Ragweed Pollen Concentration Predict Seasonal Oculo-rhinitis and Asthma Severity in Subjects Allergic to Ragweed.

Objectives: 1. To investigate the correlation between ragweed pollen concentration and conjunctival, nasal and asthma symptoms severity in patients allergic to ragweed using ambient pollen exposure in the Milan area during the 2014 ragweed season; 2. to calculate the pollen / symptom thresholds and 3. to assess the effectiveness of ragweed Allergen Immuno Therapy (AIT). Patients: 66 subjects allergic to Amb a 1 enrolled in the study and were divided into two cohorts: AIT treated (24) and non-AIT treated (42). Measurements: Pollen counts and daily symptom/medication patient diaries. Autoregressive Distributed Lag Models were used to develop predictive models of daily symptoms and to evaluate the short-term effects of temporal variations in pollen concentration on the onset of symptoms. Results: We found signicant correlations between ragweed pollen load and the intensity of symptoms, for all three symptom categories respectively, both in non-AIT treated ( 𝛕 = 0.341, 0.352, 0.721 and ρ = 0.48, 0.432, 0.881, p-value < 0.001) and in AIT treated patients (O= 0.46, 0.610, 0.66 and ρ = 0.692, 0.805, 0.824; p-value < 0.001). In both cohorts, we observed a positive correlation between the number of symptoms reported and drug use. Mean symptom levels were signicantly greater in non-AIT treated than in AIT treated patients (p < 0.001) for all symptom categories. Pollen concentration thresholds for three symptom severity levels were calculated. Conclusions: Ragweed pollen concentration is predictive of symptom severity in ragweed (Amb a 1) allergy patients. AIT treated patients had signicantly reduced mean symptom levels compared to non-AIT patients. with a high prevalence of seasonal conjunctival, nasal and bronchial symptoms due to sensitization to ragweed. Therefore, our results can be generalized in situations with similar contexts. We strongly believe that our ndingsthat ragweed hay fever symptoms are positively correlated with ragweed pollen concentration, are generalizable to many areas of the world where there are high concentrations of airborne ragweed pollen. Furthermore, our results have signicant implications for the metropolitan area of Milan, as they conrm the validity of the efforts made to contain the expansion of the ragweed plant in the territory.


