Participants and Methods

Study populationFrom January 1989 through August 1990, a total of 9226 pregnant women who were attending 2 antenatal clinics in Kingston, Jamaica, were screened for HTLV-I antibodies, in a study of the risk factors for maternal-child transmission of HTLV-I. This study population has been described in detail elsewhere [20]. In brief, of the 350 women (3.8%) identified as being HTLV-I seropositive, 212 (60.6%) were enrolled in the present study, along with 145 randomly selected, HTLV-I–seronegative women. A total of 181 children born to HTLV-I–seropositive women and 127 children born to HTLV-I–seronegative women attended at least 1 postnatal clinic visit. These 308 children were seen at a clinic every 6 weeks for the first 6 months after birth, then every 3 months up to age 2 years, and then every 6 months up to age 10 years

Twenty-eight children born to HTLV-I–seropositive women became infected with HTLV-I. Blood samples obtained from the 28 mothers at or around the time of delivery, as well as blood samples obtained from their children over the course of follow-up, were the basis of the present analysis. Written, informed consent was obtained from all participating women before their enrollment in the study. The study was approved by the institutional review boards at the University of the West Indies and the US National Cancer Institute

Data and sample collectionNurses interviewed mothers at enrollment concerning weekly household income and at each clinic visit regarding breast-feeding status. A physician who was blinded to the HTLV-I status of the mothers and their children conducted physical examinations at each clinic visit. Eczema was diagnosed if the child had clinical signs of the condition at a minimum of 1 clinic visit. Infantile eczema appeared as an intensely pruritic rash of nonflexural surfaces and of the dorsum of the hands, limbs, and chest of infants and children who were 1 month to 3 years of age. Childhood eczema occurred among children 4–10 years of age; it appeared as a dry, papular, and intensely pruritic rash involving scaly patches that were distributed on the wrists, ankles, antecubital and popliteal fossae extensor surfaces of the limbs, and palmar and plantar surfaces of the hands and feet

At each clinic visit, a physician performed phlebotomy on the children. Mothers had phlebotomy performed at or around the time of delivery. Peripheral blood samples were separated into plasma and lymphocyte samples, which were used for HTLV-I serologic and molecular testing, respectively. Plasma samples were stored at −70°C, and lymphocytes were stored in liquid nitrogen at a central repository, until they were used for testing

Age at HTLV-I infectionThe age of the children at the time of HTLV-I infection was determined based on the detection of HTLV-I, by use of polymerase chain reaction (PCR), in samples obtained during the period of the estimated date of seroconversion, which previously had been determined by Western blot analysis of serially obtained samples [20]. For each child, age at HTLV-I infection was calculated as the midpoint between the date when the first sample with a PCR-positive result was obtained and the date when the last sample with a PCR-negative result was obtained

HTLV-I serologic testingThe HTLV-I antibody titer was determined using the Vironstika HTLV-I/II Microelisa System (Organon Teknika). The reciprocal of the end-point dilution was reported on the basis of 4-fold dilutions. HTLV-I Tax-specific antibody titers were measured using ELISA (Tsukuba Research Laboratories) as described elsewhere [21]

HTLV-I provirus loadThe HTLV-I provirus DNA load was measured in triplicate by use of the Taqman system (PE Applied Biosystems) [14]. DNA was extracted from 1×10⁶ frozen peripheral blood mononuclear cells (PBMCs) by use of the PureGene DNA isolation kit (Gentra Systems). For determination of the provirus load, a total of 10 μL of DNA solution per sample was analyzed in a 96-well format, by use of the ABI Prism 7700 Sequence Detector (Perkin-Elmer Applied Biosystems). The amount of HTLV-I provirus DNA was calculated using the formula [(number of copies of HTLV-I pX)/(number of copies of β-actin/2)] × 10⁵. This assay detects as few as 10 copies/10⁵ PBMCs

Statistical methodsAnalysis of all viral markers in 28 HTLV-I–positive children was based on examination of samples obtained at a total of 106 clinic visits over the course of 10 years of follow-up (median follow-up, 6.2 years). Nine children (32.1%) had 5 clinic visits, 7 children (25.0%) had 4 clinic visits, 9 children (32.1%) had 3 clinic visits, and the remaining 3 children (10.7%) had 2 clinic visits. Analysis of viral markers in the mothers was conducted using samples collected at or near the time of delivery

In cross-sectional analyses, maternal income level (all dollar amounts are expressed as Jamaican dollars) was categorized as low (<$100), medium ($101–$200), or high (>$200), and the age at HTLV-I infection was categorized as 0–12, 12.1–18, or >18 months. Duration of breast-feeding was dichotomized (>12 vs. ⩽12 months). Maternal provirus load and maternal antibody titer were treated as continuous variables. Maternal Tax-specific antibody status was categorized as either positive or negative

