Objectives To determine the impact of HIV on child mortality and explore potential risk factors for mortality among HIV-infected and HIV-exposed uninfected children in a longitudinal cohort in rural Uganda.
Methods From July 2002 to March 2010, HIV-infected and HIV-exposed uninfected children aged 6 weeks–13 years were enrolled in an open population-based clinical cohort. Antiretroviral therapy (ART) was introduced in 2005. Clinical and laboratory data were collected every 3 months. Person-years at risk were calculated from time of enrolment until earliest date of ART initiation, death or last visit. Cox regression was used to estimate hazard ratios (HR) for mortality.
Results Eighty-nine (30.2%) HIV-infected and 206 (69.8%) HIV-exposed but uninfected children were enrolled. Twenty-one children died. The mortality rate was six times higher in ART-naive HIV-infected children than in HIV-exposed but uninfected children (HR = 6.4, 95% CI = 2.4–16.6). Among HIV-infected children, mortality was highest in those aged <2 years. Decreasing weight-for-age Z (WAZ) score was the strongest risk factor for mortality among HIV-infected children (HR for unit decrease in WAZ = 2.6, 95% CI = 1.6–4.1). Thirty-five children (aged 7 months–15.6 years; median, 5.4 years) started ART.
Conclusions Mortality among HIV-infected children was highest among those aged <2 years. Intensified efforts to prevent mother-to-child transmission of HIV and ensure early HIV diagnosis and treatment are required to decrease child mortality caused by HIV in rural Africa.
Objectifs: Déterminer l’impact du VIH sur la mortalité infantile et explorer les facteurs de risque potentiels pour la mortalité chez les enfants infectés par le VIH et ceux non infectés mais exposés au VIH, dans une cohorte longitudinale en zone rurale en Ouganda.
Méthodes: De juillet 2002 à mars 2010, des enfants infectés et non infectés mais exposés au VIH, âgés de 6 semaines à 13 ans ont été inscrits dans une cohorte clinique ouverte basée sur la population. L’ART a été introduit en 2005. Les données cliniques et de laboratoire ont été recueillies tous les 3 mois. Les risques en personnes-années ont été calculés à partir du moment de l’inscription jusqu’à la première date d’initiation de l’ART, du décès ou de la dernière visite. La régression de Cox a été utilisée pour estimer les rapports de risque (HR) pour la mortalité.
Résultats: 89 (30,2%) enfants infectés par le VIH et 206 (69,8%) enfants non infectés mais exposés au VIH ont été inscrits. 21 enfants sont décédés. Le taux de mortalitéétait 6 fois plus élevé chez les enfants infectés par le VIH, naïfs pour l’ART que chez les enfants non infectés mais exposés au VIH (RH: 6,4; IC95%: 2,4–16,6). Parmi les enfants infectés par le VIH, la mortalitéétait la plus élevée chez ceux âgés de moins de 2 ans. La diminution du score Z du poids-pour-l’âge (WAZ) était le principal facteur de risque de mortalité chez les enfants infectés par le VIH (RR: par unité de diminution de WAZ = 2,6; IC95%: 1,6 – 4,1). 35 enfants (âgés de 7 mois à 15,6 ans, médiane: 5,4 ans) ont commencé l’ART.
Conclusions: La mortalité chez les enfants infectés par le VIH était plus élevée chez ceux âgés de moins de 2 ans. L’intensification des efforts visant à prévenir la transmission mère-enfant du VIH et d’assurer un diagnostic précoce et le traitement du VIH est nécessaire afin de diminuer la mortalité infantile due au VIH en zone rurale, en Afrique.
Objetivos: Determinar el impacto del VIH en la mortalidad infantil y explorar los potenciales factores de riesgo de mortalidad en niños infectados con VIH y niños no infectados pero expuestos al VIH en una cohorte longitudinal en Uganda rural.
Métodos: Entre Julio del 2002 y Marzo del 2010, se incluyó en una cohorte clínica basada en la comunidad a niños infectados con VIH y niños no infectados pero expuestos al VIH y con edades comprendidas entre las 6 semanas y los 13 años. El TAR se introdujo en el 2005. Se recolectaron datos clínicos y de laboratorio cada 3 meses. Se calcularon las personas-años de riesgo desde el momento de inclusión hasta la fecha de iniciación del TAR, muerte o última visita (la primera en ocurrir). Se utilizó una regresión de Cox para estimar los cocientes de riesgo (CR) de mortalidad.
