Annual intravenous zoledronic acid for three years increased cancellous bone matrix mineralization beyond normal values in the HORIZON biopsy cohort



The efficacy of 3 years of annual intravenous administration of zoledronic acid (ZOL) in reducing vertebral and nonvertebral fractures in postmenopausal osteoporosis has been shown by the HORIZON pivotal fracture trial. Histomorphometric analysis of transiliac bone biopsies from the HORIZON participants revealed significantly improved trabecular architecture and reduced bone remodeling for the ZOL-treated versus placebo-treated patients. The aim of our study was to evaluate the cancellous and cortical bone mineralization density distribution (BMDD) in these biopsies by quantitative backscattered electron imaging (qBEI). The study cohort comprised 82 patients on active treatment (ZOL, yearly doses of 5 mg) and 70 treated with placebo, and all received adequate Ca and VitD supplementation. Comparison of ZOL-treated versus placebo-treated cancellous (Cn.) and cortical (Ct.) BMDD-derived variables resulted in significantly higher average (Cn.CaMean + 3.2%, Ct.CaMean + 2.7%) and mode calcium concentrations (Cn.CaPeak + 2.1%, Ct.CaPeak + 1.5%), increased percentages of highly mineralized bone areas (Cn.CaHigh + 64%, Ct.CaHigh + 31%), lower heterogeneity of mineralization (Cn.CaWidth −14%, Ct.CaWidth −13%), and decreased percentages of low mineralized bone areas (Cn.CaLow −22%, Ct.CaLow −26%) versus placebo (all p < 0.001). Cn. BMDD from the patients on active treatment also revealed a statistically significant shift to higher Ca concentrations when compared to a historical normal reference BMDD. These differences in BMDD from ZOL patients compared to the other groups were in line with the correlation of BMDD variables with previously determined cancellous mineralizing surface per bone surface (Cn. MS/BS, a primary histomorphometric index for bone turnover), showing that those with lower Cn. MS/BS had a higher degree of bone matrix mineralization. However, the differences in BMDD variables between the study groups remained when adjusted for Cn. MS/BS, suggesting that other factors in addition to reduced bone turnover might contribute to the higher bone matrix mineralization after ZOL treatment. © 2013 American Society for Bone and Mineral Research.


The efficacy of 3 years of once-yearly infusions of zoledronic acid (ZOL) in reducing the risk of hip, vertebral, and other fractures was shown by the HORIZON pivotal fracture trial in postmenopausal osteoporotic patients.1, 2 The former was accompanied by a beneficial effect on bone mineral density (BMD) as shown by BMD increases at different skeletal sites.1, 3, 4 In a subgroup of HORIZON participants, transiliac bone biopsy samples were obtained after 3 years of treatment. Recker and colleagues5 conducted histomorphometric analysis of these biopsy samples and reported higher bone volume in the ZOL-treated patients versus placebo. As expected, bone turnover indices were reduced in ZOL-treated patients compared to placebo-treated ones, and this reduction was comparable to the findings with other bisphosphonates.5

In the present work, we aimed to obtain information on the bone mineralization density distribution (BMDD) by quantitative backscattered electron imaging (qBEI). BMDD is one important determinant of bone material properties, essentially influencing its stiffness and hardness, thus contributing to whole bone mechanical features such as Young's modulus and ultimate strength. Among other techniques used for analysis of bone matrix mineralization, qBEI measures BMDD with high spatial resolution and high sensitivity to differences in calcium concentrations.6 In a healthy adult population, we have previously observed that cancellous BMDD was independent of gender, ethnicity, age, or skeletal site. Using the results from this healthy population, a reference BMDD for trabecular bone has been established,7 which has been used for comparison with BMDD in disease and/or after treatment.6 Previous works have demonstrated that BMDD is essentially influenced by mineralization kinetics and bone turnover (the determinant of average bone tissue age).6, 8, 9 The term “kinetics of mineralization” refers to the speed of mineral accumulation within a newly formed bone packet after the onset of mineralization (and is NOT equal to the mineral apposition rate, which describes the speed of propagation of the mineralization front). Consequently, because BMDD depends on mineralization kinetics and bone turnover, deviations from normal BMDD indicate alterations in one or in both of these processes. For instance, the high bone turnover with a negative bone balance in postmenopausal osteoporosis not only decreases the bone volume but also lowers average tissue age and therefore bone matrix mineralization density.10, 11 The goal of therapies (such as bisphosphonates, teriparatide, strontium ranelate) is to stop bone loss and/or to increase bone volume. Depending on their anabolic and/or antiresorptive effects on bone, these therapies result in specific changes in BMDD.10–17

