Координаты времени и пространства лимфомы Ходжкина в Сардинии, Италия.

Introduction

In 1832, the Thomas Hodgkin's article “ On Some Morbid Appearances of the Absorbent Glands and Spleen ” described the clinical and autopsy findings in seven patients with lymph node and spleen enlargement. 1 Almost two centuries afterwards, despite achieving high survival rates and the progress in the diagnosis and therapy of the disease, the etiology of Hodgkin's lymphoma (HL) is still a matter of debate. In 1994, the Revised European-American Lymphoma (REAL) classification, using morphologic, phenotypic, genotypic, and clinical findings, classified the various HL forms into 2 main subtypes: nodular lymphocyte predominant and classic HL (about 90–95% of incident HL cases), 2 , 3 , 4 the last including four specifically defined entities: nodular sclerosis, mixed cellularity, lymphocyte rich, and lymphocyte depleted. The subsequent World Health Organization classification of lymphomas adopted the same scheme. 2

HL shows a peculiar bimodal age distribution, and a median age at diagnosis of 39 years, substantially lower than most lymphohemopoietic malignancies, peaking around age 30 and after age 50 years. 5 The nodular sclerosis subtype is the most prevalent (74% of classical forms) and differs from the other forms as it affects mainly adolescents and young adults, shows upward incidence trends, 4 and prevails in the high socio-economic status and rural areas, so originating the two-disease hypothesis. 6 Also, HL accounts for around 15% of all cancers among young adults (15–24 years) and the prognosis is still unfavorable for the elderly (>50–60 years), who show lower complete remission rates and overall survival. 5 Cancer registry data only consider all HL forms combined, which show a wide geographic variation in incidence with lowest rates among the American natives (men: 0.3–0.9 per 100,000; women: 0.3–1.3) and highest in Europe (men: 1.8–6.7; women: 1.2–4.7), and among the North American populations of non-Asian descent (men: 1.8–4.6; women: 1.0–3.1), and male rates prevailing in most cancer registries. 7 HL incidence among immigrants of Chinese descent in British Columbia, Canada were lower than the British Columbia population but more elevated than in the country of origin, with the difference mostly accounted for by young immigrants with nodular sclerosis as the prevalent HL subtype. 8 A robust evidence supports the role of genetic factors 5 consistent with recessive inheritance. Interleukin gene polymorphisms and, particularly, IL6 gene polymorphisms might be implicated. 5 Millennia of geographic isolation and selective pressure from malaria resulted in a high prevalence of gene polymorphisms associated with multiple sclerosis, 9 childhood type I diabetes, 10 thalassemia, 11 and other rare diseases among the population of the regional island of Sardinia, Italy. On the other hand, some areas of the island exhibit the highest prevalence of male centenarians worldwide. 12 It is of interest to note that genome-wide studies revealed a genetic overlap particularly in the HLA region between HL and multiple sclerosis, 13 which link was also supported by a geographic correlation study. 14 For these reasons, the genetic peculiarity of the Sardinian population 15 provides an excellent setting for exploring HL time trend, geographical spread, and environmental correlates.

Several HL clusters have been described, including one among the Italian military deployed in the aftermath of the Balkan war, 16 and another around the mount Aetna volcanic area. 17 Evidence of geographic clustering exists for youngster cases 6 although, overall, is thought to be scarce. 5 Livestock breeding was a risk factor for HL in a U.S.A. study, 18 while studies of environmental, chemical, and occupational risk factors provided inconsistent results. 19 , 20 , 21

A previous Epstein-Barr virus (EBV) infection carries an increase in HL risk inversely related to time since infection, 5 suggesting a role for immune and inflammatory factors, 22 although results are inconsistent for other viruses and tonsillectomy. 5 Interestingly, we previously observed a correlation between COVID-19 incidence in the pre-vaccine year and the past incidence of non-Hodgkin's Lymphoma 23 ; the availability of population data on HL incidence might allow to explore whether a similar geographic connection also exists with past HL incidence.

As it concerns socio-economic and lifestyle variables, smoking has shown a significant association with HL risk among adult cases but not among subjects below age 35 years, 24 whilst alcohol 24 , 25 and dietary wholegrain intake, 26 and exposure to ultraviolet radiation during childhood and adulthood showed an inverse relationship, particularly among EBV positive subjects. 27

To further inquiry on the HL epidemiology, we explored the time trend and the geographical distribution of HL incidence (International Classification of Diseases 10th Revision (ICD-10) codes: C81.0–C81.9) 28 in the Sardinia region of Italy, and we used population data and an ecological study design to investigate potential risk factors.

