phase 2, henceforth referred to as ITN group 2 (under ITNs 0–2 years during phase 2 only); and former intervention villages during phase 2 henceforth referred to as ITN group 3 (under ITNs 2–4 years during phase 2). Thus, the baseline group serves as a contemporaneous control to ITN group 1 and as a historical control to ITN groups 2 and 3.
Post-neonatal infant mortality is defined as the rate of death in infants 28–364 days old (henceforth referred to as 1–11 months) per 1,000 person-years, while post-infant mortality is defined as the rate of death in children 365–1,824 days old (henceforth referred to as 12–59 months) per 1,000 person-years. Child mortality is defined as the rate of death in children 1–59 months old.
The effect of ITNs on all-cause child mortality was assessed on an intention-to-treat basis for all analyses. A two-sided P value < 0.05 was considered statistically significant for all analyses. Results of the survival analyses described below are presented as adjusted hazard ratios (HRs). Protective efficacy (PE) is defined as the percent reduction in all-cause mortality between comparison groups, calculated as (1 – the adjusted HR) x 100. Because randomization of treatment status was done at the village level, 95% confidence intervals (CIs) were derived for all survival models using a robust sandwich estimator to correct for the effect of clustering.
Survival analysis for historical comparison. Age-specific mortality rates between ITN group 3 and the baseline were compared to determine whether mortality rates had increased in children as a result of sustained ITN use of 2–4 years. ITN groups 1 and 2 are also compared with the baseline to assess the impact of ITN use for up to two years. An individual-level survival analysis was used to fit a multivariate Cox Proportional Hazard model to the data (SAS Phreg Procedure) to control for time-dependent covariates. All person-time from Asembo and Gem was pooled for this analysis because the effect of ITNs did not differ significantly between study sites. The analysis was stratified a priori by 1–11- and 12–59-month age groups given the findings from phase 1 that showed ITNs to have a significant impact on infant mortality only.9 Subsequent analyses here showed the effect of ITNs did not differ by age within each of the 1–11- and 12–59-month intervals. Dummy coding was used to establish a single model for comparing the three ITN groups to the baseline. Rainfall lagged by 60 days was included in this model to control for seasonality across calendar years. While the effect of delayed net re-treatment beyond six months was a significant factor on ITN efficacy in phase 1,9 it was not in phase 2 and was subsequently dropped from the final model. To control for the effect of underlying changes in child mortality in this area over the course of the study, reported monthly mortality rates from the contiguous areas without ITNs were included in the model. Monthly mortality rates estimated from contiguous areas were not associated with lagged rainfall, and thus both were eligible for inclusion in the final model. The final survival model included sex, plus time-dependent covariates for age, total monthly rainfall, and monthly child mortality rates from contiguous areas. Because age in months was not a significant confounder within the model comparing children 1–11 months old, it was excluded from this analysis.
Survival analysis for birth cohorts. To further evaluate whether protection by ITNs throughout infancy was associated with increased mortality rates at an older age, we compared two cohorts of children: those born after their villages received ITNs but at least one year prior to the start of phase 2, and those born into control villages during the same time period without ITNs. The two groups differed only according to their ITN use during the first 12 months of life. Since the ITN study lasted just over four years, children within these cohorts during phase 2 were between 12 and 51 months of age, which is the age group in which a rebound in mortality would be expected following decreased malaria transmission in infancy. A multivariate Cox Proportional Hazard model was used to compare mortality rates in these two cohorts during phase 2 (SAS Phreg Procedure), adjusted by sex, plus time-dependent covariates for age, and total monthly rainfall. Since this was a contemporaneous comparison, child mortality from contiguous areas was not a significant confounder and was thus excluded. All person-time from Asembo and Gem was pooled for this analysis because the effect of ITNs did not differ significantly between study sites.
Analysis of proportional mortality. To test if sustained ITN use resulted in an upward shift in the age distribution of all-cause child mortality, as would be expected if there was a rebound in mortality at older ages, the proportion of child deaths at 12–59 months of age of all deaths at 1–59 months of age were compared between each of the three ITN groups and the baseline. Furthermore, the proportion of child deaths at 12–59 months of age of all deaths at 1–59 months of age among former intervention villages in 2002 are also compared with the baseline, providing up to six years of follow-up for detecting an upward shift in mortality following sustained ITN use. To ensure that the age distribution did not change over the period of study, the proportion of living children 1–11 months old and 12–59 months old was compared between each of the three ITN groups, the former intervention villages in 2002 and the baseline. Logistic regression (SAS Proc Genmod) was used to compare the odds of dying at 12–59 months of age between the baseline and the four comparison groups outlined above, adjusted by study site and rainfall. Empirically estimated standard errors were derived using generalized estimating equation methods to account for the effect of clustering at the village level.