Introduction
Ambrosia artemisifolia (Common Ragweed) is an invasive plant whose highly allergenic pollen causes seasonal respiratory allergy in people in many countries around the world (1)(2)(3)(4). In particular, the North-West metropolitan area bordering the city of Milan (Italy) is one of the most infested areas with ragweed in Europe (5)(6)(7). Currently, the Agenzia di Tutela della Salute (ATS) is the regional agency responsible for healthcare management in the Metropolitan Area of Milan. Inthe North-West area it comprises two districts: ASST Ovest Milanese and ASST Rhodense (Fig. 1). Since the 90s, the ragweed plant has spread enormously in this area, becoming the leading cause of oculorhinitis and asthma in the summer-fall seasons. For example, an epidemiological survey in the municipality of Magenta, found that the prevalence of cases of ragweed oculorhinitis increased in the general population from 9.2-14.00% in the period from 1996 to 2005, and in the same period the prevalence of ragweed asthma increased by more than 40% (8). Furthermore, in 2012 a survey by the ATS Health Agency, involving the Allergy Units in the two districts, found that 54% of the patients visited for the rst time for oculorhinitis, and 38% of those hospitalized for asthma, were allergic to ragweed (9). Given the high prevalence of ragweed allergy sufferers in this area, it is important to achieve an adequate control of oculorhinitis and asthma symptoms, as well as to limit the social and psychological consequences of the allergy during ragweed season. In fact, ragweed allergy in this area is a major limiting factor for people in schools andat work, affecting their learning ability. Consequently, since the 90s, a series of primary prevention measures have been carried out by the Province of Milan and the Lombardy Region in collaboration with the local Health Agency (today ATS). More recently, these preventive measures have become part of the COST SMARTER action (Sustainable management of Ambrosia artemisiifolia in Europe) (10). In practice, the following measures have been carried out in recent years: i) control of the territory, e.g. through aerobiological monitoring, surveillance and monitoring of the infested area, ii) urban planning, iii) epidemiological studies, and iv) studies on the effectiveness of various methodologies to limit the spread of ragweed, i.e.: mowing the weeds before owering, covering the land, plowing the soil, harrowing discs and chemical control (11)(12)(13)(14)(15).
Furthermore, since 2013 this area has been infested with a beetle (Ophraella communa) (16)(17)(18)(19)(20), which feeds preferably on ragweed weeds, causing them to dry out and die. As a result of these preventive measures and the beetle infestation, in the years following 2013 there has been a reduction in both the number of ragweed plants and the levels of airborne ragweed pollen (21).
However, reducing the concentration of pollen in an area, does not make the conclusion certain that there will be a reduction in the allergy symptoms in the residents of that area. More speci cally, concerning the north-western area of the Metropolitan Area of Milan, to our knowledge,there are no published studies on the correlation between the concentration of airborne ragweed pollen and the severity of allergy symptoms. Indeed, in other countries only a few studies have been carried out on this topic and their results are con icting: some have found a positive correlation between the concentration of ragweed pollen and the severity of both oculorhinitis and asthma (22)(23)(24)(25), while others have not found any correlations (26-28), and two studies have even found an inverse correlation between ragweed pollen concentration and symptom severity (29,30).
Given the ambiguity of these ndings, we planned a cohort study investigating the correlation between the concentration of ragweed pollen and the severity of seasonal oculorhinitis and asthma in subjects allergic to ragweed. The second aim of our study was to determine theconcentrations of ragweed pollen to be considered as thresholds for the onset and worsening of symptoms. Finally, our third aim was to evaluate the effectiveness of AIT in reducing the severity of ragweed allergy symptoms.

Study design
The primary goal of this study was to investigate the association between exposure to airborne ragweed pollen and daily ocular, nasal and asthmatic symptoms in two sub-cohorts of patients throughout one whole ragweed season. We followed individuals sensitized to ragweed and suffering from seasonal ragweed oculo-rhinitis with/or without asthma, from July 16 to September 15, 2014. Study subjects weredivided in two cohorts: one consisted of individuals who had never received a ragweed AIT, hence "non-AIT",while subjects in the other cohort had been treated with ragweed AIT, hence "AIT treated". Patients were treated with AIT prior to, and independent of, their enrollment in the study. These patients received AIT treatment either in the same year as the study or in the three years immediately preceding the study.
Originally, we actually envisioned a different study design. Ragweed and mugwort coexist in the study area, and since mugwort and ragweed blooms are partly overlapping, at rstwe also wanted to investigate the in uence of mugwort exposure on allergy symptoms attributed to ragweed. However, since the number of patients with dual sensitization (i.e. ragweed and mugwort) was very small (n=25), the original study design was modi ed to include only subjects sensitized to ragweed.
From July 16, 2014 to September 15, 2014, all patients lled in a daily Clinical Diary of Symptoms and Drugs (CDSD). During the same period, pollen counts were measured from three pollen traps located in the study area (see the 'Pollen concentration' section for exact locations). The daily averages of ragweed and mugwort pollen concentration obtained from the three pollen traps were used for statistical calculations against the daily mean of the symptom/drug scores for each of the two sub-cohorts.

Setting
Patients were recruited and followed in the areas of the ASST Ovest Milanese and ASST Rhodense Districts, both of which are under the milanese health protection agency ATS (Fig. 1 Units was located in proximity of one of the three pollen traps (i.e. min-max distance: 0.2 -19.6 km) (Fig. 1).

