Report · Phase 1c
Esophageal cancer: did the excess already exist from 1996 to 2014?
A historical series of almost 30 years of mortality in the 16 municipalities with a signal and the 24 excess areas. Versão em português.
Author: Clériston Almeida Capistrano (ORCID 0009-0008-5280-1869) · independent researcher with no formal training in health sciences · cledon.org
Exploratory study — not conclusions. Ecological description of recorded mortality; it does not indicate causation, does not allow estimating any person's risk and does not change any medical recommendation. Version 1.7 · November 13, 2026. DOI: 10.5281/zenodo.24195057. This is a translation of the Portuguese report; in case of discrepancy, the Portuguese version prevails.
Abstract
The 16 municipalities with a signal and the 24 excess areas were identified with deaths from 2020–2024. Phase 1c goes back in time, with deaths from 1996 to 2014 that played no part in the selection, and asks whether the excess already existed. The analysis plan was committed to git before the data (commit 9847190); the report incorporates an AI-assisted technical review (not independent peer review).
- The excess already existed before 2015. Taken together, in 2000–2014 the 16 municipalities had esophageal cancer mortality 1.91 times that expected from their states' rates (95% CI 1.74–2.10; p < 0.001). Individually, 10 of the 16 meet the excess criterion (SMR > 1 and p < 0.05) and 14 have an SMR above 1.
- So do the areas. 18 of the 24 scan areas meet the criterion in 2000–2014; 22 have an SMR above 1.
- The full series, from 1996 to 2024, stays above 1 in every period: in the 16 municipalities, from 1.63 (1996–1999) to 3.44 (2020–2024); in the 24 areas, from 1.23 to 1.65.
- Specific to the esophagus. In the same municipalities, lung and other cancers were below expectation in 2000–2014 (SMR 0.87 and 0.84).
- Constant territory. With minimum comparable areas (4 of the 16 municipalities lost territory after 1991), the 16 together have an SMR of 1.91 (1.74–2.08) in 2000–2014; 11 of the 16 and 18 of the 24 areas meet the criterion.
- Main caveat: recording quality improved greatly over the period. In the 16 municipalities, ill-defined causes fell from 25% of deaths in 1996–1999 to 11% in 2010–2014. Part of the rise in the SMR over time may reflect this improvement rather than a real increase.
1. Methods
Data. Deaths of residents from SIM/DATASUS (the national mortality information system), 1996 to 2014 (513 files, all states), with the same aggregation rules as in earlier phases. In the older files the municipality code has 7 digits; we use the first 6. Population by municipality, sex and age group from the 1991, 2000, 2010 and 2022 Censuses. Expected deaths from the state's own rates in the same year, by sex and age group. SMR with exact 95% CI and one-sided Poisson test.
Departures from the plan (not pre-specified). (1) For the 1,079 municipalities created after 1991, the plan specified the population of the following Census; this double-counts their territory in 1996–1999 (it was already part of the municipality of origin in 1991) and inflated the country's population by about 2%, which would artificially increase the excess. In the main analysis, 1996–1999 uses the 2000 municipal division, scaled by each state's population change, by sex and age group, between 1991 and 2000; the plan's method is kept as a sensitivity analysis (section 6). (2) A coding error inverted the p-value; it was fixed before interpretation and checked against the exact Poisson test. (3) After an AI-assisted technical review (not peer review): an analysis with minimum comparable areas (section 6); redistribution of ill-defined causes presented as descriptive only, with treatment of strata with no defined cause (section 5); the quality table extended to the areas' 14 states; results stored without rounding counts.
Data security. Public SIM microdata, with no names, processed locally into counts aggregated by municipality, year, sex, age group and cause group. No individual data are published; the original files are kept outside the repository, with their SHA-256 recorded for reproduction.
2. The full series, 1996–2024
Esophageal cancer SMR by period, relative to the state. 1996–2014: this phase; 2015–2024: Phase 1b (same method, with the state's rates in the same year). Above 1 = more deaths than expected.