Resultados: Se incluyeron 89 (30.2%) niños infectados con VIH y 206 (69.8%) expuestos al VIH pero no infectados. 21 niños murieron. La tasa de mortalidad era 6 veces mayor en niños infectados con VIH que no habían recibido TAR que en niños expuestos al VIH pero no infectados (CR 6.4, 95%CI = 2.4–16.6). Entre los niños infectados con VIH, la mortalidad era mayor en los menores de 2 años. La disminución en el Z-score de Peso-por-Edad (PEZ) era el factor de riesgo más importante de mortalidad entre niños infectados con VIH (CR la disminución unitaria de PEZ = 2.6, 95% IC = 1.6–4.1). 35 niños (con edades entre los 7 meses-15.6 años, mediana 5.4 años) comenzaron TAR.
Conclusiones: La mortalidad entre niños infectados con VIH era mayor en aquellos con menos de 2 años. Se requiere de mayores esfuerzos para prevenir la transmisión del VIH de madre a hijo y asegurar un diagnóstico y tratamiento temprano del VIH con el fin de disminuir la mortalidad infantil debida al VIH en África rural.
An estimated 2.5 million children worldwide were living with HIV by the end of 2009 with 370 000 new infections in that year (UNAIDS 2010). The majority of these infections occur in sub-Saharan Africa with more than 90% through mother-to-child transmission (MTCT) (Dabis et al. 2000). In Uganda, the Ministry of Health estimates that 27 300 babies annually are infected with HIV through MTCT. Despite scale-up of prevention of mother-to-child transmission (PMTCT) programmes in Uganda, only 51.6% of HIV-infected mothers are accessing PMTCT interventions.
Paediatric HIV infection is still a leading cause of death among children in developing countries with mortality rates highest during infancy (Little et al. 2007). Early initiation of antiretroviral therapy (ART) in infancy is associated with slower disease progression and reduction in mortality (Violari et al. 2008). However, care of HIV-infected children in resource-limited settings still poses a great challenge given the limited and overburdened healthcare systems, inability to perform early infant diagnosis, poor nutritional status (Thurstans et al. 2008) and frequent exposure to multiple co-infections (Obimbo et al. 2004).
We describe a longitudinal paediatric clinical cohort of HIV-infected and HIV-exposed but uninfected children in rural south-west Uganda and determine the impact of HIV on child mortality. We also explore potential risk factors for mortality among HIV-infected and HIV-exposed but uninfected children.
Materials and methods
The study site is in rural south-western Uganda about 30 km from Masaka town. This paediatric cohort was recruited from a larger general population cohort (GPC) that was established in 1989. The GPC initially comprised 15 villages from 1989 to 1999; another 10 villages were added in 1999/2000. The population is surveyed annually to study the population dynamics and risk factors for HIV infection with the aim of assisting the development of appropriate intervention programmes (Mbulaiteye et al. 2002). HIV-1 prevalence among this population declined from 8.5% in 1990/1991 to 6.2% in 1999/2000, but thereafter rose to 7.7% in 2004/2005 (Shafer et al. 2008).
From July 2002 to March 2010, children aged 6 weeks–13 years of HIV-positive women were enrolled into an open paediatric clinical cohort. Children were identified as potential participants from three sources: (i) Babies of HIV-positive mothers identified from the GPC by birth recorders. (ii) Babies born to HIV-positive women in a project that offered a comprehensive package of PMTCT interventions where single-dose nevirapine was offered to both mother and infant from 2002 until March 2007 (Barigye et al. 2010). Thereafter, HIV-positive mothers not eligible for ART received zidovudine from 28 weeks of gestation, and then zidovudine/lamivudine and single-dose nevirapine from labour until 1 week post-partum while their babies received single-dose nevirapine and zidovudine syrup for 1 week post-partum. (iii) Older HIV-infected children identified through voluntary counselling and testing services (VCT).
