You are here

Systematic characterization of gastrointestinal clinical trials

Digestive and Liver Disease, In Press, Corrected Proof, Available online 15 January 2016,

Abstract

Background

Clinical guidelines are commonly based on inadequate evidence, suggesting deficiencies in the present portfolio of clinical research.

Aims

To investigate characteristics of clinical trials examining gastrointestinal (GI) diseases registered in ClinicalTrials.gov.

Methods

A cross-sectional analysis of 13,647 GI trials and 111,535 non-GI trials initiated between January 1997 and September 2013 was performed. Entries were sorted by operational status, purpose, interventions, trial design, and epochs to identify trends and interactions in trial properties.

Results

The global production of GI trials has remained static in recent years and a majority of research efforts are focused on a few diseases. While GI trials are generally produced by highly populated US states and countries, they are also seldom larger than 500 patients. The likelihood of using data monitoring committees, randomization, and double blinding in GI trials has increased over time, though a substantial fraction of GI trials still do not employ rigorous trial designs. While levels of GI trials correlate with disease burden, the explained variance of GI trials by disease burden worldwide is poor.

Conclusion

GI trials are chiefly concentrated in few diseases and highly populated regions, exhibit heterogeneous trends and methodologies, and are sensitive to disease burdens, though more so within North America than worldwide.

Keywords: Clinical trials, Disease burden, Gastrointestinal diseases, Trial designs.

1. Introduction

Advances in medical practice are dependent on our ability to scrutinize how preventative, diagnostic, therapeutic, and prognostic interventions applied in clinical trials may alleviate the burden of disease. However, preceding the advent of clinical trial registries, an accessible and comprehensive resource for identifying clinical trials regardless of their publication status was lacking. In 1997 the US Food and Drug Administration Modernization Act mandated the public registration of interventional clinical trials, culminating in the creation of a web-based clinical trial registry, ClinicalTrials.gov [1]. By 2005, the International Committee of Medical Journal Editors mandated that all interventional clinical trials submitted for publication would require prior disclosure of study characteristics in a registry [2]. Since these pivotal measures, the number of entries in public clinical trial registries has accelerated, with the number of entries in ClinicalTrials.gov doubling between 2009 and 2014.

While GI disorders account for a large proportion of the global burden of disease, the evidence-based foundation of current clinical practices remains inadequate. For instance, 78% of the current strong recommendations by the American College of Gastroenterology on fourteen diseases based on ranked quality of evidence are only supported by moderate or low quality evidence (where further research is likely or very likely to have an important impact on our confidence in the estimate of effect). This is likely due to a dearth of clinical studies employing sufficient sample sizes, reliable analyses, appropriate and relevant outcomes, and proper oversight [3]. In addition, while observational studies including diagnostic and prognostic trials are not currently required to be registered, the fact that many routine tests and biomarkers with little prior validation or fantastical preliminary success are popularly applied indicates that the scientific process for developing clinical tools along with interventions deserves further scrutiny [4], [5], [6], [7], [8], and [9].

Prior research on GI trials suggest that rigorous trial designs have increased over time but are only applied to a limited number clinical studies [10], [11], [12], [13], [14], [15], [16], [17], [18], and [19]. However, past analyses have been limited in scope and scale due to a focus on published trials in select journals or specific GI disorders. In addition, trial data curated prior to the establishment of standardized trial registries was chiefly obtained from the published corpus, which may be more relevant to understanding the reporting or publishing patterns of GI trial results and designs [13]. Here, we highlight contemporary trends and characteristics of GI trials registered in ClinicalTrials.gov and explore how study properties may relate to design elements that may lead to high quality trials such as randomization, blinding, and the use of data monitoring committees (DMCs). Our systematic approach provides a comprehensive and unbiased perspective on the methodology of GI trials. Lastly, we compare the alignment of clinical research efforts with the burden of GI diseases and non-GI diseases in both the US and worldwide.

2. Methods

125,182 entries with completed data fields registered between January 1, 1997 and September 27, 2013 in ClinicalTrials.gov were downloaded through the Aggregate Analysis of ClinicalTrials.gov portal. Clinical trials were categorized into GI and non-GI trials by examining Medical Subject Heading (MeSH) terms assigned to clinical trials from the condition browse field. MeSH terms encompass descriptors organized in a hierarchy, with broader headings including anatomical regions and diseases. MeSH terms classified within the high level descriptor “digestive system diseases” were manually reviewed by all authors for their pertinence to GI diseases and associated clinical trials were designated as GI trials. All other clinical studies were designated as non-GI trials. MeSH terms are non-redundant and clinical trials linked to multiple MeSH terms were counted towards each disease entity. If a clinical trial was associated with multiple clinical sites, it was counted towards each address. Temporal changes in GI trial characteristics were analyzed after clinical trials were segregated by their starting dates, which corresponded to the first year of actual or planned patient enrollment.

Logistic regression analyses were performed to calculate the odds ratios (ORs) and Wald 95% confidence intervals for clinical trial characteristics associated with the reporting of DMC use, randomization, and double blinding [20]. Logistic regression models included the following variables: sponsors, collaborators, trial phase, purpose, enrollment size, start year, and intervention. A sponsor is designated as the primary organization overseeing the clinical trial, while a collaborator is designated as an organization in a supportive role. Sponsors and collaborators were classified into four categories: National Institutes of Health (NIH), US Fed (non-NIH funding including the Food and Drug Administration, Centers for Disease Control and Prevention, and other US governmental agencies), industry, and other (including universities and charities). Trial purposes were classified into four categories: diagnostic, prevention, treatment, and other (including basic science, education, health services research, screening, and supportive care). Trial interventions were classified into five categories: biologicals and drugs, devices, procedures, dietary supplements and behaviors, and other. Relative risk ratios (RRRs) and Wald 95% confidence intervals were calculated from multinomial regression analyses to determine how sources of sponsoring, enrollment sizes, and phases were associated with the operational status of clinical trials. The sponsor variable used in multinomial regression models was agglomerated with collaborating organizations as previously described for ease of analysis and interpretation [3].

Mortality and DALYs associated with diseases were obtained from the 2010 Global Burden of Disease datasets [21] and [22]. Number of clinical trials, deaths, and DALYs were log transformed prior to analysis given the skewed nature of the data resulting from the extremely high and low prevalence of certain diseases.

SPSS Statistics 20 (IBM) was used for all statistical analyses.

