SEARCH

SEARCH BY CITATION

Keywords:

  • interstitial lung disease;
  • lung damage;
  • pulmonary toxicity;
  • radiation-induced effects;
  • respiratory epidemiology

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

Background and objective:  Radiotherapy and an increasing number of substances are implicated in the pathogenesis of interstitial lung disease (ILD). While the frequency of published data on more common ILD entities such as the idiopathic interstitial pneumonias has increased in recent years, less attention has been given to relatively rarely occurring forms such as drug-/radiation-induced ILD.

Methods:  Data from the UK-based General Practice Research Database (GPRD) was used to estimate the incidence of drug-/radiation-induced ILD over a 12-year period (1997–2008). Crude incidence rates were stratified by gender, age group and calendar period, and rate ratios were adjusted using Poisson regression. All-cause mortality was modelled using Cox regression, and characteristics at diagnosis were compared with a random sample of matched, non-ILD controls using conditional logistic regression.

Results:  A total of 128 patients with an incident diagnosis of drug-/radiation-induced ILD were identified, and the overall incidence density during the study period was 4.1 (95% confidence interval 3.4–4.9) per million person-years. Incidence rates increased during the time period 1997–2005 and decreased thereafter. The adjusted all-cause mortality was >4 times higher in cases compared with controls.

Conclusions:  This UK population-based study characterizes patients diagnosed with drug-/radiation-induced ILD and quantifies incidence and all-cause mortality during 1997–2008. No statistically significant time trend in incidence was found, despite having observed numeric increases in incidence rates during the study window. Future research using the GPRD and other data sources is required to better understand the disposition of patients diagnosed with drug-/radiation-induced ILD and to investigate potential trends incidence and mortality over time.


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

An increasing number of prescription drugs and novel biologics, such as monoclonal antibodies, antitumour necrosis factor, interferons and immunoglobulins, are implicated in the pathogenesis of infiltrative and/or parenchymal lung disease.1–3 Additionally, certain herbal remedies, dietary supplements, chemotherapeutics and irradiation are also known to induce interstitial lung disease (ILD).4–7 During the 1990s, data from European ILD registries indicated that drug-induced ILD cases represented between 1.9% and 3.3% of all ILD cases reported.8 While the amount of published epidemiological data on more commonly occurring types of ILD (e.g. the idiopathic interstitial pneumonias and sarcoidosis) has increased in the recent past, less attention is devoted to the epidemiology of less frequently occurring subgroups such as drug- and radiation-induced ILD. Moreover, published data on the frequency and on mortality of patients with drug- and radiation-induced ILD cases in the UK general population are—to the best of our knowledge—non-existent. To better understand the epidemiology of this ILD subgroup and to predict demand on health-care resources, there is a need to describe and quantify drug- and radiation-induced ILD in the UK general population. We used the UK General Practice Research Database (GPRD) to characterize patients diagnosed with ILD induced by drugs or radiation, and to examine the incidence and all-cause mortality of these ILD subgroups in the UK general population.

METHODS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

Data source

The GPRD has been described elsewhere in detail.9 In summary, it is a large and well-validated UK-based database that was established in June 1987. It contains anonymized electronic patient records collected from routine general practice, and the geographical, age and gender distributions of the GPRD population have been shown to be representative of the UK general population.9 The GPRD encompasses some 9 million patients who are or were enrolled with selected general practitioners throughout the UK, covering approximately 50 million patient-years of follow-up. The participating general practitioners has been trained to record medical information in a standard manner and to provide anonymized records to the GPRD group within the Medicines and Healthcare products Regulatory Agency, the UK's medicines and devices regulator. The recorded data include demographical information, medical diagnoses and drug prescriptions. Data from the GPRD have been used in previous studies involving ILD, and use of the data for respiratory epidemiology has been validated.10,11 The GPRD is managed by the Medicines and Healthcare products Regulatory Agency, and the present study protocol was approved by the Independent Scientific Advisory Committee for Medicines and Healthcare products Regulatory Agency database research.

