Retrospective study on the association between paternal occupational exposure to agrochemicals and childhood leukemia in Michoacán de Ocampo, México

Study design

A retrospective observational cohort study was conducted. Medical records and death certificates from the period 2010 to 2023 were reviewed. All medical records of patients with leukemia under 18 years of age from the oncology department of the Children’s Hospital ‘Eva Sámano de López Mateos’ located in Michoacán de Ocampo, México (Fig. 1), were included. The medical records were analyzed once without subsequent follow-up; this review was conducted from January to April 2024. In Michoacán, oncology care—including the treatment of childhood leukemia—is concentrated in three healthcare centers: the Mexican Social Security Institute (IMSS) and the Institute of Security and Social Services for State Workers (ISSSTE), which together manage between 10% and 15% of childhood leukemia cases in the state; and the Children’s Hospital “Eva Sámano de López Mateos”, which accounts for 85% to 90% of all leukemia cases in Michoacán. It is important to note that there are no specialized hospitals in the various municipalities of the state.

Additionally, death certificates of children (under 18 years of age) whose underlying cause of death (ICD-10) was leukemia and whose care was provided at the Children’s Hospital ‘Eva Sámano de López Mateos’ were included to complete the follow-up of medical records without outcome. These records were obtained from the ’Health Intelligence Center’ (CIS) of the Ministry of Health of Michoacán de Ocampo, México. The CIS collects and manages all mortality data for monitoring, health planning, and evaluating public health policies. These records were reviewed from March to May 2023.

The following data were collected from the medical records and death certificates for each patient and their parents: patient sex, age, occupation (paternal and maternal), education level (paternal and maternal), residential area, type of leukemia, risk, medical course, survival months and/or date of death with the causes of death. Overall survival was defined as the time from diagnosis to death or until the last progress note in the medical record for those patients not listed in the death registry. The causes of death obtained were recorded according to the International Statistical Classification of Diseases and Related Health Problems (ICD-10) by the WHO. Age groups were classified as high risk (under 1 year and from 10 to 17 years) and standard risk (from 1 to 9 years).

This research project was carried out in accordance with the ethical standards of the General Health Law Regulations on Health Research (Cámara de diputados, 2014). According to Article 17 of this regulation, this research is classified as type I. Risk-free research: retrospective documentary studies without intervention in physiological, psychological, and social variables, such as questionnaires, interviews, and review of medical records, without identification of subjects or treatment of sensitive behavioral aspects; therefore, informed consent is not required. A letter of exemption from informed consent, detailing data collection and confidential data management, was submitted to the Ethics Committee of the Children’s Hospital of Morelia “Eva Sámano de López Mateos” (CI/10/2023) along with the research protocol for approval. The protocol was also submitted to the Local Health Research Committee 1602 of the Hospital General Regional No. 1 with register R-2023-1602-060.

This protocol was developed following the guidelines for strengthening the reporting of observational studies in epidemiology (STROBE).

Standardized incidence ratio (SIR)

To assess the incidence of leukemia, the state was stratified into two zones: avocado-growing and non–avocado-growing. The cumulative incidence per 100,000 children (0–17 years) over a 10-year period was calculated for each zone using population census data from 2010 and 2020—the two available censuses within the study period—provided by the National Institute of Statistics and Geography (INEGI). Calculations were based on the number of observed cases and the average child population.

Calculation of expected cases and standardized incidence ratio (SIR). The expected number of cases for each zone and municipality was estimated by multiplying the corresponding average child population (0–17 years) by the state leukemia incidence rate. The state incidence rate was obtained by dividing the total number of cases recorded in the state by the total state population, expressed per 100,000 person-years, with population data obtained from INEGI. Subsequently, the SIR was calculated as the ratio between the observed number of cases and the expected number of cases in each zone or municipality. An SIR value of 1 indicates that the observed incidence matches the expected incidence, values greater than 1 indicate that more cases were observed than expected, and values less than 1 indicate fewer observed cases than expected. To assess the precision of the SIR, 95% confidence intervals (CIs) were calculated under the assumption that events follow a Poisson distribution.

Maps

The delimitation for the avocado agricultural frontier was updated from the one obtained by Morales-Manilla et al. (2011), using information from Climate Rights International (CRI, 2023) certified orchards for export. Additionally, a 2015 land use map (Benítez-Franco, 2021) and a map of avocado orchards in eastern Michoacán (2020) (López-Sánchez, 2024) were used. Very high resolution (ESRI, Google Earth Pro) and high resolution (Sentinel-2, February–March 2024) images were utilized. The maps were created with ArcGIS Pro. Zip codes were downloaded in Shapefile format and a join was performed, defining ranges of leukemia cases and assigning color palettes according to the number of cases.

