Primary Hyperparathyroidism from Parathyroid Microadenoma

2
Clifford Ko, MD, FACS Los Angeles, CA Karen Richards, BA Chicago, IL The authors thank Dr Wagner and Dr Schifftner-Smith for their commentary on our examination of missing data within the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP). 1 We ac- knowledge Dr Wagner’s and Dr Schifftner-Smith’s exper- tise and appreciate their contributions to the understand- ing of this topic as well as to the broader evaluation of quality within health care. We also agree that further dis- cussion of these topics is warranted and will be valuable, and regret that there are many additional issues around the topic of missing data, including additional investigations such as those noted by Dr Wagner and Dr Schifftner- Smith, which were beyond the scope of our brief article. As Dr Wagner and Dr Schifftner-Smith pointed out, the terminology of missingness is exacting, and despite this there has been some variability in usage across authors and disciplines. We apologize if our attempts to clarify and emphasize fell short of the mark in any way. We agree that we presented an emphasis distinguishing between the rela- tionships of missingness to independent and dependent variables that, although typical of some earlier applications, does not generalize across all situations. 2 Because we dem- onstrated a relationship (though imperfect) between albu- min missingness and outcomes, the data are not likely miss- ing completely at random. Regarding an inquiry by Dr Wagner and Dr Schifftner-Smith, we do know that miss- ingness cannot be predicted perfectly based on the ob- served variables in the NSQIP data, including outcomes (data not previously shown for the sake of brevity). It there- fore remains possible that data are missing in a partially at-random fashion (an imperfectly covariate-dependent fashion), and that there is a not-at-random aspect. 3,4 Given these issues, we would like to respectfully re- emphasize for the reader the primary conclusions from our work, which remain true: 1. The single imputation approach traditionally used within the NSQIP under-represents the uncertainty of the imputation performed; a multiple imputation ap- proach can better represent this uncertainty. 2. In the case of albumin within the NSQIP, missingness is not independent of outcomes: patients with known al- bumin values have greater likelihood of adverse out- comes. The missingness of albumin is not random in relation to outcomes. 3. Altering the statistical approaches to handling of miss- ing data (among the models we present for the NSQIP data set) has demonstrable effects on coefficients gener- ated for the albumin covariate, but less impressive ef- fects on predictions of hospital morbidity or mortality, or risk-adjusted hospital performance assessments. We could not agree more with our colleagues that addi- tional analyses, including sensitivity analyses, will be valu- able and should be continued. For instance, although we were optimistic about the value of a missingness indicator alone, we recognize that improper reliance on an indicator can result in bias, and that because multiple imputation remains likely to have a low risk of bias, it is probably the approach of choice. 3-6 Still, these issues will certainly ben- efit from further investigation, and we look forward to pursuing these topics along with Dr Wagner and Dr Schifftner-Smith. We thank them again for expanding the discussion of this important subject. REFERENCES 1. Hamilton BH, Ko CY, Richards K, Hall BL. Missing data in the American College of Surgeons National Surgical Quality Im- provement Program are not missing at random: implications and potential impact on quality assessments. J Am Coll Surg 2010; 210(2):125–139. 2. Ambler G, Omar RZ, Royston P. A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome. Stat Methods Med Res 2007;16:277– 298. 3. Janssen KJ, Donders AR, Harrell FE Jr, et al. Missing covariate data in medical research:To impute is better than to ignore. J Clin Epidemiol 2010;63:721–727. 4. Knol MJ, Janssen KJ, Donders AR, et al. Unpredictable bias when using the missing indicator method or complete case analysis for missing confounder values: an empirical example. J Clin Epide- miol 2010;63:728–736. 5. Donders AR, van der Heijden GJ, Stijnen T, Moons KG. Review: a gentle introduction to imputation of missing values. J Clin Epidemiol 2006;59:1087–1091. 6. Kenward MG, Carpenter J. Multiple imputation: current per- spectives. Stat Methods Med Res 2007;16:199–218. Disclosure Information: Nothing to disclose. Primary Hyperparathyroidism from Parathyroid Microadenoma Ritesh Agrawal, MS Sudhi Agarwal, MS Anjali Mishra, MS Gaurav Agarwal, MS 436 Letters J Am Coll Surg