In this work, we studied the effect of ZOL on bone matrix mineralization in both the spongiosa and cortex of transiliac bone biopsy samples from the HORIZON pivotal fracture trial. BMDD from patients on 3 years of active treatment were compared to those treated with placebo, and to those from a healthy reference population (published previously7). Additionally, BMDD outcomes from all patients were correlated with cancellous mineralizing surface/bone surface (Cn. MS/BS, a measure of bone turnover) from a previous study.5

Materials and Methods

Bone biopsy samples

Transiliac bone biopsy samples from the HORIZON pivotal fracture trial5 were kindly provided by Novartis, Pharma AG, Basel, Switzerland. Biopsies were collected from a total of 152 participants (82 ZOL and 70 placebo). To be enrolled in the study, the patients had to have a T-score < –2.5 or a T-score ≤ −1.5 with evidence for at least one vertebral fracture. The patients were classified into two strata (stratum I, without concomitant treatment; stratum II, with concomitant osteoporosis therapy, such as hormone replacement, selective estrogen receptor modulators, etc.). All patients received either yearly doses of active drug (5 mg) or placebo plus daily calcium (1.0–1.5 g) and vitamin D (400–1200 IU). Patients previously treated with parathyroid hormone (PTH), fluoride, or strontium were excluded. Bisphosphonates were either excluded or included if there was an adequate time period without bisphosphonates before study entry; 2 years (if used for more than 48 weeks), 1 year (if used for more than 8 weeks but less than 48 weeks), and 6 months (if used for more than 2 weeks but less than 8 weeks). A minor reduction in bone turnover markers before ZOL treatment started was tolerated. For more details see Recker and colleagues.5


BMDD of the entire trabecular and entire available cortical bone compartment of the transiliac bone biopsy samples was determined using qBEI. For full details see previous works.6, 18 Prior to analysis, the undecalcified iliac bone samples were embedded in polymethylmethacrylate. The resulting surfaces of the block samples were ground, polished, and carbon-coated before they were analyzed for BMDD in a digital scanning electron microscope (DSM 962; Zeiss, Oberkochen, Germany). Digital images from the samples were obtained using the following microscope settings: an accelerating voltage of 20 kV, a probe current of 110 pA, and a working distance of 15 mm. Backscattered electrons were collected by a four-quadrant semiconductor backscattered electron detector. The entire bone tissue area was recorded in a series of images of the same size (2 mm × 2.5 mm) at × 50 nominal magnification (corresponding to a resolution of 4 µm per pixel) using a scan speed of 100 seconds per frame. The intensity of the signal of the backscattered electrons (the different gray levels in the image) is mainly dependent on the local calcium concentration of the sample (the higher the calcium concentration the brighter the pixel gray level in the image; see Fig. 1A). The digital images for cancellous and cortical bone were used to construct gray-level histograms, which were further transformed to weight percent Ca histograms (so called BMDD) with a bin width of 0.17 weight percent Ca (Fig. 1B). Five variables were evaluated to characterize the BMDD7: CaMean, the weighted mean Ca concentration of the bone area; CaPeak, the peak position of the histogram, which indicates the mode (most frequently occurring) calcium concentration of the studied bone area; CaWidth, the full-width at half-maximum of the distribution, describing the variation in mineralization density; CaLow, the percentage of mineralized bone with a calcium concentration less than the 5th percentile of the reference BMDD7 (less than 17.68 weight percent calcium), which reveals the amount of bone area undergoing primary mineralization, and CaHigh, the portion of bone areas with a calcium concentration higher than the 95th percentile (higher than 25.30 wt% Ca) of the reference BMDD, predominantly representing interstitial bone.

Figure 1.