3. Results

The global production of GI trials paralleled the growth of non-GI trials over the last decade but has been static in most recent years (Supplementary Fig. 1). After categorizing clinical trials into relevant diseases based on their associated MeSH terms, we highlight the most popularly represented GI diseases in clinical trials in Table 1. Consistent with their high incidence and extensive burden of disease, viral hepatitides and malignancies were the most common disease terms associated with GI trials. We also examined PubMed primary research articles published in the last ten years for each disease which largely demonstrated a positive correlation between clinical trials and research efforts. Notably, the top five MeSH terms for GI trials accounted for nearly 50% of all clinical trials and publications.

Table 1 Popular GI and GI-related oncologic MeSH terms registered in Clinicaltrials.gov.

All GI terms GI oncology terms
MeSH Clinical trials Publications MeSH Clinical trials Publications
Hepatitis A 2022 (15.9) 1393 (0.6) Colorectal Neoplasms 1964 (31.9) 56,141 (34.4)
Colorectal Neoplasms 1964 (15.4) 56,141 (22.5) Pancreatic Neoplasms 1082 (17.6) 19,994 (12.3)
Hepatitis C 1260 (9.9) 24,302 (9.8) Stomach Neoplasms 777 (12.6) 21,478 (13.2)
Pancreatic Neoplasms 1082 (8.5) 19,994 (8) Carcinoma, Hepatocellular 749 (12.2) 23,355 (14.3)
Stomach Neoplasms 777 (6.1) 21,478 (8.6) Esophageal Neoplasms 559 (9.1) 12,538 (7.7)
Carcinoma, Hepatocellular 749 (5.9) 23,355 (9.4) GIST 152 (2.5) 3478 (2.1)
Hepatitis B 694 (5.5) 14,468 (5.8) Cholangiocarcinoma 125 (2.0) 3267 (2.0)
Cystic Fibrosis 560 (4.4) 7751 (3.1) Gallbladder Neoplasms 96 (1.6) 1820 (1.1)
Esophageal Neoplasms 559 (4.4) 12,538 (5) Duodenal Neoplasms 94 (1.5) 1236 (0.1)
Crohn Disease 516 (4.1) 9325 (3.7) Biliary Tract Neoplasms 87 (1.4) 6812 (4.2)
Liver Cirrhosis 485 (3.8) 17,427 (7) Anus Neoplasms 81 (1.3) 1463 (0.1)
Gastroesophageal Reflux 483 (3.8) 7886 (3.2) Ileal Neoplasms 79 (1.5) 685 (0.5)
Colitis, Ulcerative 376 (3) 6878 (2.8) Jejunal Neoplasms 79 (1.3) 571 (0.0)
Irritable Bowel Syndrome 320 (2.5) 3383 (1.4) Salivary Gland Neoplasms 66 (1.3) 3608 (2.2)
Fatty Liver 266 (2.1) 8885 (3.6) Adenoma, Islet Cell 51 (0.1) 1215 (0.1)
GIST 152 (1.2) 3478 (1.4) Hereditary Nonpolyposis 29 (0.0) 1646 (0.1)
Cholangiocarcinoma 125 (1) 3267 (1.3) Adenomatous Polyposis Coli 29 (0.1) 1766 (0.1)
Hepatic Encephalopathy 118 (0.9) 1562 (0.6) Carcinoma, Islet Cell 19 (0.0) 601 (0.0)
Barrett esophagus 118 (0.9) 3215 (1.3) Gastrinoma 18 (0.0) 262 (0.0)
Gallstones 104 (0.8) 2481 (1) Hepatoblastoma 15 (0.0) 1263 (0.1)

We next examined the geographic location of GI trial sites. Fig. 1A displays a pictorial representation of clinical trial output across all US states showing the dominance of several states, namely California, New York, Maryland, and Texas, in the production of GI clinical research. We suspected that the level of clinical trial output may be aligned with other social dimensions, and indeed clinical research activity correlated linearly with state populations (Fig. 1B and C). The location of clinical trial sites at the national-level was also examined in Fig. 1D. Clinical trials were most often associated with North America, Europe, and China, with nearly 44.5% of GI trials linked to the US, which is comparable to the proportion of all clinical trials (including other medical specialties) attributed to the US. Of note, other clinical trial registries exist and may be more relevant for non-US countries. Therefore, the aforementioned localization of GI trials across the globe requires further clarification in future studies.

gr1

Fig. 1 Geographic distribution of GI trials. (A) Aggregate state-level production of GI trials from 1997 to 2013. (B) State populations in 2013 are portrayed in a heat map. Dark blue and light blue denote the highest and lowest state population levels. (C) Relation between clinical trials and populations among US states. (D) Aggregate country-level production of GI trials from 1997 to 2013

In Table 2, we characterized the properties of GI and non-GI trials that were categorized by their operational status. We defined open studies as recruiting and not yet recruiting trials, completed studies as clinical trials that have ended normally or are active and no longer recruiting, and closed studies as terminated, suspended, or withdrawn trials. Though the number of closed trials only accounted for a small fraction of all registered clinical trials (8.2% for non-GI and 9.1% for GI trials), we hypothesized that the completion of clinical trials may be associated with specific qualities. For example, among closed trials, drugs were the dominant interventions for GI trials (63% vs 43.1% for recruiting and 58.3% for complete) and non-GI trials (61.6% vs 38.2% for recruiting and 52.2% for complete). Treatments followed by preventions were the most common purpose of registered clinical trials, and contributed disproportionately to closed studies in both GI trials (74.6% vs 56.8% for recruiting and 64.5% for complete) and non-GI trials (71.3% vs 55% for recruiting and 63.3% for complete).

Table 2 Characteristics of non-GI and GI clinical trials by status.