Cohort study population

We identified in the GPRD all patients with a first-time diagnosis of drug- or radiation-induced ILD between January 1, 1997 and December 31, 2008. Diagnoses in the GPRD are coded using the Read coding system, and our list of diagnoses included any Read code explicitly mentioning drug- or radiation-related lung disease: acute pulmonary radiation disease, acute radiation pneumonitis, chronic pulmonary radiation disease, chronic pulmonary fibrosis following radiation, drug-induced interstitial lung disorders, acute drug-induced interstitial lung disorders or chronic drug-induced interstitial lung disorders. Read codes without explicit mention of a drug- or radiation-related lung disease, such as ‘lung disease due to external agents not otherwise specified (NOS)’, were not used. In order to increase the likelihood of capturing incident rather than prevalent cases, we excluded patients with less than 3 years of active recording history prior to the date of the first drug/radiation ILD diagnosis. In addition, we identified at random from the GPRD control subjects without diagnosis of drug/radiation ILD, matched 1:1 on age (same year of birth), gender, general practice and calendar time (i.e. by using the date of the first ILD diagnosis of the case). We applied the same exclusion criteria to the control group as to the case patients. Within this study population, we assessed and compared demography, prescription drug use, comorbidities, and other characteristics of cases and controls at or prior to the date of incident diagnosis.

Data analysis

An initial descriptive analysis was conducted to summarize demographics and other baseline characteristics of cases and their matched controls. Comparisons between cases and controls were then analysed using conditional logistic regression. With respect to differences in comorbidities and prescription drug use at or prior to incident diagnosis, an a priori decision was taken to present and further adjust those odds ratios (OR) that reached a conventional level of statistical significance (i.e. P ≤ 0.05).

Person-time denominators for the calculations of incidence density were derived from the GPRD. For each calendar period, only those patients in the GPRD who were alive, actively registered and had at least 3 years of recorded history contributed person-time to the denominators. Incident cases were grouped into 10-year age bands, and calendar time was divided into four 3-year periods (1997–1999, 2000–2002, 2003–2005, 2006–2008). Crude incidence rates were calculated for the group of drug/radiation ILD as a whole and stratified by gender, age group and calendar period, with exact 95% confidence intervals (CI) calculated from the Poisson distribution. In order to model disease incidence and estimate incidence rate ratios, fixed-effects Poisson multivariate regression was used. The earlier methods were repeated for cases diagnosed with drug-induced and radiation-induced ILD separately.

Initially, the comparison of survival between patients with drug-/radiation-induced ILD and matched non-ILD controls was summarized using Kaplan–Meier plots. Survival of cases and controls was then modelled using stratified Cox proportional hazards regression. Covariates included smoking history, body mass index and cancer as comorbidity. All statistical analyses were performed with the statistical software SAS (release 9.2, SAS Institute, Inc., Cary, NC, USA). The proportional hazards assumption was assessed using the RESAMPLE option within the SAS procedure PHREG.

RESULTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

A total of 128 patients with an incident diagnosis of drug-/radiation-induced ILD were identified during the period 1997–2008, of which 106 (83%) cases were diagnosed with radiation-induced lung disease. The frequency distribution of incident diagnoses is shown in Table 1. The mean age (standard deviation) at diagnosis was 67 (11.9) years, and 73 (57%) cases were female. Males and females were of similar mean age at the time of first diagnosis (67 years (±standard deviation 12.1) vs 66 years (±standard deviation 11.8), respectively), and the majority of cases were diagnosed in the age group 60–79 years. The age distribution of cases is shown in Figure 1. With respect to health-care resource utilization, cases (as opposed to their matched controls) were almost eight times more likely to have had ≥20 versus <20 practice visits and were greater than three times more likely to have had ≥25 versus <25 drug prescriptions during 1 year prior to incident diagnosis (Table 2). Likewise, cases were ≥1.8 times more likely than their matched general-population controls to have a range of both comorbid diagnoses and prescription drug use at or prior to diagnosis; the highest OR was observed for having a diagnosis of cancer (in situ metastasis) (OR = 43.5, 95% CI 10.7–177.0) and use of immunosuppressives (OR = 11.0, 95% CI 1.42–85.20), respectively. After adjusting for cancer diagnosis, elevated OR achieving statistical significance was noted for diagnoses of RA, constipation, IHD and CHF, as well as for prescription drug use of immunosuppressives, corticosteroids, anti-arrhythmics, non-steroidal anti-inflammatory drugs and diuretics. Results of comorbidities and prescription drug use can be found in Table 3 in which the data for each category is sorted in descending order according to the magnitude of the crude OR.