Participants
Participants in the study were a random sample of citizens residing in the designated area, with a con rmed diagnosis of ragweed seasonal rhinoconjunctivitis (with or without asthma), some of which were treated with AIT. 66 subjects (male: 32, female: 34) were enrolled in the study. The Eligibility Criteria to participate in the study were: 1) subjects were sensitized to ragweed pollen, demonstrated by a positive SPT with commercial ragweed extracts (Lofarma SpA, Milan; ALK-Abellò SpA, Milan) and IgEs for the total ragweed extract and for rAmb a 1 (pectate lysase) (Immuno -CAP Thermo Fisher Scienti c Inc., Monza, Italy). In case of discordant test results, ragweed sensitization was con rmed based on positive IgEs for rAmb a 1. 2) Subjects had an established and documented history of clinical manifestations of ragweed allergy (i.e. symptoms of oculo-rhinitis, associated or not with asthma), coinciding with the ragweed owering period, in the years immediately preceding enrollment. 3) Subjects had given their informed consent. 4) Subjects were able to adhere to the study protocol, and 5) subjects would remain in their residence throughout the observation period.
At the end of the recruitment period, from February 01, 2014 to July 15, 2014, 71 subjects were assessed to the Eligibility Criteria. One subject was excluded because he did not perform the IgE test and four because the IgEs for rAmb a 1 were negative. The remaining 66 patients were con rmed eligible and were included in the study. The mean age was 36.5 years (range: 8 -69 years). None of the subjects were pregnant, nor suffering from chronic diseases. All subjects resided in the study area, and in the period from February 01, 2014 to July 15, 2014, they had an allergic visit for respiratory symptoms in one of the ve Allergy Units participating in the study. Moreover, 42 subjects had never been treated with AIT for ragweed, while 24 subjects had previously received ragweed AIT.Thus, patients were divided into two sub-cohorts: non-AIT and AIT treated. All recruited subjects (100%) completed the follow-up and were subjected to statistical analysis. The characteristics of the study base at enrollment are shown in Table 1. Ragweed and mugwort's pollenwas sampled daily from three Hirst-type pollen traps located respectively in Legnano (Lat 45° 35' 44" N, Long 8° 55' 23" E), Magenta (Lat 45° 28' 16" N, Long 8° 53' 33" E) and Rho (Lat 45° 32' 51" N, Long 9° 02' 42" E). The Hirst volumetric trap continuously draws 10 liters of air per minute onto an adhesive coated tape. Particles in the air stick to the tape, which moves at 2mm per hour to provide a time sample. The pollen collected from the traps was identi ed and quanti ed by a specialized technician and then correlated to the average air volume over 24 hours. The reference standard adopted for the sampling and counting of pollen was the UNI 11108 of 2004, valid at the time of the study.
The following de nitions were adopted from the Quality Control Working  All study participants were asked to complete a CDSD during the follow up period. Patients were asked to record their daily symptoms and medication use every evening from July 16th to September 15th, 2014. Each symptom was rated on a 4-point scale: 0: no symptoms; 1: mild symptoms; 2: moderate symptoms; 3: severe symptoms. These ratings were used to score each one of the following 9 distinct symptoms: Nasal: 1. nasal congestion/nasal di culty of breathing, 2. runny nose, 3. itchy nose, 4. itchy throat/ears; Ocular: 5. itching and/or burning in the eyes, 6. tearing and/or wet eyes; Bronchial: 7. cough, 8. breathing di culty while moving, 9. breathing di culty at rest. Therefore, the total symptom score ranged from 0 to 24, while the nasal score ranged from 0 to 9, the ocular from 0 to 6 and the bronchial from 0 to 9. Furthermore, patients were also asked to record every single unit of drug taken, i.e. number of tablets/day of oral antihistamines, number of spray/day of inhaled topical nasal corticosteroids or topical bronchial corticosteroids or topical bronchial corticosteroids + Long Acting Bronchodilator Agents (LABA) or + Short Acting Bronchodilator Agents (SABA), and number of drops/day of antihistamine eye drops.
Patient follow up during the exposure period All patients enrolled in the study went to their reference Allergy Unit immediately before the exposure period began. During the visit, their clinical history was collected, a medical examination was performed and patients gave their informed consent for participation in the study. All patients were also provided with a CDSD, which was returned compiled for all dates during the second visit after September 15, 2014.
Comparability of evaluation methods.
Patients in both cohorts, non-AIT and AIT treated, were recruited for the study in the same way, underwent the same diagnostic tests, received the same CDSD, were visited in the follow up on the same dates, and received the same Informative Consent.