By region, the story differs. In the South, the excess is high and stable since 1996 (SMR 2.47 in 1996–1999, 57 deaths). In Minas Gerais, the 1996–1999 SMR is 0.90 (CI 0.55–1.37), with no clear excess. These 9 municipalities had 34.9% of deaths with an ill-defined cause in that period, against 14.8% in the state; with the descriptive redistribution of these causes (section 5), the SMR becomes 1.36. In Bahia, the 2 municipalities recorded no esophageal cancer deaths in 1996–1999 (expected 1.4), with 33.8% ill-defined causes.
3. The 16 municipalities in 2000–2014
| Municipality | Deaths 2000–14 | Expected | SMR (95% CI) | p | Replicates |
|---|---|---|---|---|---|
| Brasília de Minas (MG) | 37 | 23.4 | 1.58 (1.11–2.18) | 0.006 | yes |
| Braúnas (MG) | 14 | 5.4 | 2.61 (1.43–4.38) | 0.001 | yes |
| Capelinha (MG) | 33 | 18.7 | 1.77 (1.22–2.48) | 0.002 | yes |
| Caraí (MG) | 11 | 16.0 | 0.69 (0.34–1.23) | 0.923 | no |
| Cipotânea (MG) | 12 | 5.6 | 2.13 (1.10–3.72) | 0.013 | yes |
| Dom Feliciano (RS) | 25 | 19.4 | 1.29 (0.84–1.91) | 0.123 | no |
| Ervália (MG) | 26 | 14.1 | 1.84 (1.20–2.70) | 0.003 | yes |
| Itaiópolis (SC) | 37 | 17.1 | 2.17 (1.53–2.99) | < 0.001 | yes |
| Mallet (PR) | 42 | 13.2 | 3.19 (2.30–4.32) | < 0.001 | yes |
| Matina (BA) | 7 | 3.9 | 1.80 (0.72–3.71) | 0.099 | no |
| Palmas de Monte Alto (BA) | 7 | 7.7 | 0.90 (0.36–1.86) | 0.653 | no |
| Poté (MG) | 25 | 13.0 | 1.92 (1.24–2.83) | 0.002 | yes |
| Prudentópolis (PR) | 113 | 44.3 | 2.55 (2.10–3.07) | < 0.001 | yes |
| Riozinho (RS) | 8 | 6.2 | 1.29 (0.56–2.54) | 0.285 | no |
| Santa Maria do Suaçuí (MG) | 16 | 11.0 | 1.46 (0.83–2.37) | 0.090 | no |
| Serro (MG) | 34 | 14.8 | 2.30 (1.59–3.21) | < 0.001 | yes |
Did not meet the criterion: Caraí (MG), Dom Feliciano (RS), Matina (BA), Palmas de Monte Alto (BA), Riozinho (RS), Santa Maria do Suaçuí (MG). The two in Bahia and the two in Rio Grande do Sul have few expected deaths.
4. The 24 areas in 2000–2014
| Area | Deaths 2000–14 | Expected | SMR (95% CI) | p | Replicates |
|---|---|---|---|---|---|
| 01 — Angelândia (MG) | 592 | 350 | 1.69 (1.56–1.83) | < 0.001 | yes |
| 02 — São Mateus do Sul (PR) | 1,222 | 580 | 2.11 (1.99–2.23) | < 0.001 | yes |
| 03 — Sebastião Laranjeiras (BA) | 212 | 185 | 1.15 (1.00–1.31) | 0.026 | yes |
| 04 — São José do Belmonte (PE) | 514 | 293 | 1.75 (1.61–1.91) | < 0.001 | yes |
| 05 — Alvorada de Minas (MG) | 421 | 237 | 1.78 (1.61–1.96) | < 0.001 | yes |
| 06 — Patis (MG) | 656 | 463 | 1.42 (1.31–1.53) | < 0.001 | yes |
| 07 — Goioxim (PR) | 767 | 551 | 1.39 (1.30–1.49) | < 0.001 | yes |
| 08 — Itaocara (RJ) | 676 | 330 | 2.05 (1.90–2.21) | < 0.001 | yes |
| 09 — Santa Cruz do Escalvado (MG) | 924 | 621 | 1.49 (1.39–1.59) | < 0.001 | yes |