The nonparametric Wilcoxon&amp;rank sum test and Kruskal-Wallis test were used to assess differences in the provirus loads and antibody titers of the children, according to the following categories: sex, maternal income level, age at infection, and duration of breast-feeding. Pearson’s χ² test or Fisher’s exact tests were used to compare the proportions of children who had Tax-specific antibody, with regard to these same independent factors. The Spearman&amp;rank correlation coefficient was used to assess the linear associations between nonnormally distributed continuous variables

Generalized estimating equations (GEEs) were used to describe the associations of exposure variables with sequential levels of HTLV-I viral markers, to accommodate interclass correlations of HTLV-I titers and provirus loads within children [22]. When considered as covariates in the longitudinal analyses, the antibody titers and provirus loads of children were dichotomized at the median values obtained at the point of equilibrium. Antibody titer was dichotomized into the following categories: >1:7786 and ⩽1:7786; provirus load was dichotomized into the following categories: >6695 and ⩽6695 copies/100,000 PBMCs. Breast-feeding status was treated as a time-dependent variable. Age at HTLV-I infection and duration of breast-feeding were measured in months and were treated as continuous variables. Maternal provirus loads and antibody titers were treated as continuous variables, and maternal Tax-specific antibody status was considered to be either positive or negative. The locally weighted scatterplot-smoothing technique was used to characterize changes in HTLV-I provirus loads and antibody titers with time [23]. The cumulative probabilities of a positive Tax-specific antibody status were calculated using life-table estimates with 6-month risk periods. Multivariate models were used to compare β-estimates or odds ratios (ORs) for each viral marker, adjusted for all variables for which P<.05 in the respective univariate models. SAS software (version 8.02; SAS Institute) was used to conduct all statistical analyses

Results

Changes in HTLV-I viral markers over timeThe 28 HTLV-I–infected children included 16 boys (57.1%) and 12 girls (42.9%) who were born to women predominantly of a low-income (35.7%) or medium-income (42.9%) stratum. The median age of the children at HTLV-I infection was 13 months (range, 4–27 months), and the median duration of breast-feeding was 17 months (range, 2–52 months)

Figure 1A shows a scatterplot of HTLV-I provirus load measurements, presented according to the number of years from initial infection, with the solid line denoting a moving average of the provirus loads over time. On the basis of a GEE regression model, HTLV-I provirus loads increased by 0.24 log₁₀ units/year, on average, for the first 2 years after infection (P=.0004). Provirus loads plateaued after 2 years, as was evidenced by a slope that was not different from zero (P=.61). At 2 years after infection, the median provirus load in children was 6695 copies/100,000 PBMCs. This load was comparable to the median maternal HTLV-I provirus load noted at the time of delivery (9010 copies/100,000 PBMCs), although the proviral loads in children and their mothers were not significantly correlated (r=0.28; P=.15)

Figure 1B shows a scatterplot of the HTLV-I antibody titers in children, presented according to the number of years from infection. Antibody titers increased by 1.26 log₁₀ units/year from the time of infection to 1 year after infection (P<.0001). Antibody titers plateaued after 1 year, after which time changes in log₁₀ antibody titers were not different from zero (P=.28). For children, the median antibody titer at 1 year after infection was 1:7786, which was moderately correlated with the maternal antibody titer (1:17,370) noted at the time of delivery (r=0.40; P=.04)

For the 27 children who were tested for the presence of Tax-specific antibody, the cumulative probability of positive reactivity was 24% at 6 months after infection, and it increased to 80% at 2 years of follow-up (figure 1C). All children who developed Tax-specific antibody remained antibody positive during subsequent follow-up. In comparison, 81.5% of mothers had Tax-specific antibody. The presence of maternal Tax-specific antibody was associated with the development of Tax-specific antibody in children (P=.06)

Cross-sectional associations between HTLV-I viral markers in childrenAmong children, the antibody titer measured at 1 year after infection was significantly correlated with the provirus load determined 2 years after infection (r=0.65; P=.0002). In addition, compared with children who were Tax-specific antibody negative, children who were Tax-specific antibody positive at 1 year after infection had a higher median antibody titer at 1 year after infection (1:15,243 vs. 1:1961; P=.0004) and a higher median provirus load 2 years after infection (9220 vs. 1530; P=.002). However, none of the children’s viral markers at equilibrium were associated with sex, maternal income, or duration of breast-feeding (table 1)

