The Geographical Pattern of Thyroid Cancer Mortality Between 1980 and 2009 in Italy

Abstract

Background: Mortality for thyroid cancer (TC) is low and has been decreasing worldwide; yet few population studies based on mortality have been conducted. Several nonradiation risk factors have been associated with TC, including residence in goiter-endemic areas (as an indicator of iodine deficiency). We used mortality data to perform a spatial-temporal analysis regarding TC in Italy and investigated the association between mortality and socioeconomic status and geographical features (residing in a mountainous area is a proxy for iodine deficiency).

Methods: We analyzed data from Italy's National Mortality Database (1980–2009). To evaluate temporal trends in mortality the age-standardized death rate (ASR) was used; to identify geographic areas with excess deaths due to TC standardized mortality rates (SMR) were calculated. We also calculated the rate ratios (RR) of the ASR and the 95% CI by sex. We performed a cluster analysis to identify municipalities with major departures from expected mortality, both in the entire study period and in two separate periods to evaluate the spatial-temporal variability. Finally, we evaluated the association between mortality and index of deprivation and altitude.

Results: There were 16,473 deaths due to TC (10,690 females, 5783 males). The mean ASR was unsurprisingly low (0.58/100.000). There was a trend of decrease in mortality throughout Italy (−42% for 2007–2009 vs. 1980–1984), more pronounced among women. The decrease was greater in the north. Four geographic clusters were identified when considering the entire study period, two in the north and two in the south; however, the clusters in northern Italy refer to the earlier period (1980–1994) and those in southern Italy to the later period (1995–2009). Mortality was associated with residing in a mountainous area. A slight association with high socioeconomic status was found.

Conclusions: This study reveals space-time differences in TC mortality in Italy. It shows an association between mortality and residing in mountainous areas, which is a proxy of iodine deficiency. The observed temporal north–south shift cannot be explained by socioeconomic differences, whereas the efficient prophylaxis program implemented in the 1980s in some areas of northern Italy can help to explain the disappearance of the clusters in those areas in the period 1995–2009.

Introduction

Thyroid cancer (TC) constitutes a group of extremely diverse malignancies. The vast majority are indolent tumors with low mortality, of which the most common is papillary carcinoma. However, some of these malignancies are lethal, with anaplastic carcinoma being the most aggressive type.

In the past 30 years, the incidence of TC has been increasing worldwide. By contrast, TC mortality is low and has been decreasing. In Europe, decreases have been reported in many countries, although the rates vary by individual country and very few studies have focused exclusively on TC mortality (1–3). Mortality varies greatly by sex as well, with higher rates among women (0.40/100,000, compared to 0.3/100,000 among men) (4).

The malignancies that comprise TC also differ in terms of risk factors. The best-known risk factor is exposure to ionizing radiation (therapeutic irradiation or environmental pollution), whereas nonradiation risk factors include history of goiter, inadequate or excess iodine intake, obesity, hormonal and reproductive factors, genetic factors (5), and socioeconomic status (6). Geographical area has been hypothesized to be a further risk factor for TC (7), as reported in a recent study conducted in Spain in which a nonuniform pattern of the geographic distribution of TC mortality was found (8). Areas with various degrees of iodine deficiency may contribute to explain these differences. In fact, the occurrence of nodular goiter is common in areas characterized by iodine deficiency, and a direct relationship between nodular thyroid diseases and TC has been confirmed by a number of papers published so far (9). Furthermore, more aggressive histotypes of TC are reported to more frequently occur in iodine-deficient areas than iodine-sufficient ones (10,11).

With specific regard to Italy, a high incidence of TC has been reported in the areas covered by the Italian Network of Cancer Registries (1991–2005), and these rates have been increasing (12). This increase has been attributed, in part, to improved health surveillance; however, this is almost exclusively the case for papillary carcinoma, which is rarely lethal and which shows wide geographical heterogeneity in Italy (13). Regarding the geographic distribution of TC in Italy, the Network of Cancer Registries only covers about 40% of the population, and most of the data are from northern areas; thus, to determine geographic distribution, only mortality data covering the entire country can be used.