| 10 — Itambé (BA) | 276 | 265 | 1.04 (0.92–1.17) | 0.262 | no |
| 11 — Nanuque (MG) | 530 | 438 | 1.21 (1.11–1.32) | < 0.001 | yes |
| 12 — Alpercata (MG) | 493 | 338 | 1.46 (1.33–1.59) | < 0.001 | yes |
| 13 — Ivorá (RS) | 854 | 821 | 1.04 (0.97–1.11) | 0.129 | no |
| 14 — Rio das Flores (RJ) | 671 | 424 | 1.58 (1.46–1.71) | < 0.001 | yes |
| 15 — São Domingos (GO) | 50 | 38 | 1.30 (0.96–1.71) | 0.042 | yes |
| 16 — Padre Carvalho (MG) | 118 | 110 | 1.08 (0.89–1.29) | 0.228 | no |
| 17 — Poço Dantas (PB) | 513 | 401 | 1.28 (1.17–1.40) | < 0.001 | yes |
| 18 — Riozinho (RS) | 51 | 32 | 1.59 (1.18–2.08) | 0.001 | yes |
| 19 — São José do Cedro (SC) | 175 | 160 | 1.09 (0.94–1.27) | 0.126 | no |
| 20 — Casa Branca (SP) | 589 | 600 | 0.98 (0.90–1.06) | 0.674 | no |
| 21 — Alto Rio Doce (MG) | 83 | 31 | 2.70 (2.15–3.35) | < 0.001 | yes |
| 22 — Rosário do Sul (RS) | 738 | 514 | 1.44 (1.33–1.54) | < 0.001 | yes |
| 23 — Arapeí (SP) | 690 | 520 | 1.33 (1.23–1.43) | < 0.001 | yes |
| 24 — Eunápolis (BA) | 53 | 55 | 0.97 (0.73–1.27) | 0.603 | no |
5. Recording quality
Percentage of deaths with an ill-defined cause (ICD-10 R00–R99), by period, in the sets analysed and in the 14 states with municipalities in the 24 areas.
| Set | 1996–1999 | 2000–2004 | 2005–2009 | 2010–2014 |
|---|---|---|---|---|
| The 16 municipalities | 25.0% | 21.2% | 16.5% | 10.8% |
| The 9 in Minas Gerais | 34.9% | 29.9% | 23.7% | 14.4% |
| The 5 in the South | 14.2% | 9.8% | 8.0% | 5.9% |
| The 2 in Bahia | 33.8% | 29.1% | 10.1% | 9.8% |
| The 24 areas | 20.7% | 17.5% | 11.9% | 8.6% |
| State: BA | 28.8% | 26.7% | 16.4% | 13.2% |
| State: CE | 26.5% | 23.6% | 8.0% | 5.6% |
| State: ES | 19.3% | 11.3% | 2.9% | 1.1% |
| State: GO | 14.2% | 8.4% | 6.5% | 3.7% |
| State: MG | 14.8% | 13.3% | 11.3% | 8.7% |
| State: PB | 52.9% | 39.2% | 9.9% | 6.7% |
| State: PE | 25.5% | 19.4% | 6.3% | 4.9% |
| State: PR | 7.7% | 5.0% | 4.9% | 4.2% |
| State: RJ | 10.2% | 10.6% | 8.2% | 6.1% |
| State: RN | 30.0% | 25.5% | 4.8% | 3.4% |
| State: RS | 5.4% | 5.4% | 4.8% | 4.6% |
| State: SC | 14.0% | 11.1% | 8.0% | 4.0% |
| State: SP | 6.3% | 6.5% | 6.3% | 5.6% |
| State: TO | 28.1% | 12.6% | 4.2% | 3.0% |
In areas with many ill-defined causes, some esophageal cancer deaths go without a recorded diagnosis. The pre-specified sensitivity analysis redistributes ill-defined deaths among the defined causes, proportionally, by municipality, year, sex and age group. Of the 1,997,178 ill-defined deaths in 1996–2014 in the country (all municipalities), 189,804 (9.5%) were in strata with no defined cause at all; for these we used the state's proportion in the same year, sex and age group (an adjustment made after the review, not pre-specified; without it, these deaths would be left unallocated).