Materials and methods

Study population

For the purposes of this study, we used a database of incident hemolymphatic cancer in Sardinia, Italy covering the years from 1974 to 2003 which was previously described 29 and validated. 30 , 31 Briefly, the database comprised the records of 14,744 cases of hematological malignancies diagnosed in the regional island of Sardinia, Italy in 1974–2003, including 1347 cases of Hodgkin's Lymphoma, 597 women and 750 men. For 24 cases (2.0%), the information on residence at the time of diagnosis was unavailable. Therefore, these cases contributed to the calculation of the age- and gender-specific regional rates but not to the geographic analysis nor the association with environmental risk factors. The database includes anonymous records grouped by gender, 10-year age-group, residence, diagnosis, and year of diagnosis, but no information on the specific HL subtype.

Statistical methods

Hodgkin's lymphoma incidence and time trend

For each administrative unit (or commune as in the Italian definition) existing in Sardinian as of 1974 ( n = 356), we calculated the age- and sex-standardized incidence rate of HL along the study period for the total population, for men and women separately, using the regional population as the standard. To allow international comparisons, we calculated the regional rates for the total population, by sex, and age up to 44 years or 45 years or more using the World Health Organization (WHO) standard population as the standard. We plotted the annual incidence rates in 1974–2003 and estimated the annual percent change (APC) using the Joinpoint software which also tests changes in the slope by identifying a pre-set number of knots over the years. The Joinpoint software, freely available online, 32 was developed by the Statistical Methodology and Applications Branch of the U.S. National Cancer Institute.

Spatial distribution of Hodgkin's lymphoma

We used a Bayesian approach 33 to inquiry into the geographical distribution of the probability of HL cases over the 356 Sardinian communes and described in detail the calculation of the posterior probabilities by commune. 31 Based on the prior regional rates, for each commune we calculated the distribution of the expected events and set at P = 0.001 the threshold to reject the null hypothesis of the posterior HL probability belonging to the distribution of the expected events. Then, we calculated the likelihood ratio between the probability associated with H 1 (alternative hypothesis) vs. that associated with H 0 (the null hypothesis of chance variation), and used a chromatic scale of increasing darkness to plot it on the map of the communes’ territorial borders (white P ≤ 0.165, light grey P = 0.166–0.335, medium-light grey P = 0.336–0.50, medium-dark grey P = 0.501–0.80, dark grey P = 0.801–0.95, black P ≥ 95%). The maps are publicly available on the Italian Institute for Statistics (ISTAT) website ( https://www.istat.it/it/archivio/104,317 ) under the Creative Commons BY 3.0 IT License. The analysis was conducted with Bespoke Python code.

Socio-economic factors and environmental exposures

We also ranked the 356 communes by the HL likelihood ratio below and above age 45, representing the phase shift in the age distribution ( Fig. 1 ) and used Spearman's correlation analysis to explore the spatial correlation between HL incidence in these two age-groups.

Incidence of Hodgkin's lymphoma by age in Sardinia, Italy.

We used publicly available population data to investigate possible environmental determinants of HL. These included: 1) the Italian Institute of Statistics (ISTAT) deprivation index ( http://istat.it ); 2) the kilometric distance from the nearest hospital on the shortest route ( https://maps.google.co.uk ); 3) the quintiles of the probability of α-radiation emission from the natural background above 300 Bq/m 3 (≤5%, 6–10%, 11–20%, 21–30%, 31% or more) 34 ; 4) the prevalent geology (quaternary or subsequent marine deposits, basalt and other effusive volcanic rocks, granite and intrusive volcanic rocks, and metamorphic rocks) of the communal territory 35 ; 5) a previously described score representing five community services to define the urban or rural type of commune 36 ; 6) a modified version of proximity to industrial or military areas or cork harvesting first described by Biggeri et al. 37 ; 7) the per-capita size of cattle, sheep, and goat breedings in each commune 38 as representative of potential exposure to transmissible zoonotic agents; 8) prevalence of multiple sclerosis by local health authority in 2016 39 ; and 9) cumulative incidence of COVID-19 by commune in the first pandemic year, before vaccines were introduced. 23 , 40