The objective of the present study was to use national mortality data to contribute to the knowledge of the geographical distribution of TC in Italy, where the populations of some areas are still affected by iodine deficiency (14,15). The additional objectives were (i) to investigate whether TC mortality is associated with place of residence at death (municipality) and, in particular, with its altitude, considered as indirect indicators of residence in areas of iodine-deficiency/endemic goiter; and (ii) to identify potential clustering at the level of municipality. We also investigated the relationship between TC mortality and socioeconomic status, which is a known risk factor for TC and an indicator of access to health care, which could have an effect on mortality.

Materials and Methods

Deaths due to TC

The source of data on deaths due to TC was Italy's National Mortality Database (NMD), which is managed by the Statistics Unit of the Istituto Superiore di Sanità (National Institute of Public Health) and based on data provided by the National Institute of Statistics (ISTAT; www.istat.it/it). The NMD contains the underlying cause of death, which until 2003 was coded using the International Classification of Diseases, 9th revision (ICD-9; malignant neoplasm of thyroid gland; code 193); since then, ICD-10 has been used (code C73). In the NMD, data are available from 1980 to 2009; however, since the data referring to the years 2004 and 2005 were not codified by ISTAT, they were excluded from the analysis; therefore, the analysis comprised 28 years.

Indicators used for the analysis of mortality

Mortality rates were calculated using both a direct and an indirect method. The direct method was used for calculating the age-standardized death rate (ASR), which was used to evaluate trends in mortality over time, both nationwide and for Italy's two main geographic areas (northern/central Italy, and southern Italy and the main islands [Sicilia and Sardegna]), as well as by sex. The ASR was calculated for each year in the study period; to make the calculations more robust, it was also calculated for 3-year periods (the year 2009 was not included in these calculations because it would have resulted in a final 4-year period).

The indirect method was used to calculate the standardized mortality ratio (SMR), which we used to identify geographic areas with excess deaths due to TC (first regions and then municipalities within these regions). The SMR calculated in 5-year periods is the ratio of the number of observed deaths to the number of expected deaths. We calculated the SMR for each of Italy's 20 regions and 8094 municipalities. When calculating the SMR for each region, the national population was used as reference, whereas when calculating the SMR for the municipalities, the regional SMR was used. The SMR and their 95% confidence intervals [95% CI] were estimated either based on a Poisson distribution (if there were fewer than 100 observed deaths) or using the Byar method (if there were more than 100 observed deaths).

Using a Poisson distribution, we also calculated the rate ratios (RR) of the ASR and the 95% CI for Italy's two main geographic areas (northern/central Italy, and southern Italy and the main islands) (16), using as a basis the three initial years of the study period (1980–1982).

Cluster analysis

Because deaths due to TC are rare, the expected mortality in many municipalities would be very low; to determine whether there exist concentrations of disease related to explainable geographic risk factors we performed a municipal cluster analysis, to identify the areas with major departures from expected mortality. This analysis was first performed considering the entire 28-year study period then, to evaluate the temporal variability of the clusters, we performed the analysis for two separate periods: 1980–1994 and 1995–2009. The analysis was performed according to the procedure Spatial Scan Statistics (17), using SatScan software (version 6). The procedure employs a circular or elliptical window of varying radius, from zero to some upper limit, which moves on the entire study area, centered at each step on one of the municipalities, identified by the x, y coordinates of the municipality's town hall. The method creates an infinite number of distinct geographical circles with different sets of neighboring data locations within them: each circle is a possible candidate for a cluster. Under the null hypothesis, the observed number of cases follows a uniform distribution, so that the expected number of cases in an area is proportional to its population size. The relative risk is the estimated risk within the cluster divided by the estimated risk outside of the cluster.

Association between TC mortality and altitude and socioeconomic status

We evaluated the association between TC mortality and indices of deprivation and altitude, available at municipality level. The index of deprivation, calculated using data from the 2001 Census, was based on five conditions that describe the multidimensional nature of social and material deprivation: (i) low level of education, (ii) being unemployed, (iii) not being a home owner, (iv) living in a single-parent household, and (v) living in a municipality with a high population density (18).