(A) Backscattered electron image of the cross-sectional area of one ZOL treated biopsy sample (the brighter in the image the higher the local calcium concentrations). (B) Corresponding gray-level histogram (so called bone mineralization density distribution, BMDD) of trabecular bone from the biopsy sample above (solid line), shown together with one example of the cancellous BMDD of the placebo-treated (dashed line) and the 95% confidence interval of normal cancellous BMDD curves (gray area, published previously7). Measured BMDD variables are drawn: the average (CaMean) and the mode (the most frequent) Ca concentrations (CaPeak), the heterogeneity of mineralization (CaWidth), the percentages of low (primary) mineralized (CaLow) and highly mineralized bone areas (CaHigh).

Histomorphometric MS/BS

Histomorphometric analysis on these bone biopsy samples from the HORIZON study has been described.5 In the present work, Cn. MS/BS was considered as a primary index for bone turnover and used for correlation with BMDD outcomes. Cn. MS/BS was not available for the entire study cohort but 59 patients on active treatment and 52 treated with placebo had intact biopsy samples with sufficient tetracycline labels within trabecular bone to permit assessment of this histomorphometric measure.

Statistical analysis

Statistical analysis was done using Sigma Stat for Windows Version 2.03 (Systat Software, San Jose, CA, USA). Data are presented as mean (SD) or as median (25th, 75th percentile) (the latter if the normality test failed). Comparison between the groups (ZOL versus placebo, ZOL versus normal reference, or placebo versus normal reference, entire study groups and within each stratum separately) is based on t tests or Mann-Whitney rank sum tests (the latter were used if normality or equal variance tests failed). Analysis of the relationship between MS/BS and BMDD outcomes was based on Spearman rank order correlation. Analysis of covariance was done using SPSS version 19 (trial version) (IBM Software, Ehningen, Germany). Statistical significance was assumed when p < 0.05.


ZOL versus placebo

Two typical examples for BMDD curves are shown in Fig. 1B. The results for all measured BMDD variables are summarized in Fig. 2. In both cancellous (Cn.) and cortical (Ct.) bone, the mean (Cn.CaMean + 3.2%, Ct.CaMean + 2.7%) and the mode calcium concentrations (Cn.CaPeak + 2.1%, Ct.CaPeak + 1.5%) were increased for ZOL versus placebo. The heterogeneity of mineralization (Cn.CaWidth −14%, Ct.CaWidth −13%) and the percentage of low mineralized bone areas (Cn.CaLow −22%, Ct.CaLow −26%) were decreased, whereas the portion of highly mineralized bone areas (Cn.CaHigh + 64%, Ct.CaHigh + 31%) was increased for ZOL versus placebo.

Figure 2.

Cancellous (A) and cortical (B) BMDD outcomes from the entire study groups (white placebo, back-white patterned ZOL-treated). The bars show mean (SD) or median (25th, 75th percentile). White dotted lines and gray areas in the background of A are revealing the mean/median and ± 1 SD/interquartile range of the cancellous reference BMDD (from Ref. 17), cortical reference BMDD data are not available yet. ***p < 0.001 versus placebo treatment; °°°p ≤ 0.001, °p < 0.05 versus normal reference BMDD.

Similar significant differences were observed when the patients from the two strata were considered separately (Table 1).

Table 1. Cn. and Ct. BMDD Variables For Stratum I and Stratum II patients Separately
 Stratum IStratum II
ZOL n = 55PLA n = 51p-valueZOL n = 27PLA n = 19p-value
  1. Data are mean (SD) or median (25th, 75th percentiles).

Cn.CaMean [wt%]23.03 (22.71; 23.34)22.23 (21.62; 22.81)< 0.00122.99 (0.53)22.30 (0.53)< 0.001
Cn.CaPeak [wt%]23.59 (0.45)23.12 (0.60)< 0.00123.65 (0.53)23.12 (0.47)0.001
Cn.CaWidth [Δwt%]3.29 (2.95; 3.47)3.81 (3.47; 4.16)< 0.0013.29 (3.12; 3.47)3.64 (3.47; 3.99)< 0.001
Cn.CaLow [%]4.07 (3.72; 4.64)5.75 (4.50; 7.17)< 0.0014.18 (3.75; 5.06)4.96 (4.44; 6.35)0.006
Cn.CaHigh [%]13.95 (6.23)9.64 (5.89)< 0.00115.06 (8.58)9.03 (3.26)0.006
Ct.CaMean [wt%]22.95 (0.62)22.43 (0.54)< 0.00122.94 (0.52)22.14 (0.72)< 0.001
Ct.CaPeak [wt%]23.57 (23.27; 23.92)23.22 (22.88; 23.40)< 0.00123.61 (0.50)22.93 (0.65)< 0.001
Ct.CaWidth [Δwt%]3.59 (0.37)4.04 (0.47)< 0.0013.47 (3.29; 3.81)4.33 (3.99; 4.81)< 0.001
Ct.CaLow [%]3.85 (1.11)5.12 (1.53)< 0.0013.96 (3.26; 4.45)5.96 (3.75; 7.26)0.003
Ct.CaHigh [%]12.95 (10.19; 17.35)9.82 (6.47; 12.66)0.00214.70 (6.85)9.74 (6.07)< 0.05