We calculated the relative risk for HL and its 95% confidence interval associated with increasing categories of economic deprivation, distance from the nearest hospital, type of residence, geology, background radiation, proximity to industrial and military settlements and cork farms, and size of livestock breeding relative to the resident population, with reference to the lowest category, the null category, or the quaternary marine deposits in the case of geology, using multivariable Poisson regression analysis, for the total cases and cases below the age of 45 or aged 45 or more at diagnosis. We used the 2016 incidence of multiple sclerosis by local Health Unit, 39 and that of COVID-19 by commune in 2020 23 , 40 as surrogates of the genetic make-ups and environmental conditions that modulate susceptibility to the two diseases, which link with HL was previously suggested, 14 or reported for non-Hodgkin's Lymphoma. 23 We explored with multivariable Poisson regression whether an elevated incidence of Multiple Sclerosis and COVID-19 reflected the past HL incidence by calculating HL risk associated with categories of increasing incidence of these two diseases. If so, we might corroborate the previous findings of a link between HL and these diseases. The analysis was conducted using SPSS® version 20.0.

The use of the 1974–2003 lymphohemopoietic malignancies database for scientific research was approved by the Ethics Committee of the University Hospital of Cagliari (protocol N. PG 2019/18,070, 18 December 2019) in agreement with the Code of Ethics of the World Medical Association (Declaration of Helsinki).

Results

The median age at diagnosis of the HL cases was 33.0 years (interquartile range [IQR] 23–51) with the expected bimodal distribution and peaked at age 25–34 and 65–74 years among men and 15–24 and 65–74 years among women ( Fig. 1 ). Male cases prevailed at all ages (male to female ratio 1.26:1), ranging from 1.15:1 at age 0–44 years to 1.52:1 at age ≥45 years. The HL age- and sex-standardized incidence rate (standard: world population) was 2.6 per 100,000 (95% CI, 2.5–2.8) ( Table 1 ) and it was similar below age 45 or at age ≥45 years.

Table 1 Standardized incidence rates and APC describing the time trend of HL incidence in the total population and by age group and sex in Sardinia, Italy. Table 1 Population subgroup No. cases Age- and sex-standardized incidence rate ×10 −5 (95% CI) * 1974–2003 APC (95% CI), % All ages  Men 750 2.9 (2.8–3.1) 2.00 (1.25–2.84)  Women 597 2.3 (2.2–2.5) 2.00 (1.05–3.14)  Total 1,347 2.6 (2.5–2.8) 2.00 (1.32–2.82) 0–44 years  Men 493 2.8 (2.5–3.0) 3.50 (2.51–4.67)  Women 428 2.5 (2.2–2.7) 3.10 (1.66–4.88)  Total 921 2.6 (2.4–2.8) 3.30 (2.45–4.41) 45+ years  Men 257 3.3 (2.9–3.7) −0.72 (−2.47–1.18)  Women 169 1.9 (1.6–2.2) 0.23 (−1.28–2.04)  Total 426 2.6 (2.3–2.8) −0.44 (−1.99–1.31) ⁎ World Standard population. Abbreviations: APC, annual percent change; CI, confidence interval; HL, Hodgkin's lymphoma.

Standardized incidence rates and APC describing the time trend of HL incidence in the total population and by age group and sex in Sardinia, Italy.

World Standard population.

Abbreviations: APC, annual percent change; CI, confidence interval; HL, Hodgkin's lymphoma.

Time trend in Hodgkin's lymphoma incidence

Fig. 2 shows the plots of HL time trends, and Table 1 the APC values for the total population and by sex and age. In 1974–2003, HL incidence showed a significant upward trend, with an annual APC of 2.0% (95% CI, 1.05–3.14) which did not vary by sex. The linear upward trend was confirmed among subjects aged ≤44 years, with an annual APC of 3.3% (95% CI, 2.45–4.41) also similar in men and women, but not among those aged ≥45 years.

1974–2003 trend of Hodgkin's lymphoma over the whole population and by sex and age in Sardinia, Italy.