Ethical clearance. Informed consent to participate in the study was received from all compound heads for both phases of the ITN study and the new HDSS. This research was approved by the institutional review boards of the Kenya Medical Research Institute (Nairobi, Kenya) and by the Centers for Disease Control and Prevention (Atlanta, GA).
|
RESULTS |
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
). Mortality rates after the first year of life were very similar, regardless of the length of time ITNs had been used. Peaks in monthly child mortality (1–59 months) followed seasonal peaks in total monthly rainfall throughout the study, regardless of ITN status (Figure 2).
|
|
).
The efficacy of ITNs for reducing all-cause child mortality was concentrated among post-neonatal infants (1–11 months) throughout the study (Table 1). After controlling for sex, rainfall, and underlying changes in child mortality in the area, ITN use of 0–2 years was associated with significant reductions in all-cause mortality among post-neonatal infants, compared with the baseline (PEs = 22%, 95% CI = 10–33% and 24%, 95% CI = 11–35%) for ITN groups 1 and 2, respectively. The mortality rates presented here for phase I differ slightly from those previously reported by Phillip-Howards and others9 due to the inclusion of additional follow-up time accrued since the end-point used in their analysis. Significant reductions in post-neonatal infant mortality rates were maintained among villages of ITN group 3 who had been using ITNs for 2–4 years, compared with the baseline (PE = 18%, 95% CI = 4–30%). Moreover, ITNs remained efficacious in reducing all-cause mortality among children 1–59 months old over all four years of the ITN study. There was no significant difference between post-infant (12–59 months) mortality rates among children residing in villages with ITN coverage of 2–4 years (ITN group 3), compared with the baseline of villages without ITNs, after adjusting for age, sex, rainfall, and underlying changes in child mortality in the area (HR = 0.91, 95% CI = 0.76–1.07).
|
Deaths of children 12–59 months old represented roughly half of all deaths 1–59 months old, regardless of the length of time ITNs had been used (Figure 3). There were no increased odds of dying between 12 and 59 months of age among children within villages with ITN coverage of 2–4 years (ITN group 3), compared with the baseline, after controlling for study site and rainfall (adjusted odds ratio = 1.03, P = 0.733). The proportions of living children 12–59 months old of all children 1–59 months old at the start of the baseline and intervention villages in phase 2 (ITN group 3) were nearly identical at 76.5% and 77.6%, respectively. Furthermore, there continued to be no significant increase in the odds of dying among children 12–59 months old in 2002 within villages with ITNs for 5–6 years, after adjusting for study site and rainfall (adjusted odds ratio = 1.03, P = 0.834). Older children comprised a larger proportion (80.3%) of all living children 1–59 months old in 2002, compared with the baseline (76.5%).
|
|
DISCUSSION |
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This study confirms that sustained ITN coverage does not increase the risk of all-cause mortality in children 12–59 months of age in this area of intense, year-round transmission. These results extend and strengthen our previous analysis.19 First, we detected no difference in all-cause mortality of older children (12–59 months old) between those who had used ITNs for 2–4 years and those from villages without ITNs at baseline, after adjusting for seasonality as well as changes in the underlying trends in child mortality in the area. Second, the proportion of deaths among children 12–59 months old of all child deaths was nearly identical within villages that had ITNs for up to six years, compared with villages without ITNs, a finding consistent with no upward shift in the age distribution of child mortality. This was observed even though older children comprised a larger proportion of all living children 1–59 months old in 2002, compared with the baseline. Finally, all-cause post-infant mortality rates (12–51 months old), the age range in which one would expect to find any potential rebound, did not differ between the cohort of children who had used ITNs from birth and the cohort without ITNs that was exposed to natural levels of malaria transmission over their first year of life, even after six years of follow-up.19
The benefit of ITNs for preventing all-cause mortality was concentrated among post-neonatal infants and remained significant throughout the four-year ITN study. Based on the number of children saved per 1,000 protected during the four-year ITN study, we estimate that ITNs saved the lives of approximately 345 post-neonatal infants. Moreover, these children experienced no increase in mortality rates after reaching their second year of life.
Due to the proven benefits of ITNs in our initial study, it was inappropriate to maintain a contemporaneous control group of children living without ITNs during phase 2. We were therefore only able to make a historical comparison between children during phase 2 and the baseline of children without ITNs in phase 1. This comparison could have been biased if mortality rates decreased for other reasons during phase 2. However, results from governmental burial records from areas contiguous to the study sites without widespread ITN use showed that all-cause child mortality generally increased during the four-year ITN study (1997–2002). We were able to control for this effect with the inclusion of monthly child mortality rates from these contiguous areas in our main survival model that compared follow-up time from phase 2 historically to the baseline. Adjusting for the underlying changes in child mortality was crucial in controlling for the effect of outside influences on child mortality over the course of the four year study.