Based on the delineation of the state’s avocado agricultural frontier, the analyses in this study were conducted by dividing the area into two zones: an avocado-growing zone and a non–avocado-growing zone. Additionally, the maps reflect the division of the state by municipalities and zip codes. In México, a “municipality” refers to the administrative subdivisions within each state, functioning as local government units.

Statistical analysis

A Kolmogorov–Smirnov test was performed to estimate the normality of the distribution of the parameters investigated. The age variable had an abnormal distribution, data were expressed as median (minimum-maximum value). Categorical data were reported as percentages (%), these variables include age, sex, leukemia by cell lineage, paternal and maternal age, paternal and maternal education level, paternal and maternal occupation, survival (primary outcome), deaths, dropouts. Since the study is based on the review of medical records, there are some missing data. For this reason, the sample size (n) of the data used is specified in each table and in the description of the results. A Chi-square test was used to analyze differences between categorical variables. A logistic regression analysis was performed to investigate the relationship between paternal occupation, municipality of residence, and leukemia; the analysis was adjusted for age and sex. Municipalities of residence can act as a confounding variable; therefore, it was included in the analysis to minimize its impact as a source of confusion. Survival was analyzed using Kaplan–Meier curves. The significance of the difference observed between the various curves was determined by the Log Rank test. Additionally, a Cox proportional hazards regression was performed to assess the relative risk among four groups defined by the combination of municipality of residence and paternal occupation. Statistically significant differences were established when P value < 0.05. The data were analyzed in the SPSS version 23 statistical package and the graphs were made in the GraphPad Prism 10 program.

Demographic information

Data were collected from a total of 430 medical records of leukemia and death records from 2010 to 2023 at the “Hospital Infantil Eva Sámano de López Mateos” in Morelia and the “Centro de Inteligencia en Salud”, both located in the city of Morelia, Michoacán, México. The description of the inclusion and exclusion criteria for the medical records used in this study is detailed in Fig. 1. In this study, the medical records were analyzed once without subsequent follow-up. The complete records (up to the last recorded follow-up note) were reviewed, and for those records without an outcome, all children’s death records from the Health Intelligence Center were reviewed. These records had leukemia as the underlying cause of death (ICD-10) and were treated at the ‘Eva Sámano de López Mateos’ Children’s Hospital.

The demographic and clinical characteristics of the pediatric population with leukemia are shown in Table 1. 67.8% of the population presented the disease in the age range of 1 to 9 years (based on 393 located medical records), with a median age of 6 years. The classification of the age groups was based on the risk assignment for the disease by age (high risk, under 1 year and from 10 to 17 years and standard risk, from 1 to 9 years). Regarding sex (located medical records 430), 46.3% of the population was female and 53.7% was male. Most cases (88.3%) were of lymphoid lineage, with the predominant immunophenotype being B cells at 95.6%. In cases of myeloid leukemia, acute myeloid leukemia (AML) predominated, accounting for 71.2% of cases (located medical records of leukemia categorized by cell lineage, 430).

An analysis was conducted to determine whether there was a relationship between sex and leukemia type by lineage. Among males, 203 cases of lymphoid leukemia (88%) and 29 cases of myeloid leukemia (12%) were identified. Among females, there were 174 cases of lymphoid leukemia (89%) and 22 cases of myeloid leukemia (11%). A Pearson chi-square test was performed (χ² = 0.300, p = 0.861), indicating no statistically significant association between sex and leukemia type.

An analysis was also conducted to examine the relationship between age, sex, and leukemia type with paternal occupation (related or unrelated to agrochemicals) in avocado-growing and non-avocado-growing areas (Table 2). In both regions, most cases were concentrated in the 1–9 years age group, followed by individuals over 10 years of age and, lastly, infants under 1 year old. No statistically significant differences were found between age groups (χ² = 0.639, p = 0.424 in the avocado-growing area; χ² = 0.001, p = 1.0 in the non-avocado-growing area). Regarding sex, in the avocado-growing area, 42 cases were recorded in males and 36 in females whose fathers had occupations related to agrochemicals. In contrast, among fathers with other occupations, 61 cases were found in males and 53 in females. No statistically significant differences were observed between groups related or unrelated to agrochemicals (χ² = 0.002, p = 0.963). Lastly, when comparing leukemia by cell lineage, in the avocado-growing area, 70 cases of lymphoid leukemia were recorded among children of exposed workers to agrochemicals and 100 among children of non-exposed workers. For myeloid leukemia, eight and 13 cases were recorded, respectively, with no statistically significant difference (χ² = 0.073, p = 0.786). In the non-avocado-growing area, 56 cases of lymphoid leukemia were identified among exposed individuals and 35 among non-exposed individuals, while for myeloid leukemia, 6 and 9 cases were reported, respectively, with no significant association (χ² = 2.461, p = 0.117). These results indicate that there is no statistically significant relationship between paternal occupation and age, sex, or leukemia lineage in any of the evaluated areas.