The index of altitude is based on a district statistical system created by ISTAT in 1958, which divides Italian towns into: (i) hill zones (altitude between 300 and 600 m), which are divided into inland and coastal areas; (ii) mountain areas (altitude >600 m), divided into inland and coastal areas; and (iii) lowland areas (altitude ≤300 m). The relationships between TC mortality and the two indices (deprivation and altitude) were evaluated using the standardized rate ratio (SRR) by level of index (19).

Results

Between 1980 and 2009, there were 16,473 deaths due to TC in Italy, with a mean number of approximately 600 deaths per year. Of these deaths, 10,690 (64.9%) were among women and 5783 among men. The median age of death was 75 years: 0.1% of the deaths occurred before the age of 20; 0.7% between 20 and 34; 2.3% between 35 and 44; 7.5% between 45 and 54; 18.3% between 55 and 64; 29.9% between 65 and 74; 30.8% between 75 and 84; and 10.3% after the age of 84. The median age range at death differed significantly by sex (65–69 years for men compared with 75–79 years for women).

Rates and trends

The mean ASR for the entire period was 0.58 per 100,000 [95% CI: 0.53–0.59], 0.49 for males and 0.64 for females. The mean ASR decreased from 0.69 in the period 1980–1982 [95% CI: 0.66–0.73] to 0.41 in 2006–2008 [95% CI: 0.39–0.43]. However, the decrease only began to be significant in 1995 (Fig. 1); the decrease in the RR was 42% (RR=0.58 [95% CI: 0.53–0.65]). When stratifying the ASR by the two main geographic areas (northern–central Italy and southern Italy, including the main islands), the decrease was observed both in northern–central Italy and southern Italy and main islands (Fig. 1). With regard to sex, the decrease in these two periods was greater for females (47%) compared with males (29%; data not shown); moreover, the decrease began earlier for females (i.e., in 1992–1994 vs. 1995–1997 for males). When observing the RR in the two main geographic areas (northern–central Italy and southern Italy, including the main islands), a significant trend of decrease was seen in the study period only in northern–central Italy (Fig. 2).

FIG. 1.

Age-standardized death rate (ASR) for thyroid cancer trends (International Classification of Diseases 9th Revision code: 193, 10th Revision code: C73). World standard population, period 1980–2009 (without the years 2004–2005).

FIG. 2.

Rate ratio (RR) of ASR for thyroid cancer trends (baseline 1980), periods 1980–2009 (without the years 2004–2005).

Geographic trends and risks

When considering the entire 28-year period, the SMR showed significant excesses in the northern regions of Piemonte and Trentino-Alto Adige/Südtirol (Bolzano and Trento) and in the southern region of Calabria; deficits were observed in the central regions of Toscana, Marche, Lazio, and Abruzzi and in the southern regions of Puglia and Sardegna (Table 1).

Table 1.