There are several possible explanations why such a dramatic and sustained reduction in malaria transmission did not result in an increase in mortality in older children. First, reductions in malaria transmission have been shown to increase the age of first infection within this population.28 Although this might delay development of anti-malarial immunity, illnesses occurring at older ages might be more likely diagnosed, treated, and challenged by a more mature immune response. Second, data suggest that significant protection against severe malaria may be acquired after only one or two infections.29 Cross-sectional studies conducted during phase 2 of the ITN study showed that while infants experienced a sustained decrease in malaria-related morbidity, up to one-third remained parasitemic.19 Thus, ITNs clearly do not prevent all malaria transmission and infants using them may still acquire at least some level of immunologic protection against severe malaria. Lastly, immunologic findings from this population during phase 1 have suggested lowered malaria transmission due to widespread ITN use may not necessarily compromise the acquisition of humoral immunity to malaria in young children, as a result of more efficient antibody responses.30 While further clarification of this evidence is required from additional research, it suggests young children protected by ITNs may still garner significant immunologic protection from severe malaria.
These findings support results from a seven-and-a-half-year follow-up of an ITN study in Ghana18 and a six-year follow-up of an ITC study in Burkina Faso,17 which demonstrated that older children are not at an increased risk of death following sustained protection from malaria as a result of ITN/ITC use since early in life. Historically, our findings are consistent with those found from the study of the Pare-Taveta Malaria Scheme of indoor-residual household spraying in the late 1950s in an area of intense malaria transmission on the Kenya-Tanzania border. After the program was stopped after four years, no rebound in all-cause mortality above pre-intervention levels was detected among children 1–4 years old who had previously been protected by the intervention, presumably since birth.31 Our findings are also consistent with previous research from the Gambia and Tanzania that showed lower malaria transmission pressure during infancy to have limited adverse effects on child mortality at older ages, compared with counterparts within areas of elevated transmission.32,33 Our findings also coincide with those of a recent study from Tanzania that demonstrated the benefit of ITNs in preventing malaria morbidity in young children is not reversed at older ages following 3–4 years of ITN coverage.34
While these results provide no evidence to suggest that mortality increased in older children following sustained ITN use, further investigation of the relationship between the acquisition of clinical malaria immunity during early childhood and transmission intensity is warranted. Monitoring of this study population under conditions of high ITN coverage continues, providing a unique opportunity to assess the long-term implications for malaria in children, as well as young adults, following dramatic reductions in transmission.
Results from this study demonstrate that ITNs continue to save the lives of young children, even with sustained use. Moreover, these results confirm that sustained ITN use is not associated with rebound mortality among older children. The global initiative to promote ITNs for combating early childhood deaths in malarious areas will undoubtedly result in more children surviving into adolescence and adulthood.
Received October 23, 2004. Accepted for publication January 9, 2005.
Acknowledgments: We express our gratitude to the villagers of Asembo and Gem for their participation in this research. George Olang, James Kwach, Michael Onyango, Richard Otieno, and Maurice Ombok are thanked for their field and data management skills. We are grateful to the Centers for Disease Control and Prevention/Kenya Medical Research Institute administrative team for their support. The Director of the Kenya Medical Research Institute is thanked for his permission to publish this manuscript.
Financial support: This research was supported by the United States Agency for International Development.
Disclaimer: The opinions or assertions contained in this manuscript are the private ones of the authors and are not to be construed as official or reflecting the views of the U.S. Public Health Service or Department of Health and Human Services.
* Address correspondence to Thomas P. Eisele, Department of International Health and Development, Tulane University School of Public Health and Tropical Medicine, 1440 Canal Street, Suite 2200, New Orleans, LA. E-mail: teisele{at}tulane.edu 
Authors’ addresses: Thomas P. Eisele, Department of International Health and Development, Tulane University School of Public Health and Tropical Medicine, 1440 Canal Street, Suite 2200, New Orleans, LA 70112, Telephone: 504-584-3655, E-mail: teisele{at}tulane.edu. Kim A. Lindblade, Kathleen A. Wannemuehler, John E. Gimnig, William A. Hawley, Penny Phillips-Howard, Daniel H. Rosen, and Laurence Slutsker, Division of Parasitic Diseases, Centers for Disease Control and Prevention, Atlanta, GA 30341. Frank Odhiambo and John M. Vulule, Centre for Vector Biology and Control Research, Kenya Medical Research Institute, Kisumu, Kenya. Feiko O. ter Kuile, Liverpool School of Tropical Medicine, Pembroke Place Liverpool L3 5QA, United Kingdom. Bernard L. Nahlen, Roll Back Malaria, World Health Organization, Avenue Appia 20, 1211 Geneva 27, Switzerland.
|
REFERENCES |
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