| Set | Calculation | 1996–1999 | 2000–2004 | 2005–2009 | 2010–2014 | Deaths 1996–2014 |
|---|---|---|---|---|---|---|
| The 16 municipalities | main | 1.63 | 1.73 | 1.87 | 2.09 | 525 |
| The 16 municipalities | redistributing ill-defined | 1.83 | 1.89 | 1.96 | 2.18 | ≈ 626 |
| Minas Gerais (9) | main | 0.90 | 1.51 | 1.61 | 1.92 | 229 |
| Minas Gerais (9) | redistributing ill-defined | 1.36 | 1.87 | 1.77 | 2.04 | ≈ 303 |
| South (5) | main | 2.47 | 1.99 | 2.34 | 2.40 | 282 |
| South (5) | redistributing ill-defined | 2.48 | 1.98 | 2.41 | 2.47 | ≈ 306 |
| Bahia (2) | main | 0.00 | 1.24 | 0.52 | 1.68 | 14 |
| Bahia (2) | redistributing ill-defined | 0.09 | 1.30 | 0.46 | 1.71 | ≈ 17 |
| The 24 areas | main | 1.23 | 1.33 | 1.40 | 1.51 | 13,924 |
| The 24 areas | redistributing ill-defined | 1.31 | 1.38 | 1.45 | 1.54 | ≈ 16,038 |
With redistribution, deaths are no longer whole counts and the uncertainty of the redistribution itself is not part of the calculation; it is therefore presented only as a descriptive sensitivity analysis (SMR, with no interval, p-value or "replicates" classification). The series becomes more stable over time. Redistribution assumes that ill-defined deaths are split among causes in the same way as defined ones, which is an approximation.
6. Comparison and sensitivity analyses
Comparison outcomes, 2000–2014.
| Set | Esophagus | Lung | Other cancers |
|---|---|---|---|
| The 16 municipalities | 1.91 | 0.87 | 0.84 |
| Minas Gerais (9) | 1.71 | 0.54 | 0.70 |
| South (5) | 2.25 | 1.16 | 1.04 |
| The 24 areas | 1.42 | 0.90 | 0.93 |
For the 16 together, lung and other cancers are below 1; in the South, lung is slightly above (1.16), well below the esophagus (2.25), which is compatible with tobacco as a shared exposure.
Constant territory (minimum comparable areas; not pre-specified, done after the review). When a municipality is split, the territory to which deaths and population refer changes. 4 of the 16 municipalities lost territory after 1991. The territorial sensitivity analysis in the plan (excluding municipalities created after 1991) does not address this: the municipalities of origin remain in the calculation and none of the 16 was created after 1991, so for the 16 it repeats the main analysis. We therefore used IPEA's 1991–2010 minimum comparable areas (geobr package), which group municipalities so as to keep the same territory in every year. Each unit becomes the union of its municipalities' comparable areas, with population interpolated between Censuses within each area and expected deaths from the state's rates. One comparable area spanning municipalities in Minas Gerais and Espírito Santo was split by state; the 5 municipalities created in 2013 were linked to the area of their municipality of origin.
| Set and period | Main | Minimum comparable areas | Population by the plan's method |
|---|---|---|---|
| The 16 municipalities (2000–2014) | 1.91 (1.74–2.10) | 1.91 (1.74–2.08) | 1.91 (1.74–2.10) |
| Minas Gerais (9) (2000–2014) | 1.71 (1.48–1.95) | 1.68 (1.48–1.91) | 1.71 (1.48–1.95) |
| South (5) (2000–2014) | 2.25 (1.96–2.56) | 2.28 (2.01–2.59) | 2.25 (1.96–2.56) |
| The 24 areas (2000–2014) | 1.42 (1.39–1.45) | 1.37 (1.34–1.39) | 1.42 (1.39–1.45) |
| The 16 municipalities (1996–1999) | 1.63 (1.29–2.03) | 1.50 (1.19–1.86) | 1.58 (1.25–1.97) |
| Minas Gerais (9) (1996–1999) | 0.90 (0.55–1.37) | 0.90 (0.58–1.32) | 0.86 (0.53–1.31) |
| South (5) (1996–1999) | 2.47 (1.87–3.20) | 2.25 (1.70–2.92) | 2.41 (1.83–3.13) |
| The 24 areas (1996–1999) | 1.23 (1.18–1.29) | 1.21 (1.16–1.26) | 1.21 (1.16–1.27) |
The 4 municipalities that lost territory, in the corresponding comparable area (2000–2014):