Spatial distribution of Hodgkin's lymphoma

Fig. 3 shows the geographic distribution of the likelihood ratio of a posterior probability of HL incidence above the 0.001 threshold in the overall population and by sex and age. None of the communes showed a significant excess probability of HL occurrence for the total or the male population. Only a small commune located in the central-western part of the island had a significantly elevated posterior probability of HL among the female population, based on 4 observed cases vs. 0.8 expected. Interestingly, other three bordering communes had a non-significantly elevated probability of HL among their female population. Overall, 13 female cases occurred in these four communes vs . 5.7 expected suggesting a spatial aggregation in the area ( P = 0.002); in the same communes there were six male cases vs. 6.9 expected. Twelve out of the 13 female cases occurred in 1980–2003, evenly distributed over three decades; ten out of 13 were young adults aged 15–35 years and the median age at diagnosis was 26 (IQR, 20–35) years. A significant excess of HL cases aged ≤44 years occurred in the population of both sexes (27 observed vs 13.7 expected, P = 0.008) of a commune located in the suburban area of the region's capital. The posterior probability of an increased HL occurrence at age ≥45 years was of borderline significance only in a small rural commune, based on four cases vs. 1.33 expected ( P = 0.049). Across the region, the posterior 95% probabilities of HL below or ≥45 years were weakly correlated (Spearman's correlation coefficient = 0.133, P = 0.012, degrees of freedom = 354).

Map of 1974–2003 posterior probability of a Hodgkin's lymphoma incidence higher than the prior regional rate by sex and in the total population in Sardinia, Italy. The following chromatic scale was used for the area of each commune: white P ≥ 0.165, light grey P = 0.166-0.335, medium-light grey P = 0.336-0.500, medium-light grey P = 0.501-0.800, dark grey P = 0.801-0.950, P ≥ 95%. The red circles in map C identify the 11 urban communes as defined in the Materials and methods. Maps in this figure were verified by the Ministry of Natural Resources of the People's Republic of China [verification code: GS Jing (2024) 2565].

Socio-economic factors and environmental exposures

Table 2 shows the HL risk associated with socio-economic and environmental variables. Risk showed an 11% increase among residents in urban (RR = 1.11, 95% CI, 1.04–1.18) and semi-urban communes (RR = 1.14, 95% CI, 0.98–1.33), and decreased moderately though significantly with distance from the nearest hospital ( P = 0.001), consistent with the lower risk in rural areas. A modest decrease in risk was also observed among residents in communes with a high density of goat breeding ( P = 0.033), more pronounced among women. None of the other potential risk factors we could explore showed an association with HL risk.