In terms of the demographic characteristics of the parents related to maternal occupation (available in 353 medical records) (Table 3), household-related activities predominated, accounting for 79.6%. Only 3.4% of mothers were engaged in occupations involving agrochemicals, and 1.4% in work related to paints and solvents.

As for paternal occupation (available in 333 medical records), 46.6% reported employment related to agrochemicals, including farming, day labor, and subsistence agriculture. However, information on the type of agrochemical exposure, as well as the intensity and duration, was not available in the medical records. Additionally, 12.5% of fathers worked in construction, and 6.5% were involved in activities related to paints and solvents (Table 3). Of the total number of cases where both paternal and maternal occupations were recorded (353), only 2.2% (8 cases) involved both parents working in activities related to agrochemicals. Occupations classified as “Other”, along with details of the previously mentioned activities, can be found in Table S1.

Frequency, incidence and risk of leukemia in the avocado-growing area

In Michoacán, avocado is the most important perennial crop in terms of cultivated area (INEGI, 2021), and based on this, we assessed whether the residence (near or in the same place as the orchards) of the population with leukemia under study was related to the frequency of the disease. For this purpose, the state was grouped into avocado-growing zone of residence and non-avocado-growing zone of residence. The avocado-growing zone included those municipalities in the state that are registered in the “Servicio de Información Agroalimentaria y Pesquera” (SIAP, report 2023) as avocado producers (see Table S2). Based on this information, 65.4% of cases (246) were found in children residing in the avocado-growing area, 28.5% (107 cases) in the non-avocado-growing area, and 6.1% (23 cases) in children with leukemia residing outside the state. To assess potential differences in case incidence between the studied areas, a SIR, or Observed/Expected (O/E) ratio. In the avocado-growing area, with a cumulative average pediatric population of 858,773, the 10-year incidence rate was 28.6 cases per 100,000 children, with 227.3 expected cases and an SIR of 1.08 (95% CI [0.951–1.226]). In the non-avocado-growing area, with 470,891 children, the incidence rate was 22.7 cases per 100,000 children, with 124.6 expected cases and an SIR of 0.85 (95% CI [0.703–1.03]).

When both regions were combined, the resulting SIR was 1.00 (95% CI [0.901–1.113]) (Table 4). The 95% confidence interval analysis indicates that none of the evaluated areas showed statistically significant differences compared to the expected number of cases under a Poisson distribution.

In addition, the relationship between parental occupation and residence in the avocado-growing area was evaluated using logistic regression analysis, adjusting for age and sex. Paternal occupations related to agrochemical use were associated with the avocado-growing zone, with an OR = 1.764 (95% CI [1.034–3.009], p = 0.0379). Maternal occupations related to agrochemical use were not associated with the avocado-growing area, with an OR = 1.515 (95% CI [0.769–2.988], p = 0.230). The reference category consisted of other types of occupations (OR = 1). The other paternal or maternal occupations are reported in Table 5. None of the maternal occupation categories yielded statistically significant results (Table 5), suggesting that maternal occupation does not significantly influence the likelihood of residing in an avocado-growing area.

Survival

Of the total cases (Table 3), 68% are alive, 24% died, and 8% abandoned treatment. Medical records of live patients indicate that 70% are under surveillance, 25% are in treatment, 4% had a relapse and are still in treatment, and 1% were discharged from the oncology service. Table 6 shows the breakdown of live, relapsed, and deceased cases, as well as the evolutionary status by type of leukemia. Age groups classified as high risk (under 1 year and over 10 years) together accounted for 56.1% of deaths, while the age group from 1 to 9 years (standard risk) accounted for 43.9% of deaths. Additionally, the observed frequency of mortality during remission induction chemotherapy was 5.3% (n = 25). The first cause of death recorded in the CIS was septic shock or sepsis (45%), followed by hemorrhage (respiratory tract or intracranial) (16%), and leukemia as the base cause was the third with 8% of the cases (Table S3).

A statistical analysis was conducted to evaluate the relationship between survival status (alive or deceased) and sex among the studied cases. A total of 365 records were analyzed, showing that among males, 134 were alive (68.4%) and 62 were deceased (31.6%), while among females, 108 were alive (64.3%) and 60 were deceased (35.7%). Pearson’s chi-square test indicated no statistically significant association between sex and survival status (χ² = 1.181, p = 0.554).