Comparison to normal reference BMDD

Additionally, the cancellous BMDD data were compared with historical reference BMDD for cancellous bone (published previously7) (Fig. 2). The patients on active treatment had higher Cn.CaMean ( + 3.7%, p < 0.001), Cn.CaPeak ( + 2.8%, p < 0.001), and Cn.CaHigh ( + 194%, p < 0.001), as well as lowered Cn.CaLow (−8.8%, p < 0.05) compared to reference BMDD, whereas Cn.CaWidth was normal. Patients treated with placebo revealed normal Cn.CaMean, slightly increased Cn.CaPeak ( + 0.7%, p < 0.05), and highly increased Cn.CaWidth ( + 17%, p < 0.001), Cn.CaLow ( + 17%, p < 0.001), and Cn.CaHigh ( + 79%, p < 0.001).

Relationship of Cn. BMDD variables with Cn. MS/BS in a subgroup of our study population

From a large part of our study cohort (n = 52 placebo, n = 59 ZOL), histomorphometric Cn. MS/BS (a primary index of bone formation) was available. For testing the relationship between Cn. BMDD variables and Cn. MS/BS, we conducted correlation analysis. Significant negative correlations were found for Cn.CaMean (Fig. 3), Cn.CaPeak, and Cn.CaHigh with Cn. MS/BS (r = −0.45, −0.31, −0.32, respectively, p < 0.001). Cn.CaWidth and Cn.CaLow were positively correlated with Cn. MS/BS (r = 0.58, 0.47, respectively, p < 0.001). Based on these relationships, we tested whether the Cn. BMDD differences between ZOL versus placebo were still present when they were adjusted for Cn. MS/BS. Analysis of covariance (ANCOVA) (using Cn. MS/BS as covariate) showed that the significant differences remained after adjustment for Cn. MS/BS (BMDD outcomes of this subgroup analysis is shown in Table 2).

Figure 3.

The average calcium concentration of cancellous bone (Cn.CaMean) plotted versus Cn. MS/BS.5 Full symbols indicate ZOL, empty symbols placebo treatment.

Table 2. Subgroup Analysis of HORIZON Participants: Estimated Marginal Means of Cn. BMDD Variables After Adjustment for Cn. MS/BS
 After Adjustment for Cn. MS/BS
ZOL (n=59)placebo (n=52)ZOL vs. PLA (%)p
  1. Data are estimated marginal means (standard errors).

Cn.CaMean [wt%]22.92 (0.08)22.45 (0.08)+ 2.1< 0.001
Cn.CaPeak [wt%]23.57 (0.07)23.25 (0.08)+ 1.40.008
Cn.CaWidth [Δwt%]3.29 (0.05)3.69 (0.05)−11< 0.001
Cn.CaLow [%]4.38 (0.16)5.21 (0.17)−160.001
Cn.CaHigh [%]13.92 (0.94)10.60 (1.01)+ 320.029


In this work, we present the BMDD for cancellous and cortical bone from the entire transiliac bone biopsy sample cohort of the HORIZON pivotal fracture trial.