Transcript of Primary Hyperparathyroidism from Parathyroid Microadenoma

Page 1: Primary Hyperparathyroidism from Parathyroid Microadenoma

CL

KC

TtwQktiqcatsS

ttdewtvdomiWis(faf

ew

1

2

3

tawacraepSd

R

1

2

3

4

5

6

D

PP

RSA

436 Letters J Am Coll Surg

lifford Ko, MD, FACS

os Angeles, CA

aren Richards, BA

hicago, IL

he authors thank Dr Wagner and Dr Schifftner-Smith forheir commentary on our examination of missing dataithin the American College of Surgeons National Surgicaluality Improvement Program (ACS NSQIP).1 We ac-

nowledge Dr Wagner’s and Dr Schifftner-Smith’s exper-ise and appreciate their contributions to the understand-ng of this topic as well as to the broader evaluation ofuality within health care. We also agree that further dis-ussion of these topics is warranted and will be valuable,nd regret that there are many additional issues around theopic of missing data, including additional investigationsuch as those noted by Dr Wagner and Dr Schifftner-mith, which were beyond the scope of our brief article.

As Dr Wagner and Dr Schifftner-Smith pointed out, theerminology of missingness is exacting, and despite thishere has been some variability in usage across authors andisciplines. We apologize if our attempts to clarify andmphasize fell short of the mark in any way. We agree thate presented an emphasis distinguishing between the rela-

ionships of missingness to independent and dependentariables that, although typical of some earlier applications,oes not generalize across all situations.2 Because we dem-nstrated a relationship (though imperfect) between albu-in missingness and outcomes, the data are not likely miss-

ng completely at random. Regarding an inquiry by Dragner and Dr Schifftner-Smith, we do know that miss-

ngness cannot be predicted perfectly based on the ob-erved variables in the NSQIP data, including outcomesdata not previously shown for the sake of brevity). It there-ore remains possible that data are missing in a partiallyt-random fashion (an imperfectly covariate-dependentashion), and that there is a not-at-random aspect.3,4

Given these issues, we would like to respectfully re-mphasize for the reader the primary conclusions from ourork, which remain true:

. The single imputation approach traditionally usedwithin the NSQIP under-represents the uncertainty ofthe imputation performed; a multiple imputation ap-proach can better represent this uncertainty.

. In the case of albumin within the NSQIP, missingness isnot independent of outcomes: patients with known al-bumin values have greater likelihood of adverse out-comes. The missingness of albumin is not random inrelation to outcomes.

. Altering the statistical approaches to handling of miss- G

ing data (among the models we present for the NSQIPdata set) has demonstrable effects on coefficients gener-ated for the albumin covariate, but less impressive ef-fects on predictions of hospital morbidity or mortality,or risk-adjusted hospital performance assessments.

We could not agree more with our colleagues that addi-ional analyses, including sensitivity analyses, will be valu-ble and should be continued. For instance, although weere optimistic about the value of a missingness indicatorlone, we recognize that improper reliance on an indicatoran result in bias, and that because multiple imputationemains likely to have a low risk of bias, it is probably thepproach of choice.3-6 Still, these issues will certainly ben-fit from further investigation, and we look forward toursuing these topics along with Dr Wagner and Drchifftner-Smith. We thank them again for expanding theiscussion of this important subject.

EFERENCES

. Hamilton BH, Ko CY, Richards K, Hall BL. Missing data in theAmerican College of Surgeons National Surgical Quality Im-provement Program are not missing at random: implications andpotential impact on quality assessments. J Am Coll Surg 2010;210(2):125–139.

. Ambler G, Omar RZ, Royston P. A comparison of imputationtechniques for handling missing predictor values in a risk modelwith a binary outcome. Stat Methods Med Res 2007;16:277–298.

. Janssen KJ, Donders AR, Harrell FE Jr, et al. Missing covariatedata in medical research: To impute is better than to ignore. J ClinEpidemiol 2010;63:721–727.

. Knol MJ, Janssen KJ, Donders AR, et al. Unpredictable bias whenusing the missing indicator method or complete case analysis formissing confounder values: an empirical example. J Clin Epide-miol 2010;63:728–736.

. Donders AR, van der Heijden GJ, Stijnen T, Moons KG. Review:a gentle introduction to imputation of missing values. J ClinEpidemiol 2006;59:1087–1091.