Mortality for Each of Italy's 20 Regions Considering 5-Year Periods

Region	Observed	SMR 1980–2009	SMR 1980–1984	SMR 1985–1989	SMR 1990–1994	SMR 1995–1999	SMR 2000–2006	SMR 2007–2009
Northern Italy
Piemonte	1591	113 [107–118]	123 [109–137]	113 [101–127]	116 [104–130]	104 [92–118]	103 [91–117]	112 [95–131]
Valle d'Aosta	36	102 [71–141]	66 [18–169]	155 [74–285]	103 [41–212]	176 [88–314]	33 [4–118]	54 [7–195]
Lombardia	2637	104 [100–108]	104 [95–114]	112 [103–122]	100 [91–109]	103 [93–112]	98 [89–108]	109 [97–122]
Bolzano	204	180 [156–206]	183 [127–254]	177 [124–246]	252 [189–329]	190 [134–260]	126 [82–187]	131 [75–212]
Trento	190	139 [120–160]	197 [145–263]	172 [125–232]	107 [71–155]	109 [71–159]	111 [72–163]	139 [85–215]
Veneto	1274	100 [94–105]	102 [89–116]	98 [86–112]	103 [90–116]	95 [83–109]	106 [93–120]	93 [78–111]
Friuli-Venezia Giulia	442	107 [97–118]	143 [117–172]	91 [71–115]	105 [84–130]	96 [75–122]	94 [73–120]	107 [78–145]
Liguria	645	101 [93–109]	115 [97–136]	99 [82–118]	105 [87–124]	92 [75–112]	86 [69–106]	93 [70–121]
Emilia-Romagna	1345	98 [92–103]	99 [87–113]	100 [88–113]	92 [81–105]	98 [86–111]	103 [91–117]	84 [70–101]
Central Italy
Toscana	1152	93 [88–99]	77 [66–90]	89 [78–102]	91 [80–104]	94 [82–108]	103 [90–118]	109 [91–129]
Umbria	279	99 [88–112]	106 [79–139]	97 [72–128]	88 [65–117]	88 [64–118]	97 [71–128]	133 [95–182]
Marche	391	83 [75–91]	81 [62–103]	80 [62–101]	73 [57–93]	91 [72–114]	90 [71–113]	79 [46–108]
Lazio	1326	94 [89–99]	98 [86–111]	96 [84–108]	96 [85–109]	93 [81–106]	93 [82–106]	86 [72–101]
Southern Italy
Abruzzo	315	81 [73–91]	79 [59–104]	93 [72–118]	72 [54–94]	89 [68–115]	73 [54–97]	79 [54–112]
Molise	95	92 [74–112]	133 [86–196]	81 [46–132]	88 [52–139]	77 [42–129]	82 [45–137]	79 [34–156]
Campania	1294	103 [98–109]	93 [80–107]	101 [88–115]	102 [89–115]	112 [99–127]	108 [94–122]	110 [93–129]
Puglia	823	85 [79–90]	79 [66–94]	79 [66–94]	82 [69–96]	89 [76–104]	100 [85–116]	79 [63–98]
Basilicata	167	101 [86–118]	85 [55–126]	88 [58–128]	124 [88–169]	76 [48–116]	117 [81–165]	124 [77–190]
Calabria	571	109 [100–118]	96 [77–118]	92 [74–113]	111 [92–134]	121 [100–146]	104 [84–127]	143 [113–179]
Main Islands
Sicilia	1344	104 [98–110]	86 [75–99]	105 [92–119]	111 [98–125]	106 [93–120]	113 [100–128]	100 [84–119]
Sardegna	352	86 [77–95]	96 [74–119]	77 [58–99]	94 [74–119]	102 [80–128]	73 [55–96]	72 [49–102]
Entire country (Italy)	16,473	100	100	100	100	100	100	100

Standardized mortality ratios (SMRs) greater than 100 represent excess deaths above expected and those less than 100 are lower than expected. Numbers in brackets are 95% confidence intervals; those including 100 are not significant.

When comparing the 5-year periods (Table 1), in northern Italy, the region of Piemonte showed significantly excesses before 1995; after 1995, there were still excesses, yet they were not significant. Bolzano consistently showed very high risks, though they were not significant after 2000. Trento also showed consistently high risks; however, they were not significant beginning in 1990. The remaining northern regions did not differ from the mean values, with the exception of Lombardia in 1985–1989 and Friuli in 1980–1984. In central Italy, there were no excesses in mortality; in all of the 5-year periods, all of the central regions had values close to or below the national average (the deficits were found in Tuscany 1980–1984 and in Marche in 1990–1994). In southern Italy and the islands, the situation was more varied. In the regions of Calabria and Basilicata, there were increased risks in all of the 5-year periods, though they were significant only in Calabria in the period 2007–2009. Puglia was the only southern region that showed consistently a significant deficit with respect to the national average. Sardegna showed significant deficits only in 1985–1989 and 2000–2006 (without the years 2004–2005), whereas Campania showed no differences with respect to the national average.