Management at enrolment and follow-up
After informed consent, eligible children and their mothers or guardians were invited to attend the study clinic for an enrolment visit. Follow-up visits were scheduled every 3 months. During these visits, children were seen by a trained research nurse who administered a detailed questionnaire on medical symptoms, nutritional observations and family-social conditions and performed a physical examination under the supervision of a paediatrician. Guardians were encouraged to bring their children to the clinic whenever they were unwell. At each visit, diagnostic and therapeutic services were offered. Children who needed inpatient care were referred to collaborating hospitals (the nearest being 16 km from the study clinic), and their transport and medical bills were paid for by the project. Comprehensive counselling services including infant feeding advice were offered to guardians at the study clinic. Immunisation was also offered at the clinic. Deaths were ascertained through the guardian’s direct reports, after tracing clients who had missed their scheduled appointments, or by reports from referral hospitals.
Cotrimoxazole prophylaxis was introduced in April 2003 to all HIV-exposed children. This was discontinued among those confirmed HIV-negative after cessation of breastfeeding. ART was introduced in 2005 for children with WHO stage 3 or 4 irrespective of CD4 counts, CD4% <25% for those aged <1 year, <20% for those aged 1–3 years, <15% for those aged 3–5 years and <200 cells for those aged 5 years and older. Starting in 2009, ART was initiated as soon as a diagnosis was confirmed in all infants <1 year. Other illnesses were managed according to standard protocols.
Five home visitors acted as mobilisers and visited participants at home to inform them of their clinic appointments and traced participants who had missed their appointments. A child was defined as lost to follow-up if not seen for more than 2 years; however, an attempt was always made to trace these children.
Laboratory diagnosis and monitoring
At study entry, children aged <18 months with two positive proviral DNA tests (gp41 and gag-p21) (Yang et al. 1999; Heyndrickx et al. 2000) or two independent enzyme immunoassays (EIA) in parallel test algorithms (Nunn et al. 1993) after the age of 18 months were regarded as HIV-infected. Western blot testing was used to confirm HIV status in case of indeterminate EIA results. For children aged <18 months with an initial negative test, retesting was performed after cessation of breastfeeding. CD4 T-lymphocyte cell counts were taken every 3–6 months for HIV-infected and every 12 months for HIV-uninfected children using the FACSCount method (Becton Dickinson, San Jose, CA, USA). At every 3-monthly visit, a complete blood count, malaria test, urinalysis and stool analysis were performed. Other diagnostic tests depended on the child’s clinical presentation at the routine or interim visit.
Data management and statistical methods
Data were collected using structured questionnaires and checked for completeness. Data were double-entered and verified in Access. Stata 10 (Stata Corporation, College Station, TX, USA) was used for analyses.
HIV exposure status of children was determined by identifying the mother’s serostatus either through the annual GPC serosurvey or the PMTCT programme. Children were classified as HIV-infected if they had a positive DNA PCR at study entry or a positive EIA after cessation of breastfeeding, or as HIV-exposed but uninfected if the mother was HIV-positive at the time of delivery and the child was seronegative.
Z scores for weight-for-age (WAZ), height-for-age (HAZ) and BMI-for-age were calculated using the WHO 2007 growth references charts. Children with HAZ less than −2 were considered stunted, those with WAZ less than −2 were considered underweight, and those with BMI-for-age Z scores less than −2 were considered wasted.
Person-years at risk were calculated from time of enrolment until the earliest date of ART initiation, death or date last seen. Survival was evaluated using Kaplan–Meier curves stratified by HIV status. As age is a strong determinant of mortality, age was used as the time axis, with staggered entry times because children did not enter the cohort at the same age, and Cox proportional hazards regression was used to estimate hazard ratios (HR) for mortality, to control for the effect of age. Clustering within siblings was ignored in the analysis because of the small number of events. Among children not on treatment, mortality rates in the period before ART became available (2002–2004), and after the introduction ART (2005–2010) were compared.
We used Cox regression to examine the relationship between mortality and factors measured at enrolment, and time-varying clinical and nutritional factors during follow-up as recorded at the last quarterly visit. We did not attempt to build a full multivariable model of independent predictors of mortality, because there were few deaths. Instead, we did an exploratory analysis of nutritional and clinical factors associated with mortality, after adjusting for age (as the timescale for the Cox regression) and sex as a priori confounders, in order to identify those factors that were the strongest predictors of death. For risk factors that were continuous variables, we used fractional polynomials to examine the shape of the relationship with the log HR, using a set of defined powers [−2, −1, −0.5, 0.5, 1, 2 and ln(x)] and a maximum of two power terms in the model. The differences in model deviances were compared; the linear model was used if the improvement in fit was not statistically significant at P < 0.05.