Effect of ZOL treatment versus placebo on BMDD

Generally, we observed a shift to higher calcium concentrations in the ZOL patients when compared to placebo. This is in agreement with the histomorphometric findings of reduced bone turnover.5 Because the BMDD is essentially influenced by bone turnover rates, reductions result in less new bone being formed (leading to a lower percentage of bone areas undergoing primary mineralization), which, coupled with a deceleration of resorption, results in existing bone remodeling units having increased mineral content due to the increased time available for secondary mineralization. The resulting shift of the BMDD to higher concentrations in comparison to placebo was previously also found for postmenopausal osteoporotic patients treated with other bisphosphonates, such as alendronate and risedronate.10–14 Additionally, in this study we observed that the effect of ZOL on BMDD in comparison to placebo was similar in cancellous and cortical bone. This is important to note because cortical bone represents the majority of load-bearing bone in the skeleton, and is generally considered to be metabolically less active than trabecular bone. However, significant effects on cortical bone have been previously observed as well in short-term ZOL-treated liver transplantation patients.19

Such differences between the BMDD from patients on active treatment in comparison to those treated with placebo could also be observed when the two strata were considered separately. This indicates that the concomitant therapies in patients of stratum II did not significantly influence the action of ZOL versus placebo.

The differences between groups in the present study were not observed in a previously reported study on material density using micro–computed tomography (µCT).5 This discrepancy is likely explained by the differences in the techniques used—qBEI in the present study and µCT in the previous study—because spatial resolution and sensitivity for detection of differences in degree of mineralization are much higher for qBEI than for µCT laboratory devices.

Comparison to normal reference BMDD

Most interestingly, the ZOL-treated patients had a higher degree of cancellous bone matrix mineralization compared to reference cancellous BMDD variables. It is noteworthy that the reference healthy population measured for BMDD had a larger age-range, comprised women and men, and included cancellous bone from different skeletal sites. This normal reference BMDD should therefore be considered as a “more general” control group and not specifically for the HORIZON participants, who were postmenopausal osteoporotic women. Apart from that fact, the increase in average and mode calcium concentration after ZOL treatment beyond normal values is in agreement with the lower bone turnover in the treated patients (median Cn. MS/BS was 0.45%5) compared to that considered normal (7.0% ± 4.1%, mean ± SD) as published previously.20 Interestingly, the present study population differs somewhat from previous study populations of bisphosphonate treatments in terms of BMDD. Although the absolute bone turnover rate in the present ZOL-treated cohort is very similar to those reported for the alendronate phase III study21 (Cn. MS/BS 0.25% and 0.62% for the higher doses of alendronate), the ZOL-treated patients reveal a clear shift in BMDD to higher calcium concentrations when compared to the patients treated with alendronate.10 In this context it is also interesting that the ZOL-treated patients revealed an increase in bone volume as measured histomorphometrically.5, 22 This would be expected to result in a decrease of average matrix mineralization as a higher percentage of younger bone packets with a lower degree of mineralization is present, which for instance, is seen after PTH treatment.16 However, the contrary is seen after ZOL treatment, namely an increase in bone matrix mineralization. This is a further indication that the kinetics of mineralization (the time course of increase in mineral in a bone packet formed during treatment) might be accelerated in ZOL treatment. This view is also supported by Raman measurements in newly formed bone packets in the ZOL-treated patients.23 Moreover, the currently clinically used bisphosphonates may have different courses of action in addition to bone turnover suppression, due to the different modifications they exert on apatite crystallite surface properties once adsorbed onto them.24 However, it remains unclear whether the BMDD outcomes are related to differences in the action of ZOL in comparison to other bisphosphonates.2, 22

Another possible contribution to the elevated mineralization in ZOL treatment could be that the baseline characteristic in the ZOL study is different from the alendronate study. Whereas in the alendronate phase III trial the patients had lowered bone matrix mineralization at baseline,10 the present study population might have already been at normal bone mineralization levels as a result of adequate supplementation with calcium and vitamin D and due to other treatments (such as hormone replacement therapy [HRT], selective estrogen receptor modulators [SERMs], etc.). Although we have no information on the baseline status in our study population because no baseline biopsies were obtained, our view of normal bone mineralization at baseline is strongly supported by the findings in the placebo group. The placebo-treated has normal, or even slightly increased average mode calcium concentrations compared to normal BMDD. And interestingly, the average Cn. MS/BS in the placebo group was somewhat lower than normal.