The analysis of risks at the municipality level showed significant excesses in 132 of the 8094 municipalities. Of these 132 municipalities, 70% are located in northern–central Italy and over 40% are located in inland mountain areas.

Cluster analysis

When considering the entire study period, we identified four geographic clusters of TC mortality (Fig. 3A). Two clusters were in northern Italy: one in the Trentino-Alto Adige/Südtirol (Bolzano and Trento; Cluster 1, consisting of 338 municipalities) and the other in Piemonte (Cluster 2, 211 municipalities). The other two clusters were in southern Italy: one in Sicilia (Cluster 3, 97 municipalities) and the other in Calabria (Cluster 4, 83 municipalities). Most of the municipalities (73%) in the four clusters were located in mountain areas. Whereas the cluster in Trentino-Alto Adige/Südtirol (Bolzano and Trento) was homogeneous in terms of altitude (i.e., the municipalities were all located in inland mountain areas), the municipalities in the cluster in Sicilia were located at different altitudes.

FIG. 3.

Mortality from thyroid cancer. (A) Entire period (1980–2009). Significant cluster results and distribution of index of altitude into the clusters, overall. (B) First period (1980–1994). Significant cluster results. (C) Last period (1995–2009). Significant cluster results.

When separating in two 14-year periods (i.e., 1980–1994 and 1995–2009), we identified two clusters in northern Italy in the earlier period, in the same areas as the analysis on the entire study period (Fig. 3B: Cluster 1, in Trentino-Alto Adige/Südtirol [Bolzano and Trento], and Cluster 2, in Piemonte). In the later period, although there was a small cluster in Piemonte (Cluster 2, consisting of three municipalities and only three observed deaths), we found a large cluster in southern Italy (Fig. 3C) involving mainly Calabria and Sicilia (Cluster 1, consisting of 1147 municipalities).

Associations with altitude and socioeconomic status

The SRR by altitude (Table 2) shows that the risk of death was significantly higher in mountain areas, regardless of whether they were inland or coastal areas, and significantly lower in hill areas and lowlands. No excesses in risk were found for the categories of the deprivation index (Table 3), with the exception of the category “very wealthy,” in which there was slight yet significant excess in mortality (SRR=1.06 [CI 95% 1.01–1.12]).

Table 2.

Risk of Death for Thyroid Cancer Mortality in Italy by Index of Altitude

Index of altitude	Deaths	%	SRRa	95% CI
Inland mountain area	2234	13.6	1	—
Coastal mountain area	543	3.4	0.92	[0.83–1.01]
Inland hill area	3828	23.3	0.81	[0.77–0.86]
Coastal hill area	2373	14.4	0.82	[0.77–0.87]
Lowland area	7129	43.3	0.83	[0.79–0.87]
	16,473	100

^aSRR lower than the unit means lower risk of death respect to baseline (inland mountain area).

SRR, standardized rate ratio; CI, confidence interval.

Table 3.

Risk of Death for Thyroid Cancer Mortality in Italy by Index of Deprivation

Index of deprivation	Deaths	%	SRRa	95% CI
Much deprived	3109	19.3	1	—
Deprived	3420	21.2	1.02	[0.97–1.06]
Medium	3636	22.6	1.03	[0.99–1.08]
Rich	3276	20.3	1.02	[0.98–1.07]
Very rich	2663	16.5	1.06	[1.01–1.12]
	16,104	100

^aSRR higher than the unit means higher risk of death with respect to baseline (much deprived).

Discussion

In this study, the large number of deaths due to TC (more than 16,000) and the long period considered (28 years) allowed us to obtain robust results. The decreasing trend in mortality in our country is consistent with data from the rest of Europe and, specifically, TC patients in Italy (4,21). As expected, the mean ASR for the entire period was low (0.58/100,000; reference: world population) and similar to that reported in the United States (0.5 in 1973–2002) (22). With regard to the last 3-year period (i.e., 2006–2008), the ASR of 0.41 is similar to that reported for 2000–2003 in Italy in another study (0.39) (4).