Table 2 Risk of Hodgkin's lymphoma in relation to socio-economic and environmental variables. Table 2 Environmental factors Women Men Total population Person-years No. Cases RR 95% CI Person-years No. Cases RR 95% CI Person-years No. Cases RR 95% CI Type of residence  Rural 10,517,022 236 1.00 Ref 10,495,652 280 1.00 Ref 21,012,674 516 1.00 Ref  Small towns 4,386,212 109 1.11 0.88–1.39 4,319,038 133 1.17 0.95–1.43 8,705,250 242 1.14 0.98–1.33  Urban 9,650,688 239 1.05 0.96–1.15 9,058,258 318 1.16 1.07–1.26 18,708,946 557 1.11 1.04–1.18 Deprivation index  First quintile 4,026,416 106 1.00 Ref 3,977,702 133 1.00 Ref 8,004,118 239 1.00 Ref  Second quintile 6,043,866 146 0.92 0.72–1.18 5,688,912 160 0.84 0.67–1.06 11,732,778 306 0.87 0.74–1.03  Third quintile 4,990,322 104 0.89 0.78–1.02 4,889,274 139 0.92 0.82–1.04 9,879,596 243 0.91 0.83– 0.99  Fourth quintile 5,013,282 122 0.97 0.89–1.06 4,856,370 173 1.02 0.95–1.10 9,869,652 295 1.00 0.94–1.06  Fifth quintile 4,480,036 106 0.99 0.93–1.05 4,460,690 126 0.96 0.90–1.02 8,940,726 232 0.97 0.92– 1.01 Distance from nearest hospital, km  ≤13 14,417,396 351 1.00 Ref 13,770,082 457 1.00 Ref 28,187,478 808 1.00 Ref  1–21 4,383,600 112 1.05 0.85–1.30 4,381,578 125 0.85 0.70–1.04 8,765,178 237 0.94 0.81–1.08  22–29 3,199,332 76 0.99 0.87–1.12 3,173,632 82 0.88 0.78–0.98 6,372,964 158 0.93 0.85– 1.01  ≥30 2,553,594 45 0.90 0.81–0.99 2,547,656 67 0.92 0.85–1.00 5,101,250 112 0.91* 0.85–0.97 Geological features  Quaternary/marine deposits 16,466,464 396 1.00 Ref 15,986,002 499 1.00 Ref 32,452,466 895 1.00 Ref  Basalt/effusive volcanic rocks 2,367,840 56 1.06 0.80–1.43 2,295,422 58 0.90 0.68–1.19 4,663,262 114 0.97 0.79–1.19  Granite/intrusive volcanic rocks 2,441,948 57 1.01 0.88–1.17 2,389,754 72 1.02 0.90–1.16 4,840,702 129 1.02 0.92– 1.12  Metamorphic rocks 3,224,712 73 0.99 0.91–1.08 3,141,244 101 0.99 0.91–1.08 6,365,956 174 1.01 0.96–1.07 Probability of indoor radon >300 Bq/m 3 , %  <5 5,031,970 110 1.00 Ref 4,694,092 151 1.00 Ref 9,726,062 261 1.00 Ref  5–9.9 6,721,068 167 1.28 0.98–1.67 6,587,320 204 1.06 0.84–1.33 13,308,388 371 1.15 0.97–1.37  10–19.9 7,499,446 178 1.09 0.95–1.25 7,389,862 178 1.06 0.94–1.19 14,889,308 393 1.07 0.98–1.17  20–29.9 3,331,288 79 1.06 0.95–1.18 3,256,930 109 1.02 0.94–1.12 6,588,218 188 1.04 0.97– 1.12  ≥30 1,970,150 50 1.11 1.00–1.23 1,944,744 52 0.98 0.89–1.08 3,914,894 102 1.04 0.97–1.11 Cattle per resident, n  None 16,540,922 402 1.00 Ref 16,125,972 483 1.00 Ref 32,666,894 885 1.00 Ref  0.03–0.10 4,742,288 104 0.85 0.68–1.07 4,536,312 161 1.08 0.89–1.30 9,278,600 265 0.97 0.84–1.13  >0.11 3,270,712 78 1.01 0.89–1.14 3,210,664 87 1.00 0.87–1.10 6,481,376 165 0.99 0.91–1.08 Sheep per resident, n  <1.43 15,200,900 370 1.00 Ref 18,009,736 591 1.00 Ref 33,210,636 961 1.00 Ref  1.43–3.349 4,619,162 98 0.91 0.71–1.18 4,598,648 129 1.00 0.80–1.26 9,217,810 227 0.96 0.81–1.14  3.35–6.05 2,928,986 67 1.01 0.87–1.18 2,897,032 74 0.98 0.85–1.13 5,826,018 141 1.00 0.90–1.10  >6.05 1,804,874 49 1.06 0.95–1.19 1,764,880 55 1.06 0.95–1.18 3,569,754 104 1.06 0.98–1.15 Goats per resident, n  <1.43 6,600,236 182 1.00 Ref 6,228,960 205 1.00 Ref 12,829,196 387 1.00 Ref  1.43–3.349 9,209,012 209 0.84 0.68–1.03 8,974,024 299 1.06 0.88–1.27 18,183,036 508 0.96 0.84–1.09  3.35–6.05 8,744,674 193 0.86 0.75–0.98 8,669,964 227 0.89 0.78–1.00 17,414,638 420 0.87** 0.79–0.96 Industries/mines  No 14,569,654 365 1.00 Ref 14,372,186 414 1.00 Ref 28,941,840 779 1.00 Ref  Yes 9,984,268 219 0.81 0.67–0.97 9,984,268 317 1.05 0.89–1.24 19,485,030 536 0.94 0.83–1.06 Cork harvesting  No 24,182,306 574 1.00 Ref 23,509,754 721 1.00 Ref 47,692,060 1295 1.00 Ref  Yes 371,616 10 1.11 0.60–2.08 363,194 10 0.85 0.45–1.58 734,810 20 0.96 0.62–1.50 Military settlements & shooting ranges  No 20,384,650 483 1.00 Ref 19,996,622 610 1.00 Ref 40,381,272 1093 1.00 Ref  Yes 4,169,272 101 0.97 0.77–1.23 3,876,326 121 0.89 0.72–1.09 8,045,598 222 0.92 0.79–1.08 2016 prevalence of multiple sclerosis, ×10 -5  217–277 6,829,580 176 1.00 Ref 6,645,756 213 1.00 Ref 13,475,336 389 1.00 Ref  319–388 14,276,442 324 0.89 0.74– 1.07 14,276,442 414 0.97 0.82–1.14 28,150,184 738 0.93 0.82–1.05  >388 3,447,900 84 0.98 0.85– 1.12 3,353,450 104 1.04 0.92–1.18 6,801,350 188 1.01 0.92– 1.11 Cumulative incidence of COVID-19 in 2020, ×10 -5  0.0–9.8 3,282,114 64 1.00 Ref 3,258,504 91 1.00 Ref 6,540,618 155 1.00 Ref  9.81–16.95 10,376,472 244 1.12 0.83–1.50 9,963,006 311 1.04 0.81–1.34 20,339,478 555 1.07 0.89–1.30  16.96–28.2 5,041,978 127 1.13 0.97–1.32 5,002,130 148 1.03 0.90–1.17 10,044,108 275 1.07 0.97–1.18  28.21–119.35 5,853,358 149 1.13 1.02–1.25 5,649,308 181 1.00 0.92–1.09 11,502,666 330 1.05 0.99–1.13 *, P < 0.001; **, P = 0.033. Abbreviations: CI, confidence interval; No. number; Ref, reference; RR, risk ratio.