Regarding survival by zone of residence, there is no statistically significant difference between survival in patients from the avocado-growing zone and patients from the non-avocado-growing zone. The estimated survival by zone was 121.8 months for the avocado-growing zone, 120.4 months for the non-avocado-growing zone, and 132.4 months for those living in other states of the Republic (Fig. 2A). In contrast, survival as a function of occupation shows a mean survival of 139.3 months for patients whose parents are engaged in agrochemical-related activities, 107.043 months for those whose parents are involved in construction-related activities, 89.2 months for children whose parents have an occupational activity related to paints and solvents, and 117.078 months for those whose parents have another occupational activity (Fig. 2B). The log-rank test indicates a significant difference between the survival curves (p = 0.0148), suggesting that survival appears to be more favorable for leukemia patients whose parents are engaged in some activity related to agrochemicals.

Additionally, a Cox regression analysis was conducted to evaluate the relative risk associated with combinations of residence zone and parental occupation. Occupation was grouped into activities related to agrochemicals and other types of activities. The overall model was not statistically significant, presenting a -2LL (log-likelihood) value of 817.502, p = 0.130. The study population was divided into four groups, yielding the following results: children residing in the avocado-growing zone with parents engaged in agrochemical-related activities (HR = 0.508, 95% CI [0.248–1.040], p = 0.064); avocado zone resident children with parents engaged in other activities (HR = 0.840, 95% CI [0.459–1.538], p = 0.572); non-avocado zone resident children with parents engaged in agrochemical-related activities (HR = 0.521, 95% CI [0.244–1.112], p = 1.112); while children residing in non-avocado-growing zones with parents engaged in other activities (reference group). These results suggest that, overall, the combinations of residence zone and parental occupation are not significantly associated with survival (Fig. 3, Table 7).

Mapping by municipality of residence and ZIP codes

On the other hand, based on the municipality of residence of the leukemia cases studied (from medical records 352), a map of Michoacán de Ocampo, México was created (Fig. 4 and Table 8), showing the 113 municipalities and leukemia cases using a color code. The avocado agricultural frontier (black line) and avocado growing areas (dotted in green) were delimited. The avocado-growing fringe comprises 65 municipalities with a frequency of 60% of all cases, consistent with the 2023 SIAP register (65.4%) (SIAP, 2023). The most representative municipalities are Morelia (45 cases), Uruapan (19), Hidalgo (14), Huetamo (13), Tacámbaro (13), and Apatzingán (10). By grouping, in the avocado-growing zone there are 2 municipalities with more than 15 cases, 2 municipalities with 9 to 14 cases, 2 municipalities with 6 to 8 cases, 55 municipalities with 1 to 5 cases, and 4 municipalities with no cases. In the zone outside the avocado frontier, there are 19 municipalities with no cases, 2 municipalities with 6 to 8 cases, 3 municipalities with 9 to 14 cases, and 24 municipalities with 1 to 5 cases. The Standardized Incidence Ratio (SIR) or Observed/Expected (O/E) ratio, based on a Poisson distribution, was also calculated for each municipality in the state. However, since most municipalities reported fewer than 10 cases, the SIR estimates in these areas are not statistically reliable, which limits their interpretation and validity (Table S4). Nevertheless, some municipalities showed a higher number of observed cases than expected. Specifically, the municipalities of Coeneo, Cherán, Jiménez, Tacámbaro, Tzintzunzán, and Zinapécuaro in the avocado-growing zone, and Huetamo and Múgica in the non-avocado-growing zone, all presented SIR values greater than 2. These findings may suggest areas of potential epidemiological interest, although they should be interpreted with caution due to small sample sizes and wide confidence intervals (Table S4).

Finally, Fig. 5 shows the areas of residence of leukemia cases by zip codes (located medical records 307). Within the avocado agricultural frontier, cases are clustered in areas with a high density of avocado orchards. In Uruapan (municipality 102, Fig. 4), 13 of the 19 cases are concentrated in contiguous zip codes; between Pátzcuaro (066) and Salvador Escalante (079), there are 12 contiguous cases; and in Tacámbaro (082), the 13 cases are distributed in only 4 nearby zip codes. Furthermore, in Hidalgo (34), an area with a low density of avocado orchards, 13 of the 14 cases are in contiguous zip codes. Outside the avocado agricultural frontier, in Huetamo (038), 11 of the 13 cases are in the same zip code; in Múgica (055), 7 of the 9 cases are in the same zip code; in La Huacana (035), 4 of the 6 cases are in the same zip code. In Apatzingán (006), 8 of 10 cases are in a contiguous area despite being spread across 7 zip codes. Likewise, for Zamora (108), which has 9 cases, and Jacona (43), which has 5 cases, the zip codes of the cases were nearby.