For children who did not have CD4 counts at the enrolment visit, CD4 counts measured within 6 months of enrolment were used as baseline. CD4 count data were missing at many visits. To impute the missing information, we used linear interpolation to estimate the missing CD4 counts between any two observed CD4 counts. Results from the analysis of imputed CD4 counts were compared with those from the observed data.
Lastly, we investigated the effects of ART on mortality in a separate analysis restricted to HIV-infected children. Person-years at risk were calculated from enrolment in the cohort until death or date last seen. ART was analysed as a time-updated exposure, with children contributing person-years at risk to the group on ART from the date of initiating treatment, and to the group not on ART in the period up until then.
The study was approved by the Science and Ethics Committee of the Uganda Virus Research Institute and the Uganda National Council of Science and Technology. Parents or guardians gave informed written consent.
Characteristics of participants
From July 2002 to March 2010, 295 children from 221 mothers were enrolled in the cohort; 89 (30.2%) children were HIV-infected, 206 (69.8%) were HIV-exposed uninfected (Table 1). HIV-infected children were significantly older at enrolment than HIV-exposed uninfected: median (IQR) age 3.2 (0.2–8.1) years vs. 0.2 (0.1–0.3) years, respectively. Of the 89 HIV-infected children, only 25 (28.1%) were recruited through the PMTCT programme following early infant diagnosis while the rest through community VCT. HIV-infected children whose mothers had died before enrolment were significantly older than those whose mothers were still alive (median 8.1 vs. 0.3 years, P < 0.001).
Table 1. Characteristics of the study participants at enrolment, by HIV status
HIV-infected (N = 89)
HIV-exposed uninfected (N = 206)
Data given in parenthesis are expressed as percentage.
*Question asked only whether child <2 years. Missing data for one HIV-positive child.
†Missing data on hospital admissions for one HIV-positive and three HIV-exposed children, and on illness for five HIV-positive and three HIV-exposed children.
‡Missing data on death of mother for two HIV-positive children, and death of father for three HIV-positive children.
§Missing height-for-age Z score for two HIV-exposed children, weight-for-age Z score for one HIV-exposed child, and BMI-for-age Z score for three HIV-exposed children.
¶Missing haemoglobin for one HIV-positive and one HIV-exposed children, and CD4 count for seven HIV-positive and 55 HIV-exposed children.
Age group, years
Ever breastfed (children <2 only)*
Immunisation up to date
Hospital admissions in past 12 m†
Illness without admission in past 3 m†
Height-for-age Z score less than −2§
Weight-for-age Z score less than −2§
BMI-for-age Z score less than −2§
CD4 lymphocyte, %¶
Survival among children not on ART
During the follow-up period, there were 11 deaths (12.4%) among ART-naïve HIV-infected children, compared with 8 (3.9%) deaths among HIV-exposed uninfected children (Table 2). Mortality among HIV-infected ART-naive was highest in those <2 years of age, with a crude rate of 11.6/100 person-years (Figure 1). Mortality was over six times higher among HIV-infected than HIV-exposed uninfected overall (HR = 6.4, 95% CI = 2.4–16.6), and nearly eight times higher in children who had enrolled at ≤3 months of age (HR = 7.9, 95% CI = 2.4–26.0). Among HIV-infected ART-naive, the mortality rate in the period 2002–2004 was slightly higher than in the period after ART screening was introduced (5.2 vs. 4.2 per 100 person-years, respectively); however, this difference was not statistically significant (HR = 0.7, 95% CI = 0.2–2.9; Table 2).
Table 2. Participants’ follow-up and outcomes by HIV status
HIV-infected (ART-naïve) (N = 89)
HIV-infected on ART (N = 35)*
HIV-exposed uninfected (N = 206)
ART, antiretroviral therapy.
*Represents a subgroup of the 89 HIV-infected ART-naïve children, during the period after they initiated ART.
†Includes the 35 children started on ART.
‡‘Age at entry’ refers to age at ART initiation for group on ART.
§2002–2004 is period before national roll-out of ART in 2005.
¶From Cox regression with age as timescale, so represents HR controlling for age.