Risk of Hodgkin's lymphoma in relation to socio-economic and environmental variables.

*, P < 0.001; **, P = 0.033.

Abbreviations: CI, confidence interval; No. number; Ref, reference; RR, risk ratio.

Discussion

Time trend in incidence

In 1974–2003, the standardized HL incidence in the Sardinian population was 2.6 per 100,000, lower than what was observed in northeastern Italy in 1970–1974 (4.33 per 100,000 per year among men and 3.70 among women) 41 and within the range reported by the European Cancer Registries. 7

The bimodal age distribution of the cases was confirmed. HL incidence showed a moderate upward trend among the total population and below age 45 but not at age ≥45 years, with no substantial difference from what was observed by the Italian Cancer Registries in the overlapping time frame. 42 The time trends were similar by sex. However, incidence below age 45 and at age ≥45 years were weakly correlated over the 356 Sardinian communes. Consistent with our report, increasing HL incidence rates have been described among young adults in the U.S.A., 43 the U.K., 6 Canada, 44 and the Middle East, 45 , 46 and attributed to the higher prevalence of the nodular sclerosis histological sub-type among the youngsters. 6 , 43 , 44 , 45 , 46

Spatial distribution of Hodgkin's lymphoma

Our spatial analysis detected a spatial aggregation of female HL cases in four central-western Sardinia bordering communes. Cases were for most part young adults and did not cluster in time. The lack of personal information in the data set we accessed did not allow us to inquire about a contribution from familial aggregations. An elevated probability of HL occurrence was also observed in a commune in the suburban area of the region's capital among subjects aged ≤44 years and in a small rural commune among subjects aged ≥45 years.. These communes were not among those showing an excess for other malignancies of the hemolymphopoietic system, 31 , 47 , 48 , 49 , 50 which would imply differences in the etiologic pattern, if not due to chance.

Environmental exposures and socio-economic factors

Studies from the InterLymph Consortium reported an increase in HL risk associated with the rs 1860661, a variant of the intron 2 of the TCF3 gene, which regulates B- and T-cell lineage commitment, 51 and, limited to EBV-positive subjects, the rs 6457715 polymorphism, 52 which is located near the MHC class-II HLA – DPB1 gene, indicating that gene-environment interactions are part of the complex, multifaceted HL etiology. In our study, we observed a modest excess risk of HL in urban areas and a decrease in risk by distance from the nearest hospital, a surrogate of proximity to urban areas, which might imply less likely exposure to transmissible agents through inter-personal contacts. However, an early Italian study did not observe changes in incidence between rural and urban areas nor by level of industrialization. 53

An excess risk of HL was reported in a volcanic area in Sicily. 17 We did not observe a relationship with the prevalent geology or the probability of an indoor α-radiation level above 300 Bq/m 3 , but Sardinia's geology differs substantially from Sicily, as indicated by the absence of volcanic areas and the exceptional rarity of earthquakes. Residence in proximity of industrial or military settlements or cork harvesting farms, a possible source of airborne mold spores, also did not affect risk.