Predictors of longitudinal HTLV-I marker levelsIn univariate analyses performed using GEEs, the log₁₀ provirus loads in the children increased significantly in association with increasing levels of maternal household income (P_trend=.02), increasing age of the child (P=.008), and cessation of breast-feeding at the time of measurement (P<.0001) (table 2). The provirus loads in children were associated with other viral markers in children, and they increased in association with a high antibody titer (>1:7786; P<.0001) and the presence of Tax-specific antibody (P<.0001) at any time during follow-up. In a multivariate GEE regression analysis that adjusted for variables that were found to be statistically significant in the univariate analysis, the longitudinal provirus load in children increased by 0.25 log₁₀ units among those with high antibody titers during follow-up (P=.03), and it increased by 0.32 log₁₀ units among those with anti-Tax antibody (P<.001). No other covariates were significant independent predictors of longitudinal provirus loads in children

In univariate analyses, longitudinal log₁₀ antibody titers in children increased significantly with increasing maternal income level (P=.005), the age of the child (P_trend=.02), and cessation of breast-feeding (P=.01) (table 3). Antibody titers in children were associated with other viral markers in children, including provirus loads (P=.0002) and the presence of Tax-specific antibody during follow-up (P<.001). Antibody titers in children also increased in association with maternal antibody titers (P=.0004). In a multivariate analysis, antibody titers in children increased by 0.22 log₁₀ units per increase in the maternal income level (P=.05). In addition, antibody titers increased by 0.56 log₁₀ units among children with Tax-specific antibody during follow-up, compared with children without Tax-specific antibody (P=.008). The children’s antibody titers remained associated with maternal antibody titers, and they increased by 0.73 log₁₀ units per log₁₀-unit increase in maternal antibody titer (P=.004)

The presence of Tax-specific antibody in children during follow-up was significantly associated with the age of the child and cessation of breast-feeding (P=.05 and P=.004, respectively) (table 4). In addition, in children, Tax-specific antibody–positive status was also associated with other viral markers in children, including at least 1 occurrence of a high provirus load (>6695 copies/100,000 PBMCs) (P=.0003) and a high antibody titer (P=.0006) during the children’s follow-up. In children, Tax-specific antibody status was also associated with the maternal HTLV-I titer (P=.04). In a multivariate analysis, the likelihood of children having anti-Tax antibody increased 1.03-fold per increasing month of age at the time that blood was drawn (P=.05). In addition, for children with high provirus loads at any time during follow-up, the likelihood of having Tax-specific antibody over the course of follow-up was increased 12.2-fold, compared with children who never had a provirus load >6695 copies/100,000 PBMCs (P=.003)

Sixteen of these children were given a diagnosis of eczema. The median HTLV-I provirus load at equilibrium in children with eczema was 9220 copies/100,000 PBMCs, compared with 2705 copies/100,000 PBMCs in children without eczema (P = .03). A GEE regression analysis of sequentially measured HTLV-I provirus loads stratified according to the eczema status of the children demonstrated that provirus load increased significantly over time among children with eczema (P<.001) but not among children without eczema (P=.90). In children with eczema, the provirus load increased by 0.27 log₁₀ copies for each of the first 2 years after infection (P=.002), and it then plateaued for the remainder of follow-up. Among children without eczema, the increase in the provirus load (0.20 log₁₀ copies/year) for each of the first 2 years after infection was not statistically significant. Exclusion of the 1 child who had infective dermatitis did not alter these results

Discussion

Children who acquire HTLV-I infection early in life have an increased risk of developing infective dermatitis in childhood [11, 24]. The acquisition of HTLV-I infection in childhood is a known risk factor for ATL in adulthood and may also precede the onset of HAM/TSP [19]. These diseases have been associated with elevated provirus loads and antibody titers in adults. However, sequential changes in these markers in children with HTLV-I infection have not been previously described in a longitudinal study

We found that, in children, the HTLV-I antibody titer continued to increase for 1 year after initial infection before reaching an equilibrium level. Similarly, the prevalence of anti-Tax antibody increased up to 1 year after infection and stabilized thereafter. The initial increase in the antibody responses, followed by the occurrence of a plateau, appears to indicate that the humoral immune response to HTLV-I infection in children reaches maturity within 1 year after infection. At equilibrium, the median HTLV-I antibody titer among infected children was 1:7786, which is lower than the median titer (1:17,370) noted for the mothers of these children, although the titers were moderately correlated. However, in another Jamaican study in which adult recipients of HTLV-I–infected blood were prospectively monitored, the HTLV-I titer at equilibrium, measured ∼14 months after infection, was 1:4694, which is comparable to the titer noted among children in the present study [25]