Table 1.  Frequency of incident diagnoses of drug-/radiation-induced interstitial lung disease (n = 128)
Diagnosis codesFrequencyPercent
Acute pulmonary radiation disease10.78
Acute radiation pneumonitis6953.91
Chronic pulmonary radiation disease10.78
Chronic pulmonary fibrosis following radiation3527.34
Drug-induced interstitial lung disorders1814.06
Acute drug-induced interstitial lung disorders32.34
Chronic drug-induced interstitial lung disorders10.78
image

Figure 1. Age at diagnosis of drug-/radiation-induced interstitial lung disease (n = 128).

Download figure to PowerPoint

Table 2.  Characteristics of drug-/radiation-induced interstitial lung disease cases and their matched controls at incident diagnosis
CharacteristicsCases (%) (n = 128)Controls (%) (n = 128)OR (95% CI)
  •  

    Last recording prior to index date.

  •  

    Based on recordings of diagnoses and prescriptions (falling on separate dates) during 1 year prior to incident diagnosis.

  • § 

    During 1 year prior to incident diagnosis.

  • BMI, body mass index; CI, confidence interval; OD, odds ratio; SD, standard deviation; —, not applicable.

Mean age (years) ± SD66.6 ± 11.966.7 ± 11.9
Females73 (57.0)73 (57.0)
BMI (kg/m2)   
 <2544 (34.4)45 (35.2)1.00 (ref.)
 ≥2567 (52.3)56 (43.8)1.29 (0.70–2.37)
 Unknown17 (13.3)27 (21.1)0.60 (0.27–1.35)
Smoking status   
 Nonsmokers52 (40.6)62 (48.4)1.00 (ref.)
 Current/ex-smokers70 (54.7)53 (41.4)1.68 (0.96–2.93)
 Unknown6 (4.7)13 (10.2)0.39 (0.10–1.46)
No. of cigarettes per day   
 010 (7.8)13 (10.2)1.00 (ref.)
 1–1022 (17.2)14 (10.9)2.14 (0.75–6.15)
 11+16 (12.5)10 (7.8)2.07 (0.66–6.50)
 Unknown80 (62.5)91 (71.1)1.02 (0.41–2.54)
Alcohol consumption (no. of units/week)   
 074 (57.8)72 (56.3)1.00 (ref.)
 1–726 (20.3)27 (21.1)0.92 (0.46–1.84)
 8+28 (21.9)29 (22.7)0.92 (0.48–1.79)
No. of practice visits   
 0–1933 (25.8)87 (68.0)1.00 (ref.)
 20+95 (74.2)41 (32.0)7.75 (3.71–16.18)
No. of drug prescriptions§   
 0–2446 (35.9)83 (64.8)1.00 (ref.)
 25+82 (64.1)45 (35.2)3.47 (1.95–6.16)
Table 3.  Odds of comorbidities and prescription drug use for drug-/radiation-induced interstitial lung disease cases and their matched controls at/prior to incident diagnosis
 Cases (%) (n = 128)Controls (%) (n = 128)OR (95% CI)
CrudeAdjusted
  •  

    Adjusted by matching for age, gender, geographical location, and no. of years of recorded history in database.

  •  

    Further adjusted by cancer status at/prior to index date.

  • § 

    Nonestimable due to one or more strata with null counts.

  • CHF, congestive heart failure; CI, confidence interval; COPD, chronic obstructive pulmonary disease; Dx, diagnosis; GERD, gastroesophageal reflux disorder; H2, histamine-2; i.s., in situ; IHD, ischaemic heart disease; NSAID, non-steroidal anti-inflammatory drug; OR, odds ratio; PPI, proton-pump inhibitor; RA, rheumatoid arthritis; Rx, prescription; —, not applicable.