. Kenward MG, Carpenter J. Multiple imputation: current per-spectives. Stat Methods Med Res 2007;16:199–218.

isclosure Information: Nothing to disclose.

rimary Hyperparathyroidism fromarathyroid Microadenoma

itesh Agrawal, MS

udhi Agarwal, MS

njali Mishra, MS

aurav Agarwal, MS

Page 2: Primary Hyperparathyroidism from Parathyroid Microadenoma

AASGL

Wctmdt

1

2

3

4

5

R

1

2

D

R

MNNPEP

WmtpIp

1

2

437Vol. 211, No. 3, September 2010 Letters

mit Agarwal, MS

shok K Verma, MS

aroj K Mishra, MS

yan Chand, MS

ucknow, India

e read the article by Goasguen and colleagues1 and dis-ussed it with interest in our journal club. We congratulatehe authors for rejuvenating the concept of parathyroidicroadenomas. However, there were certain queries raised

uring our discussion and it would be fruitful if we get aimely reply.

. How many frozen-section biopsies were sent from asingle gland and were the authors able to differentiatebetween hyperplasia and microadenoma during sur-gery? Did all 4 glands undergo routine frozen-sectionbiopsy when microadenoma was suspected or found? Ifnot, on what criteria were the abnormal glands identi-fied? How did they differentiate the patient having 2microadenomas from a patient having parathyroidhyperplasia?

. The study groups are too old to recommend any fruitfulresult, especially in an era when parathyroid surgeonsroutinely perform sestamibi and neck ultrasonographyfor localization of parathyroid adenoma. Why was acomparison done between the microadenoma and ade-noma groups operated in 2004? Would it not have beenbetter if a comparison was done between the microad-enoma and adenoma groups operated in or after 2005,which underwent localization with ultrasonography orsestamibi scintigraphy or both? That would have given abetter recognition to the study. Why did they discussstudies of sestamibi scintigraphy and ultrasonographyfor localization when they had not done it during thestudy period? These were not relevant to the presentstudy.

. What was the cause of increased osteoporosis and neph-rolithiasis in the microadenoma group? How many mi-croadenomas were asymptomatic on presentation andwhat was their management policy?

. Because the authors have not compared results of pre-operative serum calcium and serum parathormone lev-els in the microadenoma and adenoma groups duringmultivariate analysis, it seems the authors were biasedabout inclusion of these criteria.

. The authors quoted Agarwal and associates,2 statingthat a single positive preoperative localizing examina-tion, mainly sestamibi scintigraphy, was sufficient forminimally invasive parathyroidectomy. However, no

such recommendation was made in that study.

EFERENCES

. Goasguen N, Chirica M, Roger N, et al. Primary hyperparathy-roidism from parathyroid microadenoma: specific features andimplications for a surgical strategy in the era of minimally invasiveparathyroidectomy. J Am Coll Surg 2010;210:456–462.

. Agarwal G, Barraclough BH, Robinson BG, et al. Minimallyinvasive parathyroidectomy using the ‘focused’ lateral approach.I. results of the first 100 consecutive cases. ANZ J Surg 2002;72:100–104.

isclosure Information: Nothing to disclose.

eply

ircea Chirica, MD

icolas Goasguen, MD

icolas Munoz-Bongrand, MD

ierre Cattan, MD

mile Sarfati, MD

aris, France

e thank Chand and colleagues for their thoughtful com-ents on our article “Primary Hyperparathyroidism Due

o Parathyroid Microadenoma: Specific Features and Im-lications for a Surgical Strategy in the Era of Minimallynvasive Parathyroidectomy.”1 We provide a point-by-oint answer to the queries raised.

. The diagnosis of microadenoma was made by the macro-scopic assessment of the involved gland by one of our se-nior surgeons. Intraoperative frozen sections of the patho-logical gland and of a biopsy of a macroscopically normalgland were the subjects of a comparative analysis. A highercellular density, lower content in stromal fat cell, and theabsence of finely dispersed intracytoplasmic fat dropletslead to the diagnosis of adenoma. Findings of normal para-thyroid tissue in the biopsy ruled out hyperplasia, but ab-normal findings led to frozen-section examination of theother glands. The only patient reported in our experiencewith 2 microadenomas underwent frozen-section biopsyof the 2 remaining glands.

. These interesting remarks do not really apply to the presentstudy. As stated, the aim of the study was to describe pre-operative and intraoperative characteristics of patients withparathyroid microadenoma, with particular attention ac-corded to the intraoperative difficulties in identifying mi-croadenoma, rather than the evaluation of the results ofminimally invasive parathyroidectomy for microadenoma(which might actually be the object of some future study).

For this reason, and to avoid confusion arising from the