| Municipality | Current municipalities in the comparable area | Deaths 2000–14 | SMR (95% CI) | p | Replicates |
|---|---|---|---|---|---|
| Brasília de Minas (MG) | 4 | 57 | 1.55 (1.17–2.00) | 0.001 | yes |
| Capelinha (MG) | 2 | 39 | 1.70 (1.21–2.32) | 0.001 | yes |
| Santa Maria do Suaçuí (MG) | 2 | 23 | 1.57 (0.99–2.35) | 0.026 | yes |
| Itaiópolis (SC) | 2 | 57 | 2.35 (1.78–3.05) | < 0.001 | yes |
With comparable areas, 11 of the 16 municipalities and 18 of the 24 areas meet the criterion in 2000–2014 (14 and 21 with an SMR above 1). Series by period with comparable areas:
| Set | 1996–1999 | 2000–2004 | 2005–2009 | 2010–2014 |
|---|---|---|---|---|
| The 16 municipalities | 1.50 | 1.68 | 1.83 | 2.15 |
| Minas Gerais (9) | 0.90 | 1.45 | 1.59 | 1.91 |
| South (5) | 2.25 | 1.97 | 2.29 | 2.57 |
| The 24 areas | 1.21 | 1.27 | 1.35 | 1.46 |
7. What this adds, with caution
The excess of esophageal cancer mortality in these municipalities and areas is neither a recent phenomenon nor a fluctuation of 2020–2024: it appears in data from 2000 to 2014, which played no part in the selection, and in the South since 1996. It is specific to the esophagus in the same municipalities and holds with constant territory. The rise in the SMR over time should be read with care, because it coincides with improved recording of causes of death, especially in Minas Gerais and Bahia. The results reinforce that these regions are worth investigating, but they remain an ecological description: they do not indicate causation.
8. Limitations
- SIM quality in the 1990s, especially in rural areas of Minas Gerais and Bahia.
- Population between Censuses is interpolated; in the 1990s, with the 2000 municipal division. The 4 municipalities that lost territory were checked with comparable areas, which are larger than the municipality.
- The 63 municipalities that did not exist in 2000 use the 2010 population in 2000–2009, which double-counts a small area (about 0.3% of the population).
- The 16 municipalities and 24 areas were chosen for their excess in 2020–2024; this does not affect the 1996–2014 data, but the result remains an ecological description.
- Analyses by an independent researcher with no formal training in health sciences, assisted by AI (Anthropic's Claude), under human review.
9. References
- Breslow NE, Day NE. Statistical Methods in Cancer Research, vol. II: The Design and Analysis of Cohort Studies. IARC Scientific Publications No. 82. Lyon: IARC; 1987.
- Garwood F. Fiducial limits for the Poisson distribution. Biometrika. 1936;28(3/4):437–442.
- Reis E, Pimentel M, Alvarenga AI. Áreas mínimas comparáveis para os períodos intercensitários de 1872 a 2000. Rio de Janeiro: IPEA; 2008.
- Pereira RHM, Gonçalves CN, et al. geobr: Download Official Spatial Data Sets of Brazil. IPEA (áreas mínimas comparáveis 1991–2010).
- Mathers CD, Ma Fat D, Inoue M, Rao C, Lopez AD. Counting the dead and what they died from: an assessment of the global status of cause of death data. Bull World Health Organ. 2005;83(3):171–177.
- Morgenstern H. Ecologic studies in epidemiology: concepts, principles, and methods. Annu Rev Public Health. 1995;16:61–81.
- Ministério da Saúde. Sistema de Informações sobre Mortalidade (SIM), 1996–2014. DATASUS.
- Instituto Brasileiro de Geografia e Estatística. Censos Demográficos 1991, 2000, 2010 e 2022. Sistema IBGE de Recuperação Automática (SIDRA).
Code at https://cledon.org/codigo: historico.py, gerar_fase1c.py.
Plan: PLANO-fase1c.md (commit 9847190). Results: fase1c_*.csv on the map page and in the
Zenodo deposit.
In memory of Francisco Cledio Capistrano (F.C.C.).