Most of the deaths occurred among the elderly for both sexes. In Italy, the median age at death due to TC (75 years) is similar to but slightly higher than that reported by SEER for the period 2005–2009 (73 years) (23). Although the number of deaths is higher among women, it does not imply that women are more likely to die of TC than men. The slightly higher mortality rates in women compared to men (female to male ratio: 1.3 [0.64 vs. 0.49]) is in agreement with higher incidence rates among women (female to male ratio: 3.0, 1991–2005) (12). In terms of mortality, however, the diversity between men and women is attenuated. Death occurred 10 years earlier among men, compared with women. The trend of decrease was significant for both sexes, although the decrease occurred earlier and was more accentuated among women. Other studies have reported results that are similar to ours for women (24–26) yet different for men (8,25,26). The decrease in mortality among women has been attributed to increasing diagnostic scrutiny (24); thus, the slighter (24) decrease among men could in part be explained by lesser access for men to analogous clinical controls. The earlier death among men could also be attributed to men having more advanced (and more aggressive) disease at the time of diagnosis (26).

The regions in which we found significant excess in mortality (i.e., Piemonte and Trentino-Alto Adige/Südtirol [Bolzano and Trento] in northern Italy and Calabria in southern Italy) are in areas that in the past were known to be iodine-deficient (and thus endemic for goiter) (27–29), consistently with reports from other geographic areas (30).

Residence in areas that are endemic for goiter is known to be a risk factor for TC (7). Long-term residence in endemic areas has been associated with an increased risk of TC in northern Italy (31,32) and in Switzerland (33), where there was a high prevalence of endemic areas in the past. In a study in which data from 12 case–control studies conducted in the United States, Asia, and Europe were pooled, a history of goiter was associated with an excess risk of nonmedullary TC (9). In Spain, TC mortality has been shown to be higher in areas that had been traditionally endemic for goiter (8). Moreover, a striking difference in the relative prevalence in the TC histotypes was reported in two adjacent areas of Sicily with different iodine intake indicating that even moderate iodine deficiency is associated with an increased frequency of more aggressive histological subtypes of TC (34). However, it has been observed that after iodine prophylaxis a clear relationship between increased iodine nutrition and elevation of the papillary TC/follicular TC ratio is evident, as observed in Switzerland (33), Germany (35), and Austria (36). This has occurred even in modest increases in iodine urinary excretion. Also, anaplastic TC is reported to be reduced after years of iodine supplementation (37–39).

The degree of the decrease in mortality varied by geographic area. In particular, the decrease was much greater in northern Italy than in southern Italy, in terms of the ASR (Fig. 1), the RR for the two main geographic areas (Fig. 2), and the regional SMR (Table 1). The division of the country in two main geographic areas was not established from a thyroid disease point of view or for demographic reasons but it reflects social, cultural, and economic differences that are well known in Italy. Although regions with excesses were found in both areas (Fig. 3A), the clusters showed a shift from the north to the south over time (Figs. 3B and 3C). However, the cluster in Piemonte in the second period consists of only three cases.

The reported adequate iodine intake in some areas of northern regions can help to explain the disappearance of the clusters in those areas in the period 1995–2009. Specifically, since 1982 an efficient province-wide iodine prophylaxis program has been active in the autonomous province of Bolzano (40,41); whereas, in Piemonte no active prophylaxis had been carried out in the last 30 years until 2005, when a nationwide program of iodine prophylaxis with salt iodization on voluntary basis was implemented in Italy (41). Piemonte has been described as a moderately iodine-deficient region since the 1970s (42). However, over the years the Piemonte population has benefited from a silent iodine supplementation especially in rural/mountain areas (27,43) due to improved nutritional conditions, the presence of iodine residues remaining in milk from disinfectant agents (42) (iodophors) used in dairying, the increase in the use of industrially produced food, and to a widespread awareness among population of the risk of iodine-deficiency disorders. This has led to a condition of iodine sufficiency in this region (27,40). In contrast, recent studies on urinary iodine concentration and goiter prevalence in schoolchildren (as indicators of the severity of iodine deficiency) in some areas of southern Italy have demonstrated the persistence of a high prevalence of goiter despite improvement in urinary iodine concentration (14,15). Moreover, the Italian National Observatory for Monitoring of Iodine Prophylaxis in Italy (OSNAMI) recently reported that only 53% of consumed salt in Italy is iodized (44). This implies that a large part of Italy's population is still affected by mild iodine deficiency.