Lost to follow-up
Median (IQR) age at entry (years)‡
Median (IQR) age at exit (years)
Median (IQR) follow-up time (years)
Median (IQR) age at death (years)
Deaths/person-years (rate/100 person-years)
Mortality hazard ratio¶
HIV-infected ART-naive vs. HIV-exposed
HR = 6.36 (95% CI = 2.44–16.59)
HIV-infected ART-naïve pre 2005 vs. 2005–2010
HR = 0.73 (95% CI = 0.19–2.87)
HIV-exposed pre 2005 vs. 2005–2010
HR = 0.95 (95% CI = 0.12–7.78)
Death was certified by a medical officer for 15 children (eight HIV-infected ART-naïve and seven HIV-exposed uninfected) who died in a health facility. Bacterial pneumonia with underlying malnutrition was the commonest cause of death among both HIV-infected and HIV-exposed uninfected. Other causes included diarrhoea, malaria and anaemia in both groups while pnuemocystis jirovecii pneumonia and pulmonary tuberculosis were specific to HIV-infected children.
Factors associated with mortality
Among HIV-infected ART-naïve children, the factor most strongly associated with mortality in the age-adjusted analysis was decreasing WAZ (HR for unit decrease in WAZ = 2.6, 95% CI = 1.6–4.1; Table 3). In addition, decreasing HAZ, BMI-for-age Z score and CD4 lymphocyte percentage during follow-up were associated with mortality, with some evidence of an association with CD4 lymphocytes at enrolment. The results of the fractional polynomial models showed that the linear model adequately described the relationship between the log HR and the continuous variables. Adjusting for sex did not change the HR substantially for any of the exposures of interest, indicating that sex was not a strong confounder of the observed associations. The results based on imputed CD4 counts were similar to those using observed CD4 counts.
Table 3. Factors associated with mortality among HIV-infected and HIV-exposed
HIV-infected not yet on ART
Deaths/person-years (rate/100 person-years)
Age-adjusted HR (95% CI)*
Deaths/person-years (rate/100 person-years)
Age-adjusted HR (95% CI)*
ART, antiretroviral therapy.
*HR from Cox regression with age as timescale, so represent HR controlling for age.
†Immunisation status not known for 52 HIV-positive children (seven who died) and 48 HIV-exposed children (one who died).
‡Missing CD4 count at enrolment for seven HIV-positive (three who died) and 55 HIV-exposed children (one who died). Missing CD4 count during follow-up at 61.6% of visits for HIV-positive children and 74.6% of visits for HIV-negative children.
§HR for one-unit decrease in risk factor, modelled as linear association with continuous covariate.
¶As measured at the last visit attended.
Factors at enrolment
P = 0.26
P = 0.45
P = 0.15
Immunisation up to date†
P = 0.05
CD4 lymphocyte, %‡
P = 0.08
P = 0.03
Time-varying factors during follow-up¶
Hospital admissions since last visit
P = 0.26
P = 0.002
Illness w/o admission since last visit
P = 0.19
P = 0.35
Current height-for-age Z score less than −2
P < 0.001
P < 0.001
Current weight-for-age Z score less than −2
P < 0.001
P < 0.001
Current BMI-for-age Z score less than −2
P = 0.002
P < 0.001
Current haemoglobin, g/dl
P = 0.81
P = 0.69
Current CD4 lymphocyte, % (observed)‡
P = 0.02
P = 0.24
Current CD4 lymphocyte, % (imputed)
P = 0.02
P = 0.29
Among HIV-exposed uninfected children, the strongest predictors of mortality were a hospital admission since the last visit, and decreasing HAZ, WAZ or BMI-for-age Z score (Table 3). There was some evidence of an association with decreasing CD4 lymphocytes at enrolment, but not during follow-up, using either observed or imputed CD4 counts. Similar to HIV-infected children, the linear model adequately fitted the continuous variables, and sex was not a strong confounder of any exposure of interest. Although 55 HIV-exposed children had no CD4 results available at any visit, mortality in this group was similar to mortality among those who had available CD4 results (data not shown).
No children were on ART at the time of study entry; treatment was initiated in 35 (39.3%) children during follow-up. Because ART was only introduced in 2005, and because of the changes in guidelines over time, 19 (54.3%) children initiated ART late, based on 2008 WHO eligibility guidelines. Median age at ART initiation was 5.4 years (IQR = 2.3–10.7 years). At the time of starting ART, 71.4% of children were stunted, 48.6% were underweight, and 5.7% had severe anaemia. Median CD4 lymphocyte percentage at the visit closest to ART initiation was 10.2% (IQR = 8.6–17.5%).