Bias
The known overlap in the initial time period of ragweed MPS with mugwort MPS was a possible cause of bias with respect to the symptom score attributed to ragweed, given that the symptoms possibly caused by exposure to mugwort pollen are indistinguishable from those caused by ragweed pollen in subjects sensitive to both pollen species. However, the concentration of mugwort pollen detected during the study period was very low (see below) and therefore, unlikely to have in uenced in any signi cant way the respiratory symptoms of those patients also allergic to mugwort.
Other possible causes of bias on the symptom/drug score may have been: climatic variables, air pollution, as well as any intercurrent respiratory infectious diseases (which were not considered in orderto simplify the analyses).

Ethics
The study protocol complies with the Declaration of Helsinki and all its subsequent amendments of Tokyo 1975, Venice, 1983, Hong Kong, 1989, as well as with the current regulations for good clinical practice (GCP). The protocol also complies with all national and community regulations applicable to observational studies and all ethical and deontological principles that inspire the medical profession. The study was approved by the Ethics Committee of Milan Area C (No 80_122013).
Only ragweed pollen concentration was considered as a feature in the regression models,since the correlation between symptom/drug score and mugwort pollen concentration was overall low and not signi cant.

Statistical analyses
Spearman's rank ρ correlation coe cient (33) and Kendall's rank τ correlation coe cient (34) nonparametric statistical tests were used to examine the correlation between daily ragweed pollen concentrations and symptoms' intensity, measured as the total daily number of symptoms observed in study patients. We adopted a non-parametric statistical approach as the target variables were not normally distributed.
The non-parametric Kruskal-Wallis test (35) and the Wilcoxon rank sum test (36) were used to compare mean symptom level scores between the two patient cohorts: non-AIT and AIT treated.
Time series analysis was used to analyze the in uence of the daily pollen concentrations cycle on the onset of allergic symptoms. It is reasonable to assume that the total number of symptoms currently observed depends on the daily pollen concentrations, as well as on the previous day's (lag=1) symptom values. For this reason, we applied a rst-order autoregressive distributed lag (ARDL) model to explore these short-run relationships (37).
Let x t (t=1, …,T) be the daily pollen concentrations and y t the total number of symptoms observed in the study patients. The ARDL model is de ned by where t is a white noise process, independent of x t , y t and y t − 1 , so that the model (1) can be estimated using ordinary least squares (OLS). The regression coe cients can be interpreted as measures of the in uence of the feature on the target variable.
We also considered a rst-order autoregressive distributed lag model similar to the model formulated in (1) with the addition of a lagged x t − 1 as a further explanatory variable, Furthermore, the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were used for model selection. Models with the lowest AIC and BIC values were considered to be the 'best'. Finally, since the pollen concentration data showed high variance, we used a square root transformation to stabilize the variance.
All statistical analyses were conducted using R 4.0.2 (38).