Comorbid diagnoses
Cancer (i.s. metastasis)97 (75.8)12 (9.4)43.5 (10.7–177.0)
RA13 (10.2)2 (1.6)6.50 (1.47–28.9)21.32 (2.46–184.82)
Constipation33 (25.8)10 (12.8)5.60 (2.16–14.5)6.09 (1.32–28.13)
Hypothyroidism18 (14.1)5 (3.9)4.25 (1.43–12.6)0.14 (0.01–2.96)
Arrhythmia25 (18.5)7 (5.5)4.00 (1.64–9.79)Nonestimable§
COPD17 (13.3)5 (3.9)3.40 (1.25–9.22)2.04 (0.54–7.62)
IHD30 (23.4)15 (11.7)2.50 (1.20–5.21)22.18 (2.66–184.98)
CHF15 (11.7)6 (4.7)2.50 (0.97–6.44)6.27 (1.38–28.43)
GERD/esophag./dyspep./gastr.40 (31.3)28 (21.9)2.00 (1.05–3.80)0.87 (0.34–2.19)
Asthma26 (20.3)16 (12.5)1.83 (0.91–3.70)1.41 (0.49–4.10
Rx drug use at/prior to Dx
Immunosuppressives11 (8.6)1 (0.08)11.0 (1.42–85.2)27.72 (1.91–402.86)
Corticosteroids (systemic)67 (52.3)31 (24.2)4.00 (2.12–7.53)4.70 (1.70–12.98)
Anticholinergics24 (18.8)6 (4.7)4.00 (1.64–9.79)3.18 (0.96–10.58)
B—Agonists and comb.62 (48.4)28 (21.9)3.62 (1.96–6.68)2.50 (0.99–6.33)
Antiarrhythmics17 (13.3)7 (5.5)2.67 (1.04–6.81)27.93 (2.88–271.26)
Thyroid gland therapy20 (15.6)9 (7.0)2.57 (1.07–6.16)3.02 (0.77–11.95)
PPI54 (42.2)33 (25.8)2.31 (1.29–4.16)2.86 (0.88–9.32)
NSAID98 (76.6)79 (61.7)2.12 (1.19–3.77)3.09 (1.08–8.80)
H2-receptor antagonists47 (36.7)31 (24.2)2.00 (1.10–3.64)2.04 (0.79–5.32)
Corticosteroids (inhaled)51 (40.0)34 (26.6)1.94 (1.10–3.43)1.27 (0.55–2.90)
Diuretics59 (46.1)41 (32.0)1.86 (1.09–3.16)3.00 (1.25–7.20)

Between 1997 and 2008, the overall incidence density of drug-/radiation-induced ILD was 4.1 (95% CI 3.4–4.9) per 1 000 000 (106) person-years, and the rate in females was approximately 20% higher than in males. Across age groups, the lowest incidence density was observed for those below the age of 60 years (1.2 (95% CI 0.8–1.7)/106 person-years); the rates in older age groups were approximately 8–12fold higher. Crude incidence rates increased progressively during the first three periods and then decreased from the period 2003–2005 to 2006–2008 (5.1 (95% CI 3.7–6.9) vs 4.4 (95% CI 3.1–6.1)/106 person-years, respectively). Crude incidence rates for the aggregate group of drug-/radiation-induced ILD are shown in Table 4. With respect to each diagnosis category separately (i.e. drug- vs radiation-induced ILD), incidence density increased progressively from 1997 to 2005 for drug-induced ILD and then dropped by more than 50% between the period 2003–2005 and 2006–2008. In the radiation-induced ILD category, incidence density remained somewhat stable during the first two 3-year periods (1997–2002, range 2.9–3.0/106/year) and then increased during the following two 3-year periods (2003–2008, approximately 3.8/106/year). Category-specific incidence density by calendar period can be found in Table 5.

Table 4.  Crude incidence rates for drug-/radiation-induced interstitial lung disease
 No. of casesTotal person-yearsIR95% CI (exact Poisson)
  1. CI, confidence interval; IR, incidence rate (density) per 1 000 000 person-years.