There were two deaths in 68 person-years of follow-up in children on ART. One was attributable to disseminated Kaposi sarcoma resulting from immune reconstitution 2 months after starting treatment while the other died of unknown cause 5 months after starting ART. The crude mortality rate in children on ART was similar to that in children who were not on ART (HR = 0.99, CI = 0.20–4.75).
In this rural cohort in Uganda, we found that mortality was six times higher among HIV-infected than HIV-exposed uninfected children, similar to reports from other studies in sub-Saharan Africa (Bobat et al. 1999; Schim Van der Loeff et al. 2003). Among HIV-infected ART-naïve children, mortality was highest in those <2 years of age, with a crude rate of 12/100 person-years, in line with other studies showing that perinatal HIV infection is associated with rapid disease progression and death during infancy (Newell et al. 2004). Although mortality in HIV-infected ART-naive children was slightly higher in 2002–2004 than after the introduction of ART in 2005, the difference was not statistically significant.
The mortality rate among HIV-exposed uninfected children was 1.34/100 person-years, slightly lower than the overall mortality in children under 5 in the GPC, 1.47/100 person-years (K. Baisley, I. Kasamba, M. Marston, unpublished data). As most HIV-infected children in our cohort were not enrolled at birth, our observed mortality rate is likely to underestimate the true mortality. Among children who enrolled within 3 months of birth, mortality was nearly eight times higher in HIV-infected than HIV-exposed. Causes of death among ART-naive HIV-infected children and HIV-exposed children were similar and are the major causes of childhood mortality in developing countries. As found elsewhere (Walker et al. 2006), malnutrition was associated with mortality among HIV-infected children in our study, stressing the need to integrate nutritional rehabilitation into paediatric HIV care (Van Kooten Niekerk et al. 2005).
Early initiation of ART during infancy is associated with reductions in mortality and HIV disease progression (Faye et al. 2004; Violari et al. 2008). The lack of a significant effect of ART on mortality in our study is likely due to delays in starting ART, the small number of children that have been treated so far and the limited follow-up time among those on ART. The median age at ART initiation in this study was 5.4 years and median CD4 at initiation was 10%, this shows that many of these children were already severely immune suppressed and therefore had a poor prognosis even after starting ART. Most deaths in children on ART occur in the first 3 months of treatment (Zanoni et al. 2011); in our study, the deaths in children on ART occurred 2 and 5 months after ART initiation.
The WHO 2010 guidelines recommend initiating ART in all infants <2 years irrespective of CD4 counts. However, in reality, early ART initiation among infants still poses a great challenge particularly in rural settings given the unavailability of diagnostic PCR testing even in areas where PMTCT is offered. Although use of clinical symptoms to make a presumptive diagnosis of HIV is recommended in the absence of virological testing, this requires trained personnel who are often lacking in rural areas.
Strong points of this study were its population rather than hospital-based nature, the close follow-up of children achieved by reviewing them every 3 months, and that we were able to study mortality among HIV-exposed but uninfected children.
One limitation was the imbalance in age distribution at enrolment by HIV status, with HIV-exposed uninfected being much younger than HIV-infected children for whom majority were recruited following testing later in childhood. As a result, HIV-exposed children were overrepresented at younger ages when the risk of mortality is highest, so we may have underestimated the effect of HIV infection on mortality. Secondly, there were few deaths overall, so we may not have had power to detect some factors associated with mortality.
Our data provide useful information for the assessment of childhood mortality in rural Africa. The mortality among HIV-infected children was very high especially in those <2 years of age, thus emphasizing the need for scaling up PMTCT services, early infant HIV diagnosis and treatment. Malnutrition was an important predictor of mortality among both HIV-infected and HIV-exposed uninfected children. In low-income countries, research to identify interventions to reduce the high burden of malnutrition is still required.
We would like to thank the families and children who participated in the study. We also thank the nursing staff, team of home visitors and laboratory staff who contributed to this work and Mr. Atuhumuza Elly for data management. We acknowledge Dr Henry Barigye for his contribution during the time he supervised this work.