In the present study, we found that the prevalence of anti-Tax antibody was similar between children and their mothers. The longitudinal evaluation of Tax-specific antibody–positive status revealed that most of the children had acquired this antibody by 1 year after infection and that Tax-specific antibody generally appeared after the detection of whole-virus titers. These data are consistent with a prospective study of adult recipients of HTLV-I–infected blood products in Jamaica [25]

The provirus load in children without eczema increased to a median of 2705 copies/100,000 PBMCs at equilibrium, which was significantly lower than the median provirus load noted in children with eczema (9220 copies/100,000 PBMCs). However, children with eczema had a median provirus load similar to the provirus loads noted for the group of mothers at the time of delivery (9010 copies/100,000 PBMCs) and for asymptomatic adults with chronic HTLV-I infection in Jamaica and elsewhere [26, 27]. In contrast, the median HTLV-I provirus load in adults with HAM/TSP, whose samples were tested in the same laboratory, was much higher (22,560 copies/100,000 PBMCs) [28]. Patients with ATL have an even higher provirus load, as measured by other laboratories [15]. It is possible that the HTLV-I provirus load in children with eczema continues to increase beyond 2 years after infection, but that a statistically significant increase was not detected in the present study because of the sparse data points beyond that time point. We previously reported the continuous increase in provirus load that occurred during the 4 years after the development of infection in a child in the present study who had infective dermatitis and evidence of immune dysregulation [24]. A cohort study of children with eczema in a region of endemicity for HTLV-I is needed to assess changes in HTLV-I provirus load over time and the risk of developing the more severe HTLV-I–associated diseases of adulthood

We found that the HTLV-I titers of children were independently associated with the HTLV-I titers of their mothers. To our knowledge, this is the first study to report this association in mothers and children with HTLV-I infection, although maternal antibody titer has been shown to be a major determinant of the neonatal titer in mothers and children with several other viral infections, including varicella-zoster virus and poliovirus infection and rubella [29–31]. This association may suggest a role for shared genetic determinants of host immune response. Associations among all HTLV-I viral markers in children were strong and were comparable to those that have been reported in adults from this population [26]

The kinetics of the antibody titer and provirus load in the 2 years after infection suggest that there is a complex relationship between these 2 viral markers. Studies have suggested that at least 2 distinct factors—viral replication and clonal expansion of HTLV-I–infected cells—may contribute to the size of the provirus load [32, 33]. The former factor is a reverse transcriptase–dependent process that results in increased production of viral antigens and, thus, is likely to be accompanied by higher antibody responses; the latter factor is described as a mitotic process that is not accompanied by increased viral antigens or an increased antibody response. Although these 2 processes can occur simultaneously, either can play a dominant role in varying stages of HTLV-I infection. The prominent increase in both the provirus load and the antibody titer during the first year after infection is consistent with a primary role of viral replication. In contrast, the continuous increase in the provirus load after the antibody response had stabilized could be due to expansion of HTLV-I–infected clones. In one child who had infective dermatitis and very high HTLV-I provirus loads throughout the 2 years of her follow-up, the provirus load correlated with the degree of infected T cell expansion [34]. Our data suggest that clonal expansion of HTLV-I–infected cells, a critical process for the pathogenesis of ATL, may be common in children. Our findings also provide evidence that eczema in HTLV-I–infected children may be a cutaneous marker of high levels of HTLV-I provirus load and, thus, may be harbinger of development of HTLV-I–associated disease in adulthood

In summary, we described the natural history and predictors of HTLV-I viral markers in vertically infected children in a well-defined epidemiologic cohort. Although the number of HTLV-I–positive children in our series was limited, the point estimate and robust confidence intervals obtained from the GEE analysis are valid. The use of longitudinal analysis with repeated measurements of HTLV-I markers has significantly enhanced our ability to detect associations and to conduct a stratified analysis to examine an effect of a diagnosis of eczema on the provirus load. We acknowledge that we used a definition of eczema (clinical evidence of eczema at ⩾1 clinic visit) that was less stringent than that which we used in a previous analysis (clinical evidence of eczema at ⩾3 clinic visits) [35]. However, the clinical definition used in the present study allowed for a more sensitive indicator for exploring this hypothesis. To expand the current observations and to further our understanding of the pathogenesis of HTLV-I in children, the continuous follow-up of these children for signs and symptoms of early onset of disease is warranted. In particular, analysis of the evolution of HTLV-I–infected clones, direct measurement of the cytotoxic T lymphocyte response against Tax, and measurements of viral replication in these children relative to HTLV-I viral markers and disease development are of critical importance