Overall12831 165 672.414.113.43–4.88
Gender    
 Female7316 421 082.124.453.48–5.59
 Male5514 744 590.293.732.81–4.86
Age groups    
 <602823 309 579.771.200.80–1.74
 60–69453 474 347.4412.959.45–17.30
 70–79382 689 162.5214.1310.0–19.40
 80+171 692 582.7010.045.85–16.10
Calendar period    
 1997–1999226 696 773.293.292.06–4.97
 2000–2002267 646 978.153.402.22–4.98
 2003–2005428 232 635.235.103.68–6.90
 2006–2008388 589 285.744.423.13–6.07
Table 5.  Crude incidence rates for drug- and radiation-induced ILD separately
 No. of casesTotal person-yearsIR95% CI (exact Poisson)
  1. CI, confidence interval; ILD, interstitial lung disease; IR, incidence rate (density) per million person-years.

Drug-induced ILD (n = 22)
Calendar period    
 1997–199926 696 773.290.300.04–1.08
 2000–200247 646 978.150.520.14–1.34
 2003–2005118 232 635.231.340.67–2.39
 2006–200858 589 285.740.580.19–1.36
Radiation-induced ILD (106)
Calendar period    
 1997–1999206 696 773.292.991.82–4.61
 2000–2002227 646 978.152.881.80–4.36
 2003–2005318 232 635.233.772.56–5.34
 2006–2008338 589 285.743.842.64–5.40

After adjusting for the effects of age and gender in the aggregate category of drug-/radiation-induced ILD, although having observed numeric increases in incidence from 1997–2005 and a subsequent decrease from the period 2003–2005 to 2006–2008, there was no evidence of statistically significant heterogeneity of rate ratios over time (P = 0.74). In contrast, after adjusting for gender and calendar period, there was evidence of statistically significant heterogeneity of rate ratios across age groups. Incidence density rates in the age groups 60–69 and 70–79 were more than 10 times higher than in those under 60 years of age. Incidence rate ratios for the aggregate category of drug-/radiation-induced are shown in Table 6.

Table 6.  Poisson regression modelling of drug-/radiation-induced incidence (aggregate group)
 IRR95% CILRT P-value
  • P for trend.

  • CI, confidence interval; IRR, incidence rate ratio; LRT, likelihood ratio test; —, not applicable.

Gender   
 Female (ref.)1.000.6254
 Male0.910.64–1.30
Age groups   
 <60 (ref.)1.00<0.001*
 60–6910.736.69–17.20
 70–7911.767.22–19.17
 80+8.224.49–15.04
Calendar period   
 1997–1999 (ref.)1.000.7387*
 2000–20021.040.59–1.84
 2003–20051.560.93–2.61
 2006–20081.310.77–2.21

A total of 64 cases died during follow-up versus 16 controls. The mean (±standard deviation) survival time for cases was 3.1 (± 3.1) versus 4.8 (± 3.1) years for general-population controls. Over the 12-year follow-up period cases were seven times more likely to die compared with their matched general-population controls (Fig. 2) (hazard ratio = 7.13, 95% CI 3.40–14.93). After allowing for the confounding effect of baseline cancer diagnosis, the difference in 12-year survival decreased yet remained elevated (hazard ratio = 4.04, 95% CI 1.56–10.56 (Table 7). No evidence was found to suggest a departure from the proportional hazards assumption in the final model.

image

Figure 2. Kaplan–Meier plot comparing survival in cases diagnosed with drug-/radiation-induced interstitial lung disease and matched general-population controls. inline image, cases; inline image, controls.

Download figure to PowerPoint

Table 7.  Survival analyses for all-cause mortality (stratified Cox proportional hazards regression)
 No. of deathsHR95% CIP-value (LRT)
  1. CI, confidence interval; HR, hazard ratio; ILD, interstitial lung disease; LRT, likelihood ratio test.