Non-GI trials GI trials
Complete Recruiting Closed Complete Recruiting Closed
Interventions Behavioral 5788 (8.3) 3465 (9) 265 (2.9) 253 (3.3) 208 (4) 24 (2)
Biological 5388 (7.7) 1963 (5.1) 587 (6.5) 804 (10.4) 274 (5.2) 109 (9.1)
Device 4951 (7.1) 3678 (9.6) 807 (9) 298 (3.9) 328 (6.3) 72 (6)
Dietary Supplement 1628 (2.3) 932 (2.4) 148 (1.6) 200 (2.6) 192 (3.7) 33 (2.7)
Drug 36,429 (52.2) 14,646 (38.2) 5557 (61.6) 4494 (58.3) 2259 (43.1) 758 (63)
Genetic 316 (0.5) 269 (0.7) 53 (0.6) 42 (0.5) 38 (0.7) 11 (0.9)
Other 3549 (5.1) 3615 (9.4) 334 (3.7) 330 (4.3) 397 (7.6) 38 (3.2)
Procedure 3829 (5.5) 2963 (7.7) 554 (6.1) 551 (7.2) 621 (11.9) 92 (7.6)
Radiation 308 (0.4) 410 (1.1) 54 (0.6) 26 (0.3) 100 (1.9) 9 (0.7)
 
Primary purpose Unreported 13,762 (19.7) 10,245 (26.7) 1380 (15.3) 1366 (17.7) 1332 (25.4) 132 (11)
Basic Science 1450 (2.1) 939 (2.4) 128 (1.4) 78 (1) 74 (1.4) 11 (0.9)
Diagnostic 1626 (2.3) 1400 (3.7) 280 (3.1) 256 (3.3) 263 (5) 52 (4.3)
Services Research 710 (1) 517 (1.3) 42 (0.5) 47 (0.6) 42 (0.8) 6 (0.5)
Prevention 6230 (8.9) 2740 (7.1) 532 (5.9) 740 (9.6) 322 (6.1) 68 (5.7)
Screening 393 (0.6) 338 (0.9) 31 (0.3) 75 (1) 88 (1.7) 2 (0.2)
Supportive Care 1301 (1.9) 1011 (2.6) 185 (2.1) 166 (2.2) 135 (2.6) 32 (2.7)
Treatment 44,226 (63.3) 21,106 (55) 6426 (71.3) 4968 (64.5) 2976 (56.8) 897 (74.6)
 
Phase N/A 23,694 (33.9) 18,773 (49) 2415 (26.8) 2099 (27.3) 2289 (43.7) 283 (23.5)
Phase 0 349 (0.5) 351 (0.9) 49 (0.5) 32 (0.4) 35 (0.7) 4 (0.3)
Phase 1 7846 (11.2) 2748 (7.2) 976 (10.8) 912 (11.8) 407 (7.8) 131 (10.9)
Phase 1/Phase 2 2620 (3.8) 1787 (4.7) 486 (5.4) 369 (4.8) 269 (5.1) 86 (7.1)
Phase 2 14,447 (20.7) 5859 (15.3) 2435 (27) 2033 (26.4) 1024 (19.6) 375 (31.2)
Phase 2/Phase 3 1480 (2.1) 838 (2.2) 222 (2.5) 157 (2) 135 (2.6) 34 (2.8)
Phase 3 11,649 (16.7) 4042 (10.5) 1351 (15) 1398 (18.2) 606 (11.6) 172 (14.3)
Phase 4 7749 (11.1) 3945 (10.3) 1080 (12) 702 (9.1) 471 (9) 118 (9.8)
 
DMC Unreported 23,341 (33.4) 5887 (15.4) 2406 (26.7) 2626 (34.1) 831 (15.9) 360 (29.9)
No 28,941 (41.4) 17,714 (46.2) 3743 (41.5) 2926 (38) 2254 (43) 447 (37.2)
Yes 17,552 (25.1) 14,742 (38.4) 2865 (31.8) 2150 (27.9) 2151 (41.1) 396 (32.9)
 
Gender Unreported 409 (0.6) 649 (1.7) 57 (0.6) 66 (0.9) 112 (2.1) 9 (0.7)
Both 58,755 (84.1) 32,279 (84.2) 7639 (84.7) 7365 (95.6) 4957 (94.7) 1158 (96.3)
Female 6715 (9.6) 3910 (10.2) 947 (10.5) 174 (2.3) 123 (2.3) 27 (2.2)
Male 3955 (5.7) 1505 (3.9) 371 (4.1) 97 (1.3) 44 (0.8) 9 (0.7)
 
Sponsor Industry 26,173 (37.5) 6430 (16.8) 3336 (37) 2877 (37.4) 776 (14.8) 405 (33.7)
NIH 6428 (9.2) 1548 (4) 479 (5.3) 487 (6.3) 154 (2.9) 61 (5.1)
Other 36,030 (51.6) 29,793 (77.7) 5106 (56.6) 4234 (55) 4254 (81.2) 725 (60.3)
U.S. Fed 1203 (1.7) 572 (1.5) 93 (1) 104 (1.4) 52 (1) 12 (1)
 
Collaborator None 45,897 (65.7) 24,180 (63.1) 5936 (65.9) 5182 (67.3) 3467 (66.2) 767 (63.8)
Industry 7070 (10.1) 3400 (8.9) 1252 (13.9) 805 (10.5) 465 (8.9) 176 (14.6)
NIH 6062 (8.7) 2537 (6.6) 695 (7.7) 682 (8.9) 259 (4.9) 113 (9.4)
Other 10,217 (14.6) 7844 (20.5) 1060 (11.8) 1002 (13) 1015 (19.4) 142 (11.8)
U.S. Fed 588 (0.8) 382 (1) 71 (0.8) 31 (0.4) 30 (0.6) 5 (0.4)

A large minority of registered clinical trials failed to report on the use of Data Monitoring Committees (DMCs). Even among clinical trials that reported on the use of DMCs, a majority of clinical studies reported not using DMCs (38% for completed GI trials and 41.4% for completed non-GI trials). The largest proportion of closed trials were in phase 2, accounting for 31.2% of closed GI trials (vs 19.6% for recruiting and 26.4% for completed) and 27% of closed non-GI trials (vs 15.3% for recruiting and 20.7% for completed). Conspicuously, industry and NIH sponsored or funded trials were more prevalent in completed studies (37.4% for GI and 37.5% for non-GI) and closed studies (33.7% for GI and 37% for non-GI).