Results
Pollen counts. Statistical analyses Table 2 reports the descriptive statistics relating to the average number of daily symptoms per person per symptom category, i.e conjunctivitis, rhinitis and asthma.  Table 3 shows the results of the non-parametric correlation tests for the two patient cohorts: non-AIT and AIT treated.  Table 4). In AIT treated patients such correlation was more moderate but still signi cant (τ= (0.495, 0.600, 0.399) and ρ = (0.628, 0.777, 0.516) with a p-value<0.001).  Hereafter, we consider only the ragweed pollen concentration as a feature in the regression models, because the correlation between symptoms' intensity and mugwort pollen concentration was low and overall not signi cant. Figure 2 and Figure 3 show the total number of symptoms observed in patients for all three symptom categories, and the mean daily ragweed and mugwort pollen concentrations (pollen grains/m 3 ) averaged over all three traps. In the non-AIT cohort, the total number of daily symptoms increased through time from mid July to mid September: symptoms increased a few days after the rise in airborne ragweed pollen concentrations, continuing to increaseup to their peak value and further, due to the known late response to the allergen. We used the Kruskal-Wallis and the Wilcoxon rank sum teststo test the hypothesis that mean symptom levels in non-AIT patients were greater than those in the AIT treated group. We found signi cant differences between daily symptoms in the two groups at a 95% con dence level or higher (Table 5).  Figure 4 shows drugs use in the two patient cohorts, AIT treated and non-AIT treated, and the daily ragweed and mugwort pollen counts, averaged over thethree stations. Table 6 presents OLS estimates of the ARLD models (1), considering as covariate the square root of the ragweed pollen daily concentration (m −3 day −1 ). We also estimated model (2), but we selected model (1) according to the AIC and BIC values. We assumed that the target variable y t , i.e. the total number of symptoms at a given day t, may be explained in terms of current x t , i.e. the square root of the ragweed pollen concentration, and past symptoms values, y t − 1 .  Figure 5 shows the bi-dimensional exposure-response relationship estimated by ARDL models. Each panel can be read using two different perspectives: it represents the increase in number of total symptoms in each t + k future day following a single exposure equal to ℓ (ℓ = 1, 5, 10, 20) ragweed pollen per cubic meter of air at day t = 0 (forward interpretation), or otherwise the contributions of each t − k past day, with ragweed pollen per cubic meter of air equal to ℓ (ℓ = 1,5,10,20), to the increase in number of total symptoms at day t (backward interpretation).
In order to de ne the airborne ragweed pollen threshold levels for the protection of human health, we considered the quantiles of order (0.25, .5, .75) for the total number of symptoms (without distinguishing them in categories) per person in non-AIT patients in the period from August 01, 2014 (the rst date in which the average number of ragweed pollen was greater than 1) until September 02, 2014 (the ragweed pollen peak). The quantiles de ned four classes based onthe total number of symptoms: low, medium-low, medium-high, and high. The threshold levels corresponding to each class were obtained by computing the averaged ragweed and pollen concentrations (pollen grains/m 3 ) for the corresponding days, i.e. L1= 0.963, L2= 4.233, and L3 =14.444.
Data on other sensitizations were not used in the elaboration of the results as they were not signi cant for the purpose of our study.

Discussion
Key results The rst objective of our study was to test the hypothesis that the severity of symptoms of ragweed hay fever is directly related to the concentration of ragweed pollen in the air. The results of our study support this hypothesis. First, we found a strong signi cant correlation in AIT treated patients, and a moderate signi cant correlation in non-AIT patients, between ragweed pollen load and symptoms' intensity for all three symptom categories (i.e. conjunctival, nasal and bronchial). Moreover, in the non-AIT cohort, we observed a strong positive correlation between the number of symptoms and drug consumption. In the AIT treatedcohort this correlation was more moderate but still statistically signi cant. Second, the OSL estimates of the ARLD models (1) demonstrated that the daily ragweed pollen concentration is predictive of the severity of each of the symptom categories considered.
The second objective of our study was to establish threshold values of ragweed pollen concentrations for different levels of symptom severity. We were able to establishthree different average ragweed pollen concentration thresholds, each corresponding to a different symptom intensity level, i.e. L1 = 0.963, L2 = 4.233, and L3 =14.444.
Our third objective was totest the hypothesis that exposure to equal concentrations of ragweed pollen, during ragweed MPS, would result in less severe ocular, nasal and bronchial symptoms in AIT treated patients compared to non-AIT. We found that the mean symptom levels for the three symptoms categories were signi cantly greater, over the entire observation period, in non-AIT patients than in AIT treated. The mean daily symptom scores and the mean daily drug consumption scores in the AIT treatedgroup were lower, each day, compared tothenon-AIT group. Furthermore, the ARDL model (1) estimate indicated that the increase in pollen level has a greater contribution on the increase in symptoms in non-AIT patients compared to AIT treated.