Crude (adjusted by matching for age, gender, location, no. of years of recorded medical history)
General-population controls161.003.40–14.93<0.0001
Drug-/radiation-induced ILD cases647.13
Further adjusted for baseline smoking status
General-population controls161.003.71–21.45<0.0001
Drug-/radiation-induced ILD cases648.92
Further adjusted for baseline smoking status and baseline cancer diagnosis
General-population controls161.001.60–13.44<0.0001
Drug-/radiation-induced ILD cases644.64

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

In this general population-based UK cohort of drug- and radiation-induced ILD patients, the overall incidence rate was approximately 4 per million population per year, suggesting that each year in the UK, there will be ∼250 new cases of drug and radiation ILD. Overall, numeric increases were observed in incidence density rates during the first 9 years of the study period, followed by a decline during the final 3 years. This drop in incidence in the aggregate category of drug- and radiation-induced ILD was clearly attributable to the >50% decrease in incidence observed in the subgroup of drug-induced ILD during the same period. Incidence density of radiation-induced ILD remained relatively stable during the first 6 years (range 2.9–3.0/106 person-years) and then increased to nearly 3.8/106 person-years.

All-cause mortality adjusted by matching on age, gender, geography (via general practice as a proxy) and number of years of recorded medical history was seven times higher in cases compared with general-population controls. After further adjustment by both baseline smoking status and cancer diagnosis (i.e. ∼75% of cases had a diagnosis of cancer at/prior to incident drug-/radiation-induced ILD diagnosis), the all-cause mortality rate for cases remained elevated at >4.5 times that of general population-based controls. The majority of cases (∼54%) in our cohort was diagnosed with acute radiation pneumonitis (as opposed to ‘classic’ or ‘sporadic’ radiation pneumonitis), a condition that is associated with significant symptoms, and in some cases develops into respiratory failure or acute respiratory distress syndrome.7,12 However, the adjusted all-cause mortality rate would suggest that many of these patients suffer progressive disease resulting in premature death. Whether or not the cause of death was attributable to radiation-induced ILD, nonremissive cancer or other comorbidities could not be examined because precise information on cause of death is not readily available in the GPRD.

A potential weakness of the present study could be the validity of the ILD diagnoses. However, the validity of many disease diagnoses has been assessed in the GPRD and has consistently found to be high.13–15 Furthermore, other researchers have reported a high validity of the diagnosis of idiopathic pulmonary fibrosis, another ILD, in the GPRD.16 Moreover, it is highly unlikely that general practitioners would record a diagnosis of drug- or radiation-induced ILD without confirmation by specialist referral. In order to explore the specificity of the ILD diagnoses in our cohort, we examined the medical history in more detail and found the following: (i) of the 106 cases with a diagnosis of radiation-induced ILD, 79 had a diagnosis of thoracic-related cancer, while the remaining 27 had a diagnosis of cancer that was NOS; and (ii) of the 22 cases with a diagnosis of drug-induced ILD, the last immunosuppressive prescribed for nine of these cases was either leflunomide (n = 1) or methotrexate (n = 8), while eight patients who did not have any history immunosuppressive drug prescription (i.e. mutually exclusive of the nine patients previously described) were prescribed amiodarone immediately prior to their ILD diagnosis. All aforementioned prescription drugs are known potential inducers of ILD. In light of the earlier reasons and findings, we expect the diagnoses used for the present study to have a high specificity.

On the other hand, sensitivity of the diagnoses used may not be as high as the specificity. Two main reasons for this are the potential for misclassification of true ILD cases into less specific disease diagnosis categories and the possibility of having missed less severe cases. For instance, we did not use less specific respiratory-related diagnosis codes, such as ‘lung disease due to other external agents NOS’ or ‘other external agent causing respiratory condition’ in our case selection algorithm. Overall, there were 649 cases with such an unspecific diagnosis that we subsequently reviewed in more detail. None of these had evidence of a drug- or radiation-induced ILD diagnosis at a later point in time in their medical record, which indicates that misclassification is unlikely. It is also possible that less severe cases have escaped detection by general practitioners and specialists. Thus, it is possible that we did not have available for analysis all true cases of drug- and radiation-induced ILD from the given sampling frame, which means that we may have underestimated the true incidence density of this ILD subgroup in the UK. Furthermore, we cannot be certain that all newly identified cases were indeed incident cases. However, we excluded cases with less than 3 years of recorded medical history in order to minimize the possibility of including prevalent cases in our incidence numerators, so we expect the occurrence of such error to be minimal. To the extent that this error occurred, it will have resulted in a small overestimation of the incidence density rates.