Temporal trends in clinical trial demographics in interventional GI and non-GI trials are displayed in Table 3. We assessed clinical trials over three specific time periods (from October 2004 to September 2007, October 2007 to September 2010, and October 2010 to September 2013) so that our analysis may be compared to prior studies examining clinical trials from other medical specialties [3]. Of note, most clinical trial characteristics exhibited stable patterns over the time periods analyzed. Exceptions include the fact that drug studies have decreased in both GI and non-GI trials. In contrast, clinical trials involving devices increased over the same time period from 4% to 7% for GI trials and from 7.2% to 11.4% for non-GI trials. Similarly, trials involving dietary supplements increased from 1.3% to 5.4% for GI trials and from 1.1% to 3.9% for non-GI trials. There were also shifts in the primary purpose of clinical trials over time, with a growing fraction of trials associated with basic science, diagnostics, and screening. On the other hand, treatments remain the principal purpose of registered trials, but have declined from 79% to 73% in GI trials and from 79.5% to 70% in non-GI trials.

Table 3 Characteristics of interventional non-GI and GI clinical trials over time periods.

October 2004–September 2007 October 2007–September 2010 October 2010–September 2013
Non-GI GI Non-GI GI Non-GI GI
Interventions Behavioral 2324 (9.9) 105 (3.9) 2819 (8.9) 130 (3.3) 3377 (10.5) 165 (4.1)
Biological 1662 (7.1) 294 (10.9) 2397 (7.6) 397 (10.2) 2203 (6.8) 288 (7.1)
Device 1688 (7.2) 109 (4) 2985 (9.5) 219 (5.6) 3659 (11.4) 284 (7)
Dietary Supplement 252 (1.1) 35 (1.3) 1075 (3.4) 149 (3.8) 1249 (3.9) 221 (5.4)
Drug 15,352 (65.4) 1886 (69.9) 18,008 (57.1) 2334 (60.1) 15,861 (49.3) 2347 (57.7)
Genetic 62 (0.3) 12 (0.4) 83 (0.3) 16 (0.4) 83 (0.3) 5 (0.1)
Other 314 (1.3) 27 (1) 2079 (6.6) 219 (5.6) 3277 (10.2) 278 (6.8)
Procedure 1771 (7.5) 229 (8.5) 1846 (5.8) 365 (9.4) 2154 (6.7) 417 (10.3)
Radiation 60 (0.3) 2 (0.1) 272 (0.9) 54 (1.4) 312 (1) 62 (1.5)
 
Primary purpose Unreported 954 (4.1) 75 (2.8) 1317 (4.2) 153 (3.9) 1686 (5.2) 141 (3.5)
Basic Science 112 (0.5) 9 (0.3) 1040 (3.3) 68 (1.8) 1361 (4.2) 85 (2.1)
Diagnostic 772 (3.3) 104 (3.9) 1100 (3.5) 204 (5.3) 1312 (4.1) 250 (6.1)
Services Research 129 (0.6) 14 (0.5) 515 (1.6) 40 (1) 624 (1.9) 41 (1)
Prevention 2518 (10.8) 302 (11.2) 3136 (9.9) 417 (10.7) 3305 (10.3) 366 (9)
Screening 42 (0.2) 11 (0.4) 134 (0.4) 39 (1) 178 (0.6) 66 (1.6)
Supportive Care 242 (1) 49 (1.8) 875 (2.8) 101 (2.6) 1199 (3.7) 151 (3.7)
Treatment 18,544 (79.5) 2125 (79) 23,447 (74.3) 2861 (73.7) 22,511 (70) 2967 (73)
 
Phase N/A 4322 (18.4) 352 (13) 8264 (26.2) 850 (21.9) 11,060 (34.4) 1045 (25.7)
Phase 0 73 (0.3) 5 (0.2) 258 (0.8) 14 (0.4) 340 (1.1) 44 (1.1)
Phase 1 2185 (9.3) 241 (8.9) 3872 (12.3) 541 (13.9) 3830 (11.9) 491 (12.1)
Phase 1/Phase 2 1118 (4.8) 142 (5.3) 1639 (5.2) 280 (7.2) 1633 (5.1) 230 (5.7)
Phase 2 6010 (25.6) 863 (32) 7181 (22.8) 1044 (26.9) 6188 (19.2) 1049 (25.8)
Phase 2/Phase 3 753 (3.2) 89 (3.3) 854 (2.7) 116 (3) 805 (2.5) 104 (2.6)
Phase 3 5497 (23.4) 697 (25.8) 5073 (16.1) 594 (15.3) 4468 (13.9) 637 (15.7)
Phase 4 3527 (15) 310 (11.5) 4423 (14) 444 (11.4) 3852 (12) 467 (11.5)
 
DMC Unreported 11,762 (50.1) 1374 (50.9) 4225 (13.4) 643 (16.6) 3156 (9.8) 452 (11.1)
No 6154 (26.2) 660 (24.5) 15,684 (49.7) 1662 (42.8) 16,206 (50.4) 1862 (45.8)
Yes 5569 (23.7) 665 (24.6) 11,655 (36.9) 1578 (40.6) 12,814 (39.8) 1753 (43.1)
 
Sponsor Industry 8839 (37.6) 995 (36.9) 11,410 (36.1) 1261 (32.5) 9638 (30) 1128 (27.7)
NIH 1389 (5.9) 124 (4.6) 793 (2.5) 90 (2.3) 586 (1.8) 85 (2.1)
Other 12,726 (54.2) 1533 (56.8) 18,906 (59.9) 2487 (64) 21,587 (67.1) 2822 (69.4)
U.S. Fed 531 (2.3) 47 (1.7) 455 (1.4) 45 (1.2) 365 (1.1) 32 (0.8)
 
Collaborator None 14,687 (62.5) 1691 (62.7) 20,174 (63.9) 2605 (67.1) 21,173 (65.8) 2816 (69.2)
Industry 3200 (13.6) 404 (15) 3917 (12.4) 455 (11.7) 2882 (9) 385 (9.5)
NIH 1934 (8.2) 196 (7.3) 1998 (6.3) 202 (5.2) 1626 (5.1) 158 (3.9)
Other 3458 (14.7) 392 (14.5) 5201 (16.5) 605 (15.6) 6167 (19.2) 688 (16.9)
U.S. Fed 206 (0.9) 16 (0.6) 274 (0.9) 16 (0.4) 328 (1) 20 (0.5)
 
Masking Unreported 505 (2.2) 60 (2.2) 470 (1.5) 93 (2.4) 300 (0.9) 44 (1.1)
Double Blind 8343 (35.6) 789 (29.3) 10,510 (33.3) 1132 (29.2) 10,221 (31.8) 1053 (25.9)
Open Label 12,711 (54.1) 1731 (64.1) 17,146 (54.3) 2394 (61.7) 17,361 (54) 2655 (65.3)
Single Blind 1926 (8.2) 119 (4.4) 3438 (10.9) 264 (6.8) 4294 (13.3) 315 (7.7)
 