Limitations
The main implicit limitation to the rst objective's results is that the symptom score attributed to ragweed pollen may have been in uenced by the coexistence of mugwort pollen. This may have occurred both in some subjects sensitized to mugwort, as well as in non-mugwort sensitized subjects, due to a crossreaction between ragweed and mugwort allergens (39). In fact, of the 66 ragweed sensitized subjects participating in the study, 21 had speci c IgE for Art v 1, and 7 for Art v 3. In 2014,the mugwort MPS partially overlapped with ragweed MPS during the study's symptom/drug-monitoring period (i.e. from July 16, 2014 to September 15, 2014). However, during the entire study period, mugwort pollen concentrations were extremely lower than ragweed concentrations, and did not show any peaks. Furthermore, mean symptom scoreswere never signi cantly positively correlated to mugwort pollen concentration values, suggesting that symptom scores did not depend on mugwort pollen concentration. Also the extremely low C max value of mugwort pollen during the 2014 mugwort MPS, supports our conclusion that the mugwort pollen concentration did not in uence the symptom scores of the study participants.
Other study limitations could arise from other factors that can affect the severity of respiratory symptoms. For example, we did not consider air pollution, which is particularly intense in the study area, due to the high industrial and commercial development. The same applies to possible intercurrent respiratory infections that, as far as we know, did not occur, but for which there was no speci c provision for reporting them in the CDSD.
The calculated threshold values of ragweed pollen concentration related to symptom severity may also have limitations. In fact, it is known that in the air we nd not only allergens contained in the pollen grains, but also free allergens in the form of bioaerosols. For example, studies found that when the concentration of ragweed pollen is 200 pollen grains/m 3 , contemporarily 2.5 mg/m 3 of free ragweed allergens are also present (40)(41)(42)(43). Consequently, a symptom threshold based on pollen counts alone cannot give a precise quanti cation of true exposure to all airborne allergens. Furthermore, atmospheric and environmental factors may alsosometimes in uence the results. Indeed, the amount of free allergen associated with pollen grains depends on a number of variables, including air humidity (which can in uence the release of the allergen-containing starch from pollen grains), rain, wind and the trauma that pollen suffers in the environment, such as that caused by intense car tra c (43). Moreover, there is evidence that ragweed pollen collected along high-tra c roads has a higher allergenicity than pollen collected in vegetated areas (44).
Concerning the third study objective,possible limits and biasesin the results could come from not having considered some of the AIT treatment variables, i.e.
the de nition of the allergenic composition of the extracts used, the maximum and cumulative dose administered, the dose of the main allergens administered, the route of administration and treatment duration.