We have been unable to find any other population-based cohort analyses in the UK with which to compare our incidence or mortality data. However, data from various European and US studies show some similarities and differences. One of the first ever population-based epidemiological studies of ILD in the late 1980s (in Bernalillo County, New Mexico) reported an incidence of 1.45/105 person-years for drug and radiation ILD, which is approximately 3.5 times higher than the rate observed in our cohort.17 One explanation might be the use of a more comprehensive case ascertainment that included ILD cases not only diagnosed by pulmonologists and general practitioners but also those cases discovered in hospital discharge reports, death certificates and autopsy reports. From a more recent survey of ILD-special interest pulmonology centres in Greece, Karakatsani et al. reported an incidence rate of 0.07/105 person-years for drug-induced ILD, which is nearly identical to our UK category-specific incidence rate for drug-induced ILD.18 Similarly, López-Campos and Rodríguez-Becerra reported an incidence of 0.08/105 person-years for drug-induced ILD from an ILD registry encompassing nine provinces in Southern Spain; data from the same registry reported an incidence rate of 0.00/105 person-years (only two cases of the total 744 ILD cases reported) for radiation-related ILD, which is strikingly different to the rate observed in our study.19 Factors influencing this difference could be the use of a different coding system (International Classification of Diseases codes vs Read codes), misclassification of ILD diagnosis, differences in prevalence of thoracic cancers (for which radiotherapy would be indicated), and differences in age and gender distribution.

It is acknowledged in the literature that the frequency of drug-induced lung disease reported from any data source is likely to underestimate true disease frequency.12,20,21 Many drugs implicated in ILD are primarily prescribed by non-respiratory physicians, and under-diagnosis is likely. Moreover, subclinical forms of drug-induced ILD are likely to escape detection by plain chest radiography as opposed to the use of high resolution computed tomography (HRCT), which would take place in close collaboration with a respiratory specialist. In the oncology setting, where drug-induced ILD is likely to be common, a diagnosis of drug-induced ILD will often remain inconclusive because the patient's condition precludes further (invasive) diagnostic assessment.

In this study of the aggregate category of drug-/radiation-induced ILD, no clear trend in incidence can be seen. This is evident from the lack of heterogeneity found in the numerically increasing incidence rates during 1997–2005, followed by a decrease during the final calendar period (2006–2008). However, due to the relatively small sample size of cases, the study did not have sufficient statistical power to detect a trend in incidence and had such a trend truly existed during the study window. Furthermore, we cannot tell whether the overall numeric increase reflects a real increase, or whether it is, at least to a certain extent attributable to secular trends. First, during the study window, major political changes in the NHS occurred, including more than a doubling in funding under the auspices of the New Deal.22 This investment in national health care was in part used to boost capacity of NHS services and to modernize facilities, thus potentially increasing the likelihood of detecting new cases of drug-/radiation-induced ILD. Second, a certain proportion of incident cases in this time period may have been due to increased recognition of ILD in general via the increased use of HRCT. Third, following publication of both the British Thoracic Society's guideline on diagnosis and care of diffuse parenchymal lung disease in 199923 and the American Thoracic Society/European Respiratory Society consensus statements on idiopathic interstitial pneumonia in 2002,24 a greater appreciation of the value of precise diagnosis according to defined criteria will likely have occurred, not to mention increased patient expectations. For the earlier reasons, an increase in case ascertainment may have played a role in the observed increase in incidence density rates.

This study makes use of data from a well-validated database that is representative of the UK general population. It is the first general population-based observational study quantifying incidence and mortality rates of diagnosed drug-/radiation-induced ILD in the UK.

ACKNOWLEDGEMENTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES

This study was funded by an unconditional grant from Actelion Pharmaceuticals Ltd.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  • 1
    Camus P. Drug-induced infiltrative lung diseases. In: Schwarz MI, King TE Jr (eds). Interstitial Lung Disease, 4th edn. Decker, Hamilton, 2003; 485534.
  • 2
    Limper AH. Chemotherapy—induced lung disease. Clin. Chest Med. 2004; 25: 5364.
  • 3
    Foucher P, Martha BW. 1997. Pneumotox Website. http://www.pneumotox.com[Accessed 22 May 2012]. Last update V1.9 June 2011.
  • 4
    Benson MK, Bentley AM. Lung disease induced by drug addiction. Thorax 1995; 50: 11257.
  • 5
    Wolff AJ, O'Donnell AE. Pulmonary effects of illicit drug use. Clin. Chest Med. 2004; 25: 20316.
  • 6
    Higenbottam TW. Bronchiolitis obliterans following the ingestion of an Asian shrub leaf. Thorax 1997; 52(Suppl 3): S6872.
  • 7
    Abratt RP, Morgan GW, Silvestri G et al. Pulmonary complications of radiation therapy. Clin. Chest Med. 2004; 25: 16777.
  • 8
    Thomeer MJ, Costabe U, Rizzato G et al. Comparison of registries of interstitial lung diseases in three European countries. Eur. Respir. J. Suppl. 2001; 32: 114s8s.
  • 9
    Garcia Rodriquez L, Perez Gutthann S. Use of the UK general practice research database for pharmacoepidemiology. Br. J. Clin. Pharmacol. 1998; 45: 41925.
  • 10
    Gribbin J, Hubbard RB, Jeune IL et al. Incidence and mortality of idiopathic pulmonary fibrosis and sarcoidosis in the UK. Thorax 2006; 61: 9805.
  • 11
    Hansell A, Hollowell J, Nichols T et al. Use of the General Practice Research Database (GPRD) for respiratory epidemiology: a comparison with the 4th Morbidity Survey in General Practice (MSGP4). Thorax 1999; 54: 4139.
  • 12
    Camus P, Fanton A, Bonniaud P et al. Interstitial lung disease induced by drugs and radiation. Respiration 2004; 71(4): 30126.
  • 13
    Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. BMJ 1991; 302: 7668.
  • 14
    Jick H, Terris BZ, Derby LE et al. Further validation of information recorded on a general practitioner based computerized data resource in the United Kingdom. Pharmacoepidemiol. Drug Saf. 1992; 1: 3479.
  • 15
    Jick SS, Kaye JA, Vasilakis-Scaramozza C et al. Validity of the general practice research database. Pharmacotherapy 2003; 23: 6869.
  • 16
    Hubbard R, Venn A, Lewis S et al. Lung cancer and cryptogenic fibrosing alveolitis. A population-based cohort study. Am. J. Respir. Crit. Care Med. 2000; 161: 58.
  • 17
    Coultas DB, Zumwalt RE, Black WC et al. The epidemiology of interstitial lung diseases. Am. J. Respir. Crit. Care Med. 1994; 150: 96772.
  • 18
    Karakatsani A, Papakosta D, Rapti A et al. Epidemiology of interstitial lung diseases in Greece. Respir. Med. 2009; 103: 11229.
  • 19
    López-Campos JL, Rodríguez-Becerra E. Incidence of interstitial lung diseases in the south of Spain 1998–2000: the RENIA study. Eur. J. Epidemiol. 2004; 19: 15561.
  • 20
    Demedts M, Wells AU, Antó JM et al. Interstitial lung diseases: an epidemiological overview. Eur. Respir. J. Suppl. 2001; 32: 2s16s.
  • 21
    Camus PH, Foucher P, Bonniaud PH et al. Drug-induced infiltrative lung disease. Eur. Respir. J. Suppl. 2001; 32: 93s100s.
  • 22
    The NHS plan. A Plan for Investment. A Plan for Reform. Stationery Office, London. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_4002960[Accessed 22 May 2012].
  • 23
    British Thoracic Society and Standards of Care Committee. The diagnosis, assessment and treatment of diffuse parenchymal lung disease in adults. Thorax 1999; 54(Suppl 1): S1S28.
  • 24
    American Thoracic Society, European Respiratory Society. International multidisciplinary consensus classification of the idiopathic interstitial pneumonias. Am. J. Respir. Crit. Care Med. 2002; 165: 277304.