Our results do not support the hypothesis of a geographic correlation between HL incidence and multiple sclerosis which was plausible because of overlapping genetic features, 9 , 13 and the observed geographic correlation between the two diseases. 15 Nor did we observe a link between the cumulative incidence of COVID-19 in the pre-vaccine months and past HL incidence, contrasting our previous report of a highly significant increase in COVID-19 related to the past incidence of non-Hodgkin's Lymphoma. 23

As it concerns livestock, risk of HL decreased with prevalence of goat-breeding, with a significant inverse trend (Wald test for trend: P = 0.033), matching a similar finding with Acute Lymphoblastic Leukemia. 50 We cannot exclude a chance finding also because of the opposite results of a previous report on livestock breeders, 18 which, however, did not inquire into specific types of livestock. Infection from Coxiella Burnetii , the agent causing Q fever, in domestic goats has been repeatedly reported in several countries. 54 To the best of our knowledge, the relationship between Q fever and HL or acute lymphoblastic leukaemia has not been formally evaluated thus far, while human studies did not find an association with risk of non-Hodgkin's Lymphoma. 55 , 56

Limitations

An important limitation in our study is that we conducted the analysis of an old database. Therefore, the time that has elapsed has made the potential impact in terms of Prevention and Public Health fade away. Still, this database offered the opportunity of exploring time trends and geographic spread of HL over three decades in a genetically peculiar population. 15 Also, thanks to this unique database, it was possible to confirm the age-related varying time trends of HL.

We did not have access to bio specimens and clinical and bio-molecular information to confirm whether any change in the prevalence of specific HL subtypes might account for the observed changing epidemiological features with age.

The association with urban residence might mask the special expertise of the local physician/s in detecting the disease and correctly addressing the patients to a nearby Hematology unit; if such expertise would have increased along the study period, detection bias might have generated spurious upward trends. However, the upward trend only concerned incident cases below age 45 but not those aged 45 years or more, which would exclude that such bias affected our results.

The information on the commune of residence at the time of diagnosis was missing for 2.0% of the cases; it doesn't seem plausible that this might have influenced the geographic patterns. Only information on the residence at diagnosis was available. As a substantial migration from rural to urban centers occurred among the Sardinian population from the 1950s onwards, we cannot exclude the possibility that the commune of residence at diagnosis differed from that linked the risk factors of interest. Such a residential bias might have created the observed excess risk associated with urban and suburban residence and the apparent protective effect of goat breeding but neither the cluster among the female population in the central-western part of the island nor the hot spot in the small rural center among the population aged ≥45 years.

Another limitation is due to the ecological study design, which identifies exposures at the population and not the individual level and does not allow us to confirm or reject the reportedly scarce relevance of environmental factors in HL etiology. 21 Using the ecological approach, spurious associations might emerge and/or true associations might not be revealed, resulting in the so-called ecological fallacy. 57 However, with recommended caution, ecological associations with widespread risk factors might suggest previously unexplored hypotheses 58 susceptible of further inquiry with carefully designed analytic studies.

Conclusions

Our results confirm age-related differences in the time trend of HL incidence, show a previously undetected cluster among young women of a central-western area of Sardinia, and suggest an excess risk in urban areas and an inverse association with goat breeding. Further in-depth investigation is warranted to verify whether the cluster persists as well as the nature of the detected associations.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests. Giorgio Broccia reports was provided by Cancer Centre Armando Businco. The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Author contributions

G.B., J.C., and P.C. conducted the conceptualization; G.B. was responsible for the data management and the validation of the dataset; J.C. and P.C. were responsible for the methodology, performed the formal analysis, and wrote the original manuscript; C.O—O. took care of the visualization; S.DeM. supervised the analysis; all the authors reviewed and edited the manuscript.

Footnotes

Given his role as Associate Editor in this journal, Pierluigi Cocco had no involvement in the peer-review of this article and has no access to information regarding its peer-review. Full responsibility for the editorial process for this article was delegated to Mei Wang.

References