Allocation Unreported 2323 (9.9) 305 (11.3) 5797 (18.4) 859 (22.1) 7377 (22.9) 1123 (27.6)
Non-Randomized 5495 (23.4) 775 (28.7) 5212 (16.5) 739 (19) 3507 (10.9) 488 (12)
Randomized 15,667 (66.7) 1619 (60) 20,555 (65.1) 2285 (58.8) 21,292 (66.2) 2456 (60.4)
 
Arms 1 5567 (39) 741 (44.1) 9772 (31.9) 1393 (37.4) 9062 (28.5) 1334 (33.3)
2 6289 (44.1) 668 (39.7) 14,916 (48.7) 1706 (45.7) 16,911 (53.1) 2037 (50.8)
3 1331 (9.3) 161 (9.6) 3226 (10.5) 340 (9.1) 3334 (10.5) 343 (8.6)
≥4 1078 (7.6) 111 (6.6) 2702 (8.8) 290 (7.8) 2540 (8) 297 (7.4)
 
Enrollment 1–50 7954 (35.7) 1029 (39.9) 13,094 (42.5) 1613 (42.7) 12,934 (40.9) 1542 (38.7)
51–100 4170 (18.7) 493 (19.1) 6124 (19.9) 786 (20.8) 7055 (22.3) 898 (22.5)
101–500 7207 (32.3) 742 (28.7) 8797 (28.5) 1089 (28.8) 9030 (28.6) 1252 (31.4)
501–1000 1642 (7.4) 180 (7) 1603 (5.2) 175 (4.6) 1484 (4.7) 180 (4.5)
≥1001 1316 (5.9) 137 (5.3) 1196 (3.9) 116 (3.1) 1114 (3.5) 117 (2.9)

Across all registered trials, relative levels of phase 2 studies remained stable, while phase 1 studies increased from 8.9% to 12.1% for GI trials and from 9.3% to 11.9% for non-GI trials. In contrast, phase 3 trials declined from 25.8% to 15.7% for GI trials and from 23.4% to 13.9% for non-GI trials. Possibly due to the importance of DMCs and emerging institutional policies, the percentage of trials reporting the use of DMCs increased over time. Nevertheless, a large fraction of clinical trials still lacked DMCs, with 45.8% of GI trials reporting the absence of DMCs.

Clinical trials were most often associated with small enrollments, typically restricted to less than 500 patients. In addition, there was a perceptible decline in relative levels of large trials involving greater than a thousand patients (from 5.3% in 2004–2007 to 2.9% in 2010–2013). The role of industry and NIH as a primary sponsor or a collaborator also declined for both GI and non-GI trials (from 36.9% in 2004–2007 to 27.7% in 2010–2013 for industry sponsoring and from 4.6% in 2004–2007 to 2.1% in 2010–2013 for NIH sponsoring).

We performed logistic regression analyses to understand how GI trial characteristics may relate to the use of DMCs, randomization and double blinding (Table 4). Compared to industry sponsored research, NIH sponsored research was more likely to use DMCs. Compared to diagnostic GI trials, treatment and prevention studies were more likely to use DMCs. DMCs were also more likely to be used in clinical trials with larger enrollments, but were not utilized differently between types of interventions.

Table 4 Logistic regression analyses on the use of DMC, randomization, and blinding in interventional GI trials.

Variables DMC Randomization Double blinding
OR (95% CI) p OR (95% CI) p OR (95% CI) p
Sponsor (vs industry)
NIH 7.82 (4.3–14.22) <0.001 0.85 (0.6–1.22) 0.39 0.33 (0.25–0.44) <0.001
U.S. Fed 1.71 (0.91–3.22) 0.096 1.69 (0.77–3.7) 0.191 1.1 (0.6–1.99) 0.767
Other 2.52 (2.22–2.87) 0.227 0.88 (0.76–1.03) 0.116 0.46 (0.4–0.52) <0.001
 
Collaborator (vs none)
Industry 1.12 (0.96–1.31) 0.158 0.85 (0.69–1.03) 0.097 0.89 (0.76–1.06) 0.191
NIH 2.24 (1.73–2.89) <0.001 0.97 (0.7–1.36) 0.864 0.74 (0.56–0.99) 0.042
U.S. Fed 1.22 (0.56–2.68) 0.613 0.87 (0.29–2.58) 0.806 0.75 (0.29–1.99) 0.567
Other 1.83 (1.56–2.13) <0.001 1.09 (0.89–1.33) 0.405 1.01 (0.86–1.19) 0.902
 
Phase (vs phase 1)
Phase 1/Phase 2 1.36 (1.08–1.71) 0.009 1.02 (0.79–1.32) 0.862 1.16 (0.89–1.51) 0.274
Phase 2 1.04 (0.88–1.24) 0.637 1.57 (1.3–1.9) <0.001 1.68 (1.39–2.03) <0.001
Phase 2/Phase 3 1.01 (0.73–1.39) 0.952 7.43 (4.49–12.32) <0.001 2.99 (2.16–4.13) <0.001
Phase 3 0.95 (0.77–1.17) 0.628 6.7 (5.09–8.82) <0.001 2.63 (2.11–3.27) <0.001
Phase 4 0.66 (0.53–0.82) <0.001 2.37 (1.86–3.02) <0.001 1.72 (1.37–2.17) <0.001
 
Purpose (vs diagnostic)
Prevention 2.16 (1.47–3.19) <0.001 5.84 (3.68–9.27) <0.001 3.51 (2.15–5.75) <0.001
Treatment 2.48 (1.76–3.51) <0.001 3.26 (2.2–4.82) <0.001 1.85 (1.16–2.96) 0.01
Other 1.61 (1.03–2.52) 0.036 4.9 (2.81–8.56) <0.001 3.81 (2.2–6.63) <0.001
 
Enrollment (vs >1000)
1 to 50 0.4 (0.27–0.59) <0.001 0.28 (0.17–0.47) <0.001 0.65 (0.47–0.9) 0.008
51 to 100 0.42 (0.28–0.62) <0.001 0.64 (0.38–1.07) 0.086 1.13 (0.83–1.56) 0.438
101 to 500 0.7 (0.48–1.02) 0.06 1.47 (0.88–2.46) 0.139 1.49 (1.1–2.01) 0.009
501 to 1000 0.86 (0.56–1.33) 0.509 2.69 (1.34–5.4) 0.005 1.58 (1.12–2.23) 0.01
Start year 1.03 (1.01–1.05) 0.002 1.1 (1.08–1.12) <0.001 1.02 (1–1.03) 0.047
 