Relationship between BMDD variables and bone turnover rates

The aforementioned findings of differences in BMDD between patients on active treatment versus placebo, and also the differences between the study groups and normal are in line with the general correlation of BMDD variables with MS/BS. As expected, those patients with relatively lower Cn. MS/BS have relatively higher BMDD and lower heterogeneity of bone matrix mineralization in our cohorts. Similar correlations between the latter histomorphometric index and the BMDD variables have been observed for other patient cohorts25, 26 and in healthy controls.26 However, to test whether the differences in BMDD outcomes between ZOL and placebo are due to bone turnover only, we analyzed the BMDD variables using MS/BS as a covariate. Our findings revealed that the significant differences remained after adjustment for Cn. MS/BS. This suggests that there are other factors in addition to bone turnover which contribute to the increased bone matrix mineralization after ZOL treatment. There is rising evidence that the osteocytes play a key role in bone and mineral homeostasis by producing specific factors such as sclerostin, phosphateregulating gene with homologies to endopeptidases on the X chromosome (PHEX), matrix extracellular phosphoglycoprotein (MEPE), dentin matrix protein 1 (DMP-1), fibroblast growth factor 23 (FGF23), DMP-1, MEPE, FGF-23, etc.27, 28 Because a recent report showed that ZOL can have antiapoptotic effects on osteocytes,29 further direct effects of ZOL on osteocytic activity cannot be excluded, which might lead to higher matrix mineralization as well as to an increase in the bone volume, as has been observed. As mentioned before, Raman microspectroscopic analysis revealed increased mineral to matrix ratio in the newly formed bone (near the mineralization front) indicative of alterations in the bone matrix after ZOL treatment.23 However, definite conclusions from the subgroup analysis of Cn. MS/BS relation to BMDD in the present study cannot be drawn. It must be noted that adjustment for Cn. MS/BS was not possible for the entire study groups because Cn. MS/BS was not available for all biopsy samples. Further, difficulties and imprecision in the measurement of Cn. MS/BS might occur when bone turnover is highly reduced and bone formation sites are rare events in the sampling area.30

Increased bone matrix mineralization and reduced fracture risk

A patient's fracture risk is, in general, dependent on several factors such as the amount and architecture of bone tissue and the quality of the bone material, although bone mineral density (BMD by densitometry) is considered an important element of fracture risk. In the HORIZON study, BMD after 3 years of treatment was significantly increased for ZOL-treated versus placebo-treated at various skeletal sites.1, 3, 4 In addition to the previously reported increase in bone volume,5 the higher calcium concentrations in the patients on active treatment, as found in the present work, contribute to this increased BMD. However, the effect of increased bone matrix mineralization on the intrinsic mechanical properties of the bone tissue remains unclear. On one hand, an increase in the form might raise stiffness and strength, whereas on the other it might also increase the “brittleness” of the material. The observed fracture risk reduction in the patients on active treatment1 supports the view that the increase in material density we observed has no adverse effect on the material quality, at least in the time frame of the study. Further, the fracture risk reduction was sustained, and no cases of atypical fracture were reported in the long-term 6-year ZOL treatment (HORIZON extension study).31


ZOL treatment for 3 years increased bone matrix mineralization in agreement with the reduction in bone turnover on treatment compared to placebo and to normal. The latter might be due to the fact that the HORIZON study population had already normal BMDD at baseline and were not typical high-turnover osteoporotic patients. However, subgroup analysis revealed that the reduction in bone turnover alone did not fully explain the difference in BMDD in the study groups, indicating that other factors (such as an accelerated mineral accumulation in the newly formed bone matrix) might be contributing to the increase in bone matrix mineralization during ZOL treatment.


RRR has received institutional grant/research support from, and is a paid consultant for Merck, Lilly, Pfizer, Procter & Gamble, Amgen, Roche, and Novartis. JAG is an employee of the Novartis Institutes for BioMedical Research, Basel, Switzerland. EFE has been an employee of and owns stock in Novartis and consults with and speaks for Novartis. All other authors state that they have no conflicts of interest.


This study was supported by the AUVA (Austrian Social Insurance for Occupational Risk), the WGKK (Social Health Insurance Vienna). We thank Petra Keplinger, Sonja Lueger, and Phaedra Messmer for careful sample preparation and qBEI measurements at the bone material laboratory of the Ludwig Boltzmann Institute of Osteology, Vienna, Austria. We thank Novartis Pharma AG, Basel, Switzerland, for providing the biopsy samples.

Authors' roles: All authors made substantial contributions to either the conception and design, acquisition of data, or analysis and interpretation of data, participated in drafting the manuscript or revising it critically for important intellectual content, and approved the final version of the submitted manuscript. BMM accepts responsibility for the integrity of the data analysis.