Interpretation
The results of this study indicate that there is a correlation between ragweed pollen concentration and the severity of ocular, nasal, and bronchial symptom scores in ragweed allergy sufferers. The data are valid for the experimental conditions of our study, and we exclude a confounding effect due to a coexisting mugwort allergy. Based on the available variables, the correlation data and OLS estimates of the ARLD models (1) show a strong relationship between ragweed pollen concentration and symptom severity, supportingour main hypothesis. We observed a greater correlation between the number of symptoms and ragweed pollen concentrations in AIT treated subjects compared to non AIT, with regard to conjunctivitis and rhinitis. This result could be due to the fact that conjunctivitis and rhinitis are treated with OCD, since they are considered less severe symptoms than asthma. This explanation is also supported by the nding of a strong correlation between the number of symptoms and drug consumption, which is greater in non-AIT than in AIT treated subjects. It is likely that AIT treated subjects tend to take fewer drugs because they feel protected by the immunotherapy. This phenomenon is not observed in asthma, where there is no difference in the correlations between the number of symptoms and pollen concentration in AIT and non-AIT treated subjects.
Future studies should take into consideration all of the AIT variables (as mentioned above). However, under similar experimental conditions, we do not expect that these additional variables would critically affect the results.
In the literature, the majority of studies investigating the association between pollen load and symptom severity, are not speci cally aimed at ragweed pollen, or have not speci cally selected populations of subjects allergic to ragweed (26-30). A few studies that investigated exposure to ragweed pollen in populations of ragweed allergy patients, do support our results. Della Valle et al. (24) found a strong association between daily asthma symptoms and drug scores, and ragweed pollen concentrations in ragweed-sensitized children. They studied a total cohort of 430 children (4-12 years old), sensitized to different pollens, from which ragweed-allergic children had been enucleated and separately studied. Other factors that are important to consider when conducting similar studies are: selecting a su cient number of patients sensitized to ragweed, recruiting only subjects with demonstrated sensitization, placing the pollen traps close the subjects' residences, and recording both symptom and drug consumption scores.
The results of our study probably depend on having selected a random sample of patients havingboth an established sensitization to ragweed (to rAmb a 1) and con rmed seasonal respiratory symptoms from ragweed allergy, residing in a relatively small area (a few km2), highly infested with ragweed, and living very close to one of the three pollen traps. Moreover,inthe analyses we used the daily average of the three pollen-traps counts, which is likely to be a very representative variable of the quantity of pollen to which patientswere exposed daily in that area.
Furthermore, even though we recognize its limits, calculating a threshold for symptom pollen levels is an important toolfor setting ragweed-allergy preventative measures. Unfortunately, comparisons with the results of previous studies are limited. In fact, while a number studies have determined symptom thresholds for grass and birch pollen, similar studies for ragweed pollen are practically non existent. An olderstudy (24), where almost all patients with a ragweed allergic rhinitis were symptomatic, proposed a range between 10 to 50 pollen grains/m 3 as threshold level. Another study (46) established the ragweed pollen concentration threshold for the onset of hay fever symptoms at 1-3 pollen grains/m 3 .. Lastly, a study (47) on environmental triggers of asthma in children,foundthat a threshold of 6-9 weed (not ragweed) pollen grain/m 3 could trigger asthma symptoms. Unfortunately, the data provided by these studies cannot be used for comparison with ours. Dataon ragweed pollen symptom thresholds can be useful for individual prevention only if they are publishedtogether with data on pollen concentrations. Studiesthat measure pollen-symptom thresholds along with aerobiological information do exist, but validated criteria on how to provide this information to the public are still lacking (48-51). Nonetheless, it can be very useful to know the symptomatic thresholds; since they can be used as a standard for all preventative measures carried out in an area, with the aim of reducing the ragweed plant's expansion.
Finally, the results of our study also support the hypothesis that subjects treated with AIT have less severe symptoms upon exposure to the same concentration of ragweed pollen, compared to subjects not treated with AIT. Even though there are limitations to these results due to some unmeasured AIT variables, such as the characteristics of the ragweed extracts used, the duration and methods of administration. Nonetheless, our general aim was to highlight the fact that AIT can modify the severity of symptoms from exposure to ragweed pollen in treated subjects. The model we used can be adopted to evaluate a speci c ragweed extract, or a particular route of administration, or duration of AIT. Our experimental model satis es the requirements cited by the EAACI Position Paper (52). Here again, comparisons with previously published studies on the "real world" effectiveness of AIT are limited. In fact, we could not nd any published "real world" comparative studies of patients treated with AIT pollen, versus untreated patients, similar to our study model. We believe that our results are unprecedented in this respect. Although not directly comparable to our study, some "real world" studies compared patients treated with AIT for different seasonal allergens, for more than one year of treatment, and with untreated control subjects. In these studies, AIT effectiveness was con rmed based only on one signi cant difference in symptom/drug score in AIT treated subjects versus controls (53-56).

Generalizability
Our data were collected in a population of subjects allergic to ragweed (all residing in a well-de ned area, highly infested with ragweed plants) and with a high prevalence of seasonal conjunctival, nasal and bronchial symptoms due to sensitization to ragweed. Therefore, our results can be generalized in situations with similar contexts. We strongly believe that our ndingsthat ragweed hay fever symptoms are positively correlated with ragweed pollen concentration, are generalizable to many areas of the world where there are high concentrations of airborne ragweed pollen. Furthermore, our results have signi cant implications for the metropolitan area of Milan, as they con rm the validity of the efforts made to contain the expansion of the ragweed plant in the territory. Time series of observed drugs use in the two cohorts of patients: AIT treated and non-AIT, and ragweed and mugwort pollen concentrations (pollen grains/m 3 )