Intervention (vs drug)
Device 1.09 (0.83–1.45) 0.535 0.92 (0.64–1.32) 0.644 0.43 (0.3–0.63) <0.001
Procedure & XRT 1.07 (0.87–1.33) 0.516 1.0 (0.76–1.33) 0.971 0.28 (0.2–0.39) <0.001
Behavioral 0.84 (0.65–1.1) 0.202 3.8 (2.37–6.09) <0.001 3.34 (2.55–4.38) <0.001
Other 0.75 (0.54–1.05) 0.093 1.42 (0.88–2.31) 0.151 0.76 (0.52–1.12) 0.165

Use of randomization did not differ between sponsoring sources, but was more likely among higher phase clinical trials, namely phase 3 studies. Compared to diagnostic GI trials, prevention and treatment studies were more likely to use randomization. Relative to enrollments greater than a thousand patients, randomization was less likely with clinical trials enrolling less than fifty patients but more likely with studies enrolling from 501 to 1000 patients. Compared to drug GI trials, randomization was also significantly more likely among behavioral clinical trials.

Compared to industry sponsored GI trials, studies sponsored by the NIH and other sources were less likely to use double blinding. Relative to phase 1 trials, higher phase clinical trials were more likely to use double blinding. Moreover, prevention and treatment studies were more likely to use double blinding than diagnostic trials. Compared to GI trials recruiting greater than a thousand patients, studies enrolling less than fifty patients were less likely to use double blinding while those enrolling from 101 to 1000 more likely to use double blinding. Compared to drug trials, device and procedure trials were less likely to use double blinding, while behavioral trials were more likely to use double blinding. Notably, the likelihood of using DMCs, randomization, and double blinding increased over time for GI trials.

We next applied multinomial logistic regression analyses to examine how clinical trial characteristics relate to the operational status of clinical trials by comparing completed and actively recruiting trials with closed trials (Supplementary Table 1). Relative to GI trials sponsored by other sources, industry and NIH sponsored trials were less likely to be actively recruiting (RRR, 0.23 [industry]; RRR, 0.29 [NIH]). Relative to enrollment sizes greater than a thousand patients, smaller enrollments were less likely to be found in completed or recruiting trials, with the smallest enrollments associated with the smallest RR. Higher phase trials were also found to be significantly more likely to be closed, i.e. phase 1 studies are more likely than any other phase trial to be completed or active.

Lastly, we inquired whether clinical research efforts for GI diseases are aligned with their burden on human health globally and in high-income North America (NA) which includes the US and Canada. In NA, the correlation between levels of clinical research and mortality or DALYs was greater for GI diseases (Pearson's r, 0.9 [deaths]; Pearson's r, 0.87 [DALYs]) than non-GI diseases (Pearson's r, 0.41 [deaths]; Pearson's r, 0.34 [DALYs]), although all relationships were significant associations (p < 0.001) (Fig. 2). This trend was conserved when we examined relationships between levels of clinical trials and mortality (Pearson's r, 0.82 [GI]; Pearson's r, 0.43 [non-GI]) or DALYs (Pearson's r, 0.76 [GI]; Pearson's r, 0.37 [non-GI]) using worldwide data. Remarkably, among GI diseases, NA deaths accounted for 73% of the variance in NA clinical research efforts while worldwide deaths could only explain 33.6% of the variance in global clinical research efforts. Similarly, NA DALYs explained 81% of the variance in NA GI research efforts while worldwide DALYs explained only 45% of the variance in global GI research efforts.

gr2

Fig. 2 Alignment of disease burden and clinical trials. 19 GI diseases and 126 non-GI diseases with sufficient health burden data were analyzed. Each point represents a disease or syndrome. Lines depict best-fit linear regressions.

4. Discussion

Our systematic characterization of GI trials in ClinicalTrials.gov reveals notable features in the ontogeny of medical knowledge. For example, while the number of registered GI trials has escalated in the past decade, the eventual plateau in growth rates suggests that research efforts have waned in most recent years or that an equilibrium or critical mass of investigators are now abiding with current recommendations regarding the registration of clinical trials. The commonality of MeSH terms associated with clinical trials and publications further denotes the concentration of most research efforts in only a handful of diseases.

An analysis of the geographic localization of GI trials indicates that the spatial production of GI trials largely follows population size and possibly other urban indicators in the US and worldwide. This would be consistent with the notion that most clinical trials originate from academic centers that are concentrated in metropolitans [23]. In support of this, the predominant sponsors of GI and non-GI trials are classified within the other category, which is chiefly comprised of research institutions and universities. In addition to the diminishing sponsorship of clinical trials by the NIH and industry, our findings collectively suggest that the production of knowledge related to GI diseases and their management is increasingly reliant on non-government or non-industry research centers.

The early termination of clinical trials is common, and is secondary to insufficient recruitment, safety or ethical concerns, treatment withdrawal from the market, futility, or in rare cases, benefit [24] and [25]. We demonstrate that closed GI and non-GI trials are constituted by a higher proportion of treatment and drug trials. This is likely related to the stricter standards and monitoring of interventional and pharmaceutical investigations, and is consistent with the increased likelihood of treatment trials to use DMCs (Table 4). It was recently reported that US governmental sponsorship was associated with a lesser likelihood of premature termination due to inadequate enrollments in cardiovascular trials [25]. Although there was a mildly decreased incidence of US governmental sponsorship among closed GI trials in our dataset, a subsequent regression analysis showed no significant association between different sources of sponsorship and early termination. Though our results are not directly comparable as we did not distinguish the cause of closed GI studies, it would be interest for future studies to assess the presence of specialty or disease specific effects on clinical trial completion. Of note, the high proportion of phase 2 trials among closed studies is consistent with high failure rate of trials at phase 2 for investigational drug studies [26], [27], [28], and [29]. However, our regression analysis adjusting for other clinical trial characteristics demonstrates that phase 2, 3, and 4 trials are all less likely to be completed than phase 1 trials.