Another possible explanation for the greater decrease in mortality in northern Italy could be that a gap exists in the quality of healthcare services between the north and the south, with the north having a greater concentration of centers of excellence for diagnosis and treatment (Working group ERA, www.e-r-a.it). Moreover, obesity, which is a risk factor for TC (45), is more prevalent in southern Italy than in northern Italy (www.epicentro.iss.it/passi/dati/sovrappeso.asp [accessed October 4, 2013]). However, it cannot explain the lower mortality for TC in the southern region of Puglia for the entire study period. Interestingly, Puglia is surrounded by the sea and is almost entirely flat.

To the best of our knowledge, our study is the first to have measured the association between TC and altitude. The significant association between TC mortality and inland mountain areas is intriguing. The results of the cluster analysis also showed this association, in that 73% of the municipalities in the four clusters are inland mountain areas, which had the greatest risk with respect to other altitudes. The variable “Inland Mountain” is an indicator of low-in-iodine territories. It is well known that geographic areas in which the population is affected by iodine-deficient disorders are usually mountainous. In Europe, endemic goiter was extensively reported in the past, particularly in isolated mountainous areas in Austria, Bulgaria, Croatia, France, Italy, Spain, and Switzerland (46). In Italy, iodine deficiency has also been reported mainly in mountainous areas (47). In the study by Lise et al. (13), in Italian Cancer Registries located near the Alps, some of the lowest TC incidence rates have been reported for papillary TC and the highest for anaplastic TC. In our study, it was not possible to evaluate the geographic distribution of the histological types of TC because the information was not available in the mortality database. However, we evaluated individual death certificates for the most recent available data (6 years; data not shown). The histological subtype of TC was not specified in the 79.3% of death certificates; when it was specified, anaplastic/undifferentiated carcinoma represented 48.8%. This distribution was identical in the 6 years examined (Kruskal Wallis test, p=0.443).

However, we hypothesize that the geographic differences in incidence observed by Lise et al. (13) and the geographic pattern in TC mortality in our study are attributable to the fact that the types of TC that show up in mortality data are lethal types (aggressive histological variants), whereas the cases of TC seen with incidence studies consist mainly of papillary cancer (nonlethal TC).

In light of reports of an association between the incidence of TC and ionizing radiation (48), TC mortality in Italy does not seem to reflect the 1986 Chernobyl accident (49) and the location of the storage of radioactive material in Italy (data not shown) (www.enea.it).

Our finding of a slight positive association between TC mortality and high socioeconomic status cannot be explained by greater access to diagnostic facilities, as found for incidence (6,50). However, it could be related to the fact that among individuals belonging to disadvantaged population groups, the most frequent causes of death are those related to lifestyle risk factors and low socioeconomic status (51).

Some potential limits of our study need to be mentioned. Mortality is not the best indicator of TC, since the survival rate is high for most types of TC. Furthermore, unlike studies based on data from cancer registries, mortality data do not provide detailed information on histological subtypes, and these data reflect mainly the most aggressive types of TC. However, given that there is no total coverage in Italy of cancer registration (the Network of Cancer Registries covers about 40% of the country's population; www.registri-tumori.it/cms/), mortality data constitute a source of important information on TC. We used residence at death as a proxy of the residence at diagnosis, taking into account that very limited migration is seen in Italy (the residential history of the patient is unknown in the death certificate). Regarding ascertainment bias, we are reasonably sure that there were no false positives. However, we cannot exclude the possibility of false negatives, a limitation that is inherent in death certificates.

To conclude, this is one of the few population-based investigations of mortality due to TC; the study include a large number of cases (total of 16,000 deaths), cover an extensive period (three decades), and refer to Italy's entire population.

Disclosure Statement

The authors have no conflicts of interest to declare.