DMCs are composed of external experts responsible for reviewing accumulating data which may be blinded to investigators as a clinical trial progresses to ensure patient safety. Formally introduced in the 1960s, DMCs have had an essential function in advising trial sponsors and even terminating clinical studies in the face of safety concerns, protocol violations, futility, or benefit [30]. Thus, while it is not practical for all trials to incorporate DMCs, such as in small early phase or exploratory studies with known safety profiles, DMCs are certainly recommended for studies with a priori concerns for patient safety or toxicity, controversial endpoints, at risk or vulnerable target patient demographics, and the involvement of large populations or multiple institutes. Given the overwhelming benefit of DMCs which provide oversight over data integrity and the safety of the human subjects, it is reassuring that GI trials have increasingly used DMCs over the past decade as compared to all other non-GI clinical trials. The likelihood of using randomization and double blinding in GI trials also increased over time in adjusted regression analyses, though a prior analysis of all registered trials indicates that recent trials are no more likely to use randomization [3]. Methodological differences among medical specialties may be due to technical constraints of studying risky interventions or unusual diseases which may preclude blinding or randomization, and the use of alternative statistical analyses or provisions for safety monitoring. Nonetheless, a substantial fraction of interventional GI and non-GI trials lack rigorous trial designs, implying limited reliability, objectivity, and substantiated efficacy in contemporary clinical research.

GI and non-GI trials are generally small, seldom larger than 500 patients in both interventional and observational studies, and characterized by an increasing proportion of clinical investigations enrolling less than fifty patients. While enrollment sizes are optimized to reach sufficient power in clinical studies, smaller trials may be completed faster, reduce cost, and enable the study of rare diseases or unique study populations and individualized therapies, particularly for preliminary investigations and during times of public health urgency [3], [31], [32], and [33]. However, there are inherent shortcomings to small trials including an inability to detect small effect sizes and the production of spurious results which precludes the study of complex diseases and convoluted outcome measures. Systematic reviews and meta-analyses may circumvent the necessity for larger or adequately powered trials, and statistical approaches have been developed to enable the analysis of small trials in aggregate [34]. Nevertheless, we find that certain clinical trial properties are scale dependent, such as the use of DMCs, scrutiny of prevention, and likelihood of study completion decreasing with smaller enrollments. Such limitations to statistical analyses and trial designs may restrict the use of small trials and reduce the degree of certainty on the effect of an intervention.

Our study also serves to highlight the value of increasing transparency in biomedical research. In fact, prior to the creation of clinical trial registries, a systematic evaluation of clinical research efforts would not have been possible. Continual reporting of clinical research efforts including their methodology and results may curb reporting or publication bias and enable the detection of ineffective or corrupt research practices [35], [36], [37], and [38]. While our study was conducted at the level of clinical trials, expanding clinical research transparency to include “raw data” such as patient-level data would have a profound impact on the ability of investigators to validate past results and reuse data for other applications [39], [40], [41], and [42].

The broad scope of this study enabled us to highlight organizational characteristics and methodological trends, including demographic shifts in trial sponsors, the limited use of DMCs, and a preponderance of small enrollments, of the current clinical research enterprise which may have implications on research policies. Congruent with past analyses, our study advocates greater scrutiny of research practices, and raises concerns regarding the ability of contemporary endeavors to yield sufficient authoritative evidence to guide clinical guidelines for GI disorders [3]. Nonetheless, our evaluation of research priorities among diseases demonstrates a greater correlation between public health needs and GI research than with non-GI research. The finding that GI research from NA appears to be better explained by either mortality or DALYs than research worldwide is likely due to greater health care disparities globally and the high health burden of neglected diseases. However, it may also indicate greater sensitivity in the allocation of research funding or execution of research aims to GI disease burdens in NA. Collectively, our findings indicate deficits in both methodological and strategic features of modern research efforts which may impair the ability of future clinical initiatives to generate high quality evidence that is aligned with global health needs and encompasses a diverse portfolio of diseases.

Conflict of interest

None declared.

Appendix A. Supplementary data

The following are the supplementary data to this article:

References

  • [1] D.A. Zarin, T. Tse, N.C. Ide. Trial Registration at ClinicalTrials.gov between May and October 2005. New England Journal of Medicine. 2005;353:2779-2787
  • [2] C. De Angelis, J.M. Drazen, F.A. Frizelle, et al. Clinical Trial Registration: a statement from the International Committee of Medical. New England Journal of Medicine. 2004;351:1250-1251
  • [3] R.M. Califf. Characteristics of Clinical Trials Registered in ClinicalTrials.gov, 2007–2010. Journal of the American Medical Association. 2012;307:1838
  • [4] N. Rifai, P.M. Bossuyt, J.P. Ioannidis, et al. Registering diagnostic and prognostic trials of tests: is it the right thing to do?. Clinical Chemistry. 2014;60:1146-1152
  • [5] A. Colli, M. Fraquelli, G. Casazza, et al. The architecture of diagnostic research: from bench to bedside – research guidelines using liver stiffness as an example. Hepatology. 2014;60:408-418
  • [6] J.P. Ioannidis, O.A. Panagiotou. Comparison of effect sizes associated with biomarkers reported in highly cited individual articles and in subsequent meta-analyses. Journal of the American Medical Association. 2011;305:2200-2210
  • [7] I. Tzoulaki, K.C. Siontis, J.P. Ioannidis. Prognostic effect size of cardiovascular biomarkers in datasets from observational studies versus randomised trials: meta-epidemiology study. British Medical Journal. 2011;343:d6829
  • [8] I. Tzoulaki, A. Seretis, E.E. Ntzani, et al. Mapping the expanded often inappropriate use of the Framingham Risk Score in the medical literature. Journal of Clinical Epidemiology. 2014;67:571-577
  • [9] G.C. Siontis, I. Tzoulaki, P.J. Castaldi, et al. External validation of new risk prediction models is infrequent and reveals worse prognostic discrimination. Journal of Clinical Epidemiology. 2015;68:25-34
  • [10] J.L. Wang, T.T. Sun, Y.W. Lin, et al. Methodological reporting of randomized controlled trials in major hepato-gastroenterology journals in 2008 and 1998: a comparative study. BMC Medical Research Methodology. 2011;11:110
  • [11] Y. Bai, J. Gao, D.W. Zou, et al. Methodological reporting of randomized clinical trials in major gastroenterology and hepatology journals in 2006. Hepatology. 2009;49:2108-2112
  • [12] L.L. Kjaergard, S.L. Frederiksen, C. Gluud. Validity of randomized clinical trials in gastroenterology from 1964–2000. Gastroenterology. 2002;122:1157-1160
  • [13] L.L. Gluud, T.I. Sorensen, P.C. Gotzsche, et al. The journal impact factor as a predictor of trial quality and outcomes: cohort study of hepatobiliary randomized clinical trials. American Journal of Gastroenterology. 2005;100:2431-2435
  • [14] L.L. Kjaergard, C. Gluud. Funding, disease area, and internal validity of hepatobiliary randomized clinical trials. American Journal of Gastroenterology. 2002;97:2708-2713
  • [15] L.L. Kjaergard, D. Nikolova, C. Gluud. Randomized clinical trials in HEPATOLOGY: predictors of quality. Hepatology. 1999;30:1134-1138
  • [16] C. Gluud, D. Nikolova. Quality assessment of reports on clinical trials in the Journal of Hepatology. Journal of Hepatology. 1998;29:321-327
  • [17] X.Q. Li, G.L. Yang, K.M. Tao, et al. Comparison of registered and published primary outcomes in randomized controlled trials of gastroenterology and hepatology. Scandinavian Journal of Gastroenterology. 2013;48:1474-1483
  • [18] F.I. Scott, R.A. McConnell, M.E. Lewis, et al. Increasing complexity of clinical research in gastroenterology: implications for the training of clinician-scientists. American Journal of Gastroenterology. 2012;107:496-500
  • [19] N. Sharma, C. Donnellan, C. Preston, et al. A systematic review of symptomatic outcomes used in oesophagitis drug therapy trials. Gut. 2004;53(Suppl. 4):iv58-iv65
  • [20] A. DasGupta, T.T. Cai, L.D. Brown. Interval estimation for a binomial proportion. Statistical Science. 2001;16:101-133
  • [21] R. Lozano, M. Naghavi, K. Foreman, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380:2095-2128
  • [22] C.J.L. Murray, T. Vos, R. Lozano, et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380:2197-2223
  • [23] L.M.A. Bettencourt, J. Lobo, D. Helbing, et al. Growth, innovation, scaling, and the pace of life in cities. Proceedings of the National Academy of Science. 2007;104:7301-7306
  • [24] A.M. McDonald, R.C. Knight, M.K. Campbell, et al. What influences recruitment to randomised controlled trials? A review of trials funded by two UK funding agencies. Trials. 2006;7:9
  • [25] S. Bernardez-Pereira, R.D. Lopes, M.J.M. Carrion, et al. Prevalence, characteristics, and predictors of early termination of cardiovascular clinical trials due to low recruitment: insights from the ClinicalTrials.gov registry. American Heart Journal. 2014;168:213-219.e1
  • [26] J.A. DiMasi, L. Feldman, A. Seckler, et al. Trends in risks associated with new drug development: success rates for investigational drugs. Clinical Pharmacology & Therapeutics. 2010;87:272-277
  • [27] J.A. DiMasi, J.M. Reichert, L. Feldman, et al. Clinical approval success rates for investigational cancer drugs. Clinical Pharmacology & Therapeutics. 2013;94:329-335
  • [28] M. Hay, D.W. Thomas, J.L. Craighead, et al. Clinical development success rates for investigational drugs. Nature Biotechnology. 2014;32:40-51
  • [29] I. Kola, J. Landis. Opinion: can the pharmaceutical industry reduce attrition rates?. Nature Reviews Drug Discovery. 2004;3:711-716
  • [30] Organization, review, and administration of cooperative studies (Greenberg Report): a report from the Heart Special Project Committee to the National Advisory Heart Council, May 1967. Controlled Clinical Trials. 1988;9:137-148
  • [31] A. Hackshaw. Small studies: strengths and limitations. European Respiratory Journal. 2008;32:1141-1143
  • [32] J.L. Wagner, S.R. Alberts, J.A. Sloan, et al. Incremental costs of enrolling cancer patients in clinical trials: a population-based study. Journal of the National Cancer Institute. 1999;91:847-853
  • [33] S. Yusuf, R. Collins, R. Peto. Why do we need some large, simple randomized trials?. Statistics in Medicine. 1984;3:409-420
  • [34] J. Inthout, J.P. Ioannidis, G.F. Borm. Obtaining evidence by a single well-powered trial or several modestly powered trials. Statistical Methods in Medical Research. 2012;
  • [35] B. Kasenda, E. von Elm, J. You, et al. Prevalence, characteristics, and publication of discontinued randomized trials. Journal of the American Medical Association. 2014;311:1045-1051
  • [36] J.P. Ioannidis. Contradicted and initially stronger effects in highly cited clinical research. Journal of the American Medical Association. 2005;294:218-228
  • [37] J.S. Ross, T. Tse, D.A. Zarin, et al. Publication of NIH funded trials registered in ClinicalTrials.gov: cross sectional analysis. British Medical Journal. 2012;344:d7292
  • [38] J.S. Ross, G.K. Mulvey, E.M. Hines, et al. Trial publication after registration in ClinicalTrials.Gov: a cross-sectional analysis. PLoS Medicine. 2009;6:e1000144
  • [39] F. Koenig, J. Slattery, T. Groves, et al. Sharing clinical trial data on patient level: opportunities and challenges. Biometrical JournalBiometrische Zeitschrift. 2015;57:8-26
  • [40] A.W. Chan, F. Song, A. Vickers, et al. Increasing value and reducing waste: addressing inaccessible research. Lancet. 2014;383:257-266
  • [41] B.L. Strom, M. Buyse, J. Hughes, et al. Data sharing, year 1 – access to data from industry-sponsored clinical trials. New England Journal of Medicine. 2014;371:2052-2054
  • [42] S. Hughes, K. Wells, P. McSorley, et al. Preparing individual patient data from clinical trials for sharing: the GlaxoSmithKline approach. Pharmaceutical Statistics. 2014;13:179-183

Footnotes

a Mount Auburn Hospital, Harvard Medical School, Cambridge, MA, USA

b Northwestern Memorial Hospital, Northwestern University, Chicago, IL, USA

Corresponding author at: Mount Auburn Hospital, Cambridge, MA 02138, USA. Tel.: +1 520 609 0073; fax: